Construction and Building Materials 60 (2014) 8–16

Contents lists available at ScienceDirect

Construction and Building Materials

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

Development of high performances concrete based on the addition

of grinded dune sand and limestone rock using the mixture design
modelling approach
R. Zaitri a, M. Bederina a,⇑, T. Bouziani a, Z. Makhloufi a, M. Hadjoudja b
a
Structures Rehabilitation and Materials Laboratory, University A. T. of Laghouat, Algeria
b
Civil Engineering Laboratory, University A. T. of Laghouat, Algeria

h i g h l i g h t s

 Dune sand and limestone rock were grinded in order to obtain economical mineral additions.
 SF and LF additions were added by substitution to the cement in order to formulate HPC.
 A mixture design modelling approach was used to highlight the effect of each addition.
 Statistical models predicting fresh and hardened HPC properties were established as function of C, SF and LF.
 Ternary contour plots were used to provide flexibility in optimizing C, SF and LF proportions.

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the results of an experimental program in which the mixture design modelling
Received 2 December 2013 approach was used in order to optimize the composition of a high performance concrete (HPC) formu-
Received in revised form 24 February 2014 lated from local materials of the area of Laghouat (Algeria). The valorised materials are mineral additions:
Accepted 25 February 2014
siliceous and limestone additions, obtained by fine grinding of dune sand and sieving aggregate crushing
Available online 19 March 2014
waste respectively. Both additions are added by substitution to the cement content. In order to model the
influence of cement content and the dosages of these additions on the properties of high performance
Keywords:
concrete, an experimental plans method with three factors was used. Mathematical models that explain
Valorisation
Dune sand
at best test results which were identified and developed in this work. The modelling was performed
Limestone fillers thanks to statistical analysis software JMP7 of Statistical American Systems SAS-Institute. The obtained
High performance concrete results showed that the introduction of siliceous fillers (SF) and limestone fillers (LF) in cement (by sub-
Mixture design approach stitution) leads to a significant improvement in mechanical strength in the medium and long term which
Rheology allows us to formulate a high performance concrete. In addition, the workability has been also improved
Mechanical properties with this substitution. Moreover, it was also shown that the use of the derived models based on the
experimental design approach is very interesting and helps us to understand the interactions between
the different parameters of the studied mixture (cement content, SF-content and LF-content). Finally it
should be noted that not only the physic-mechanical properties are encouraging, but the economic aspect
is also very interesting.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction good durability. The American Concrete Institute (ACI) defined

High performance concrete (HPC) is a specialized series of con-
cretes designed to provide several benefits in the construction of
concrete structures such as the high mechanical strength and the
⇑ Corresponding author. Address: Civil Engineering Department, University of
Amar Telidji, BP 37G Laghouat 03000, Algeria. Tel.: +213 773732230.
E-mail addresses: mdbederina@yahoo.com, m.bederina@mail.lagh-univ.dz
(M. Bederina).
HPC as ‘‘concrete which meets special performance and uniformity
requirements that cannot always be achieved routinely by using
only conventional materials and normal mixing, placing and cure’’
[1]. Indeed, HPC presents characteristic strengths at 28 days great-
er than 50 MPa and W/C ratio lower than 0.4 [2–4]. Its aggregates
constitute a conventional granular skeleton embedded in a matrix
of increased compactness thanks to the use of superplasticizer
(water-reducing) and ultrafine crystalline or amorphous products
which play both roles, as granular and pozzolanic complement

http://dx.doi.org/10.1016/j.conbuildmat.2014.02.062
0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16 9

[5]. In addition to Siliceous fume previously recognized as the best Table 2

in this role, other ultrafine (slag, metakaolin, fly ash, . . .) can be Physico-mechanical properties of the used cement.

used [6]. Currently, the application of HPC in buildings becomes
increasingly large due to its interesting properties regarding the
compressive strength, tensile strength, ease of execution and
implementation. However, the economical aspect is not com-
pletely absent [7]. Indeed, the current trend in the world is to find
materials of different origins and compositions that can meet well-
defined performance and minimizing cost [8]. It is in this aspect
that we think about the use of the dune sand. The latter is a mate-
rial of high availability in many countries around the world. Algeria
is one of those countries where a large part of its territory is cov-
ered with this material. In fact, the use of dune sand in civil engi-
neering has great economic and environmental interest. This
material has been the subject of several research topics in recent
decades [8,9]. As a mineral addition, it has been the subject of only
some research topics which showed that the dune sand, in the
form of mineral fines, is able to develop a certain pozzolanic reac-
tion in cementitious environment [10,11].
For this purpose, it is advisable to add local raw materials that
can achieve the same performance as those presented by Silica
fume. Therefore, we proceeded to the use of grinded dune sand
and limestone fines coming from crushing waste (abundant mate-
rials, and therefore relatively inexpensive) as additives in the con-
text of minimizing the cost of HPC. Moreover, and to carry out the
composition of the mixture studied, we used a new technique
called ‘‘ experimental plans method ‘‘ that permits both to stream-
line the test program and to empirically model the responses ob-
tained according to the study parameters as well [12]. Statistical
modelling approaches are commonly used to identify the relative
significance of primary mixture parameters and their coupled ef-
fects on relevant properties of concrete [13–16]. The aim of this
work is to study the effect of the partial substitution of cement
with Siliceous fillers (dune sand finely grinded (SF)) and limestone
fillers (LF) (obtained from aggregates crushing waste) on the prop-
erties of HPC. This will select the most efficient optimal couples
‘‘SF/LF’’, both in terms of mechanical strength that sustainability
point of view. For this, and as it has already been noted, we use
the theory of mixture design modelling that enables us to prioritize
influential parameters and to quantify the effects.
2. Experiments and methods
2.1. Materials
Besides the basic components (gravel, sand, cement, admixtures and water),
HPC typically includes one or more additions [5]. In this study, dune sand (siliceous)
was finely grinded and crushed waste (limestone) was sifted in order to incorporate
the obtained fines as mineral additions in HPC. All used materials are characterized
hereafter:
The used cement is a white blended Portland cement of type CPJ-CEM II/A 52.5
(initial limestone addition of about 12%) having high initial and final strengths. The
cement factory conforms to the Algerian standard NA 442 [17] (which is mainly
based on the European Standard EN 197-1). The chemical, mineralogical and phys-
ico-mechanical properties of this cement are given in Tables 1 and 2. The used addi-
tives are mineral additions of two different natures: siliceous fillers (SF) and
limestone fillers (LF) (Fig. 1). SF are obtained by finely grinding dune sand. The latter
is extracted from Algeria’s northern region, near the city of Djelfa (Algeria) and fea-
turing a maximum grain diameter of approximately 0.60 mm; the proportion of
grains smaller than 0.08 mm is below 5%. The obtained fines are intended to be
added by substitution to the cement. The sand is washed in order to remove any
impurities, then dried and finally finely grinded until obtaining the desired grain
size. The grinding was performed at dry state using a ball mill of large capacity
Characteristics White cement 52.5
Specific density (kg/m ) 3
3120
Apparent density (kg/m3) 1180
Specific surface area (m2/kg) 495
Normal consistency (%) 25
Initial set 1 h 45 mn
Final set 2 h 60 mn
Compressive strength (MPa) 2 days 30.69
28 days 57.51
which was pushed until obtaining a powder passing through a sieve of 80 lm.
The specific density of the obtained powder (SF) is 2.6 g/cm3. The mineralogical nat-
ure of ‘‘SF’’ is mainly siliceous which is shown in Fig. 2. LF are obtained by sifting (a
sieve opening of 0.08 mm) crushing waste from a quarry located in the north region
of Laghouat. The chemical analysis shows that these fillers are mainly composed of
limestone (Fig. 3). Their specific density is of 2900 kg/m3 and their specific surface is
of 312 m2/kg. The used sand is a river sand (RS) [18] which presents a continuous
particle size distribution ranging from 0.08 to 5 mm with a fraction of grains smal-
ler than 0.08 mm below 2%. Its particle size distribution is almost fully enveloped
by the limit curves recommended for concrete (Fig. 4). In a schematic manner, RS
grains present rounded shapes (Fig. 5). Table 3 lists the set of physical characteris-
tics for this sand and EDX analysis of sands demonstrates its essentially siliceous
nature (Fig. 6). The used gravel in this work is natural crushed gravel coming from
the deposit located in the area of Laghouat. Two granular fractions were used: Grav-
el (3/8) and Gravel (8/15). Their particle size distributions are showed in Fig. 7 and
their physical properties are listed in Table 3. Their chemical analysis is equally the
same as that of limestone fillers (Fig. 3). The admixture used in this study is an Alge-
rian superplasticiser, high water-reducer of third generation for ready concretes
called MEDAFLOW 30. It should be noted that the total volume of mixing water
is the volume of water to which must be added the equivalent quantity of water
provided by the plasticizer [19].
2.2. Mixture design approach
The mixture design approach consists of statistical methods that can be used to
better organize experimental tests [20]. It can be applied in many disciplines and in
all industries where several factors ‘‘xi’’ (in the case of concrete, the proportions of
individual component materials) influence one or more performance characteris-
tics, or responses ‘‘Y’’ (the fresh and hardened properties of the concrete). It is there-
fore very interesting to use this method when we have to study a function of type:
Y ¼ f ðxi Þ ð1Þ
Indeed, when we use this technique, a maximum of information is obtained
with a minimum of experiences. To do this, we have to follow mathematical rules
and adopt rigorous steps. There are many experimental plans suitable for all the
cases which can be encountered by an experimenter [21]. This method is based
on two main concepts, the experimental space and the mathematical modelling
of studied variables. The studied factors are the proportions of components of the
mixture [22]. However, these components are not independent of each other and
the proportions of a mixture must sum to 100%. The percentage of the last compo-
nent is imposed by the percentages of the first components. This is why the mixture
plans are treated aside. The mixture plans are also characterized by many con-
straints that may influence the choice of the proportions of the constituents. For
example, the concentration of a product must be at least x per-cent or the concen-
tration may not exceed a given value. Based on these constraints, the planning
study is modified and must be adapted to each case. If we denote by xi the content
of component i, the sum of the concentrations of all components of the mixture sat-
isfies the relation:
X
i¼n
xi ¼ 100% ð2Þ
i¼1
If the sum of the concentrations of various constituents is brought back to the
unit ‘‘1’’, the Eq. (2) becomes:
X
i¼n
xi ¼ 1 ð3Þ
i¼1

Table 1
Chemical and mineralogical characteristics of the cement.

Sio2 Al2O3 Fe2O3 CaO MgO SO3 Na2O K2O LIO C3S C2S C3A C4AF
23.5 3.3 0.22 63.7 0.7 2.2 0.4 0.5 4.7 55.00 23.49 8.37 0.67
10 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16

Fig. 1. General aspect of grinded dune sand and used limestone fillers.

Fig. 4. Particle size distribution of used river sand.

Fig. 2. EDX analysis of milled dune sand.

Fig. 3. EDX analysis of limestone fillers. Fig. 5. Scanning electron micrographs of river sand (general aspect of grains) [17].

2.3. Mixture proportions

This relationship is called the fundamental constraint of mixtures. The geo-
metrical representations of mixture designs are different from those used for
In this work, we are particularly interested in high performance concrete of
conventional experimental designs and mathematical models are also pro-
class I which has a compressive strength between 50 and 75 MPa. Our experimental
foundly modified. For the geometrical representation of three-component mix-
program aims to understand the role played by the various parameters separately
tures, an equilateral triangle is used. The pure products are at the vertexes of
(content of cement (C), content of grinded dune sand (SF) and limestone fillers
the equilateral triangle. Binary mixtures are represented by the sides of the tri-
(LF)) on the properties of high performance concrete. Thus, we have to consider a
angle. For example, the left side of the triangle (Fig. 8) shows the mixtures
mixture design with three factors (C, SF and LF) taken in mass proportions, whose
composed only of products A and B. The points situated on the interior of
sum is equal to unity with a substitution of cement up to 25% (125 kg of cement)
the equilateral triangle represent a tertiary mixture. The compositions of each
(Fig. 9). This means that these factors are dependent on each other. The experimen-
product can be read on the triangle sides. The geometrical properties of the
tal field is constrained by the following equation:
equilateral triangle insure that the fundamental constraint of mixtures is
respected. C þ SF þ LF ¼ 1 ð4Þ
 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16 11

Table 3
Physical properties of used river sands.

Apparent density (kg/m3) Specific density (kg/m3) Fineness modulus Compactness (CP) Sand equivalent
RS 1549.7 2580.6 2.65 57 88
G(3/8) 1358.5 2666.7 – – –
G(8/15) 1480.9 2666.7 – – –

With three factors and five levels, a mixture design comprising 21 experiments
was prepared in order to evaluate the influence of these factors on the characteris-
tics of HPC. The model can be written under the form f = (xi) where the function ‘f’ is
a polynomial development of an order greater or lesser of xi, the order of the poly-
nomial depends on the desired degree of precision. In this study, a third-order poly-
nomial model was used with three non-independent variables (C, SF and LF) and
five levels. The model is expressed as:

Y ¼ a1  ðCÞ þ a2  ðSFÞ þ a3  ðLFÞ þ a4  ðC  SFÞ þ a5  ðC  LFÞ þ a6  ðSF  LFÞ

ð6Þ

Fig. 6. EDX analysis of river sand [17].
Fig. 7. Particle size distribution of used gravels (3/8 and 8/16).
where Y is the response and (a1, a2, a3, a4, a5, a6) are the model coefficients.
The model coefficients identification is required; they express the effect of the
response to each factor and each interaction. Fig. 10 shows the triangular lattice of
the 21 studied combinations [23] and the proportions of the experience factors gen-
erated by the software are shown in Table 4. In our case, the desired responses are
the slump flow (Abrams cone), the compressive strength and the flexural strength.
The obtained results of responses are then injected into the software ‘‘JMP7’’. They
are represented by ternary diagrams connecting the components of the matrix (ce-
ment, grinded dune sand and limestone fillers) with the studied responses that
highlighted the predominant matrix components. The same diagrams have also en-
abled us to optimize the composition of reference HPC. The method used for deter-
mining the basic composition is the HPC formulation of the university of
Sherbrooke [24]. This method allows the formulation of HPC without entraining
air, and follows the same principle of ACI 211 standard. 1-91, 1991 [25]. The cor-
rected method of formulation of ordinary concrete of Dreux [4] was also used for
the pre-formulation of HPC, in particular for the determination of the granular com-
position of coarse aggregate. It is a simple method that needs only the determina-
tion of the particle size distribution curves of aggregates. The basic composition is
kept constant for all tests thereafter, except for the binder content which is com-
posed of a ternary mixture (C, LF and SF). The binder content, which is considered
as ternary mixture, was developed based on these three components for all studied
compositions with proportions resulting from the use of statistical software JMP7.
The proportions of granular classes (3/8) and (8/15) were respectively 25% and 75%
in total volume of coarse aggregate. The content of sand was adjusted for the differ-
ent compositions to obtain a cubic meter of concrete according to the following
formulas:

V paste ¼ V cement þ V fines þ V coarse aggregate þ V water þ V admixture þ V air ð7Þ

V fine aggregate ¼ 1000  V p^ate ð8Þ

The different compositions are summarized in Table 5 according to their order

of preparation in the laboratory after randomization tests.

3. Results and discussion

The results of the 21 mixtures, statistically balanced with the

three factors (C, SF and LF) can fully be exploited in the develop-
ment of mathematical models describing the effect of the type of
addition on the properties of HPC. These models facilitate the
assessment of the effect of each addition separately and in combi-
nation with the other types of addition on the properties of HPC by
using ternary diagrams with iso-response curves. Moreover, an
analysis of variance allows the dissociation of the different studied
effects starting from the variance of the measured response and al-
Fig. 8. Representation of three-component mixtures on an equilateral triangle [17]. lows us to see the contribution of factors on the variability of the
response. The validity of the models can be tested by the statistical
method based on the calculation of errors coming from the exper-
The number of experiments generated by the program is calculated by the fol-
iment and from the model. Most adapted models relatively present
lowing equation:
high correlation coefficients. This finding shows a good correlation
ðq þ m  1Þ! between the results obtained by simulations and the values pre-
Cm
qþm1 ¼ ð5Þ
ðmÞ!ðq  1Þ!
dicted by the model found thereby. The obtained test results are
where q is the number of factors and m is the number of levels. collected in Tables 6 and 7.
12 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16

Cement
125 SF
kg/m3
SF

500kg/m3
of cement 375kg/m3
of cement

(a) Initial cement (b) Substituted cement

Fig. 9. Substitution of cement by mineral additions.

3.1. Mechanical strength

The compressive strength is the most important property which

defines the use of materials. In addition, its improvement is more
spectacular. However, the compressive strength is generally a
characteristic often used to classify the HPC. The derived statistical
models for the compressive strength at 7 days and 28 days are
summarized below with variables expressed in coded values. The
models reflect the relative significance of each parameter on the
studied response (compressive strength).

Rc7 ¼ 43:13  C þ 31:96  SF þ 34:47  LF þ C  ðSF  21:17Þ

þ C  ðLF  26:31Þ ð9Þ

Rc28 ¼ 63:94  C þ 45:78  SF þ 47:94  LF þ C  ðSF

 49:50Þ þ C  ðLF  30:90Þ ð10Þ

where Rc7 and Rc28 are the compressive strengths at 7 and 28 days
respectively.
From these derived models, it can be noticed that the compres-
sive strength is first conditioned by increasing the cement content,
followed by the coupled effect of cement and grinded dune sand
Fig. 10. Triangular representation of the 21combinaisons studied with three (SF), then the content of limestone filler (LF) and finally the cou-
components [22]. pled effect of cement and limestone filler (LF). It can be also noted,
that in the first few days of hydration, the mechanical strength
develops faster in all studied HPC. Indeed, at 7 days, the control
HPC (HPC 21) reaches 68% of its compressive strength at 28 days.
Table 4 The compressive strength of HPC with SF/LF develops a rate which
Factors proportions in the studied compositions. varies from 60% to 78% depending on the percentage of substitu-
No Cement SF LF
tion. The combined effects of the addition of SF and LF offer to
the high performance concrete an improved behaviour according
1 0 0 1
to the rate of substitution of cement. The dosages of the three fac-
2 0 0.2 0.8
3 0 0.4 0.6 tors have optimum values (respectively around 60% of C, 20% of SF
4 0 0.6 0.4 and 20% of LF in substituted volume of 125 kg of cement) for which
5 0 0.8 0.2 the compressive strength reaches a maximum value. This finding
6 0 1 0
reflects the chemical role of the LF and confirms the pozzolanic
7 0.2 0 0.8
8 0.2 0.2 0.6 activity of SF at long term [9,10,26]. Beyond 7 days, the cement re-
9 0.2 0.4 0.4 leases lime during its hydration, the SF fixes the lime to form new
10 0.2 0.6 0.2 silicates that lead to significant increases in concrete strength over
11 0.2 0.8 0 time. Thus, the optimal strength is due to the physical role of the LF
12 0.4 0 0.6
which offers improved compactness to the high performance con-
13 0.4 0.2 0.4
14 0.4 0.4 0.2 cretes [27]. In fact, LF can change the pore structure of concrete and
15 0.4 0.6 0 reduce the number of pores. This change depends on fillers fine-
16 0.6 0 0.4 ness, the higher the filler particles, the more effective is their role
17 0.6 0.2 0.2
[28]. A decrease in compressive strength is observed when the per-
18 0.6 0.4 0
19 0.8 0 0.2 centage of substitution additions increases which can be explained
20 0.8 0.2 0 by the following observation: when the percentage of substitution
21 1 0 0 of cement by additions increases, the amount of cement decreases,
the resulting products fall, therefore, the amount of Ca(OH)2 with
 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16 13

Table 5
Different studied compositions.

Mixture no Cement (kg/m3) SF (kg/m3) LF (kg/m3) G (3/8) (kg/m3) G(8/15) (kg/m3) Water (l/m3) W/L SP (%) Sand (kg/m3)
HPC 1 375 + 0 0 125 262.5 787.5 150 0.3 10 720.92
HPC 2 375 + 0 25 100 262.5 787.5 150 0.3 10 720
HPC 3 375 + 0 50 75 262.5 787.5 150 0.3 10 719.08
HPC 4 375 + 0 75 50 262.5 787.5 150 0.3 10 718.16
HPC 5 375 + 0 100 25 262.5 787.5 150 0.3 10 717.25
HPC 6 375 + 0 125 0 262.5 787.5 150 0.3 10 716.33
HPC 7 375 + 25 0 100 262.5 787.5 150 0.3 10 724.14
HPC 8 375 + 25 25 75 262.5 787.5 150 0.3 10 723.22
HPC 9 375 + 25 50 50 262.5 787.5 150 0.3 10 722.3
HPC 10 375 + 25 75 25 262.5 787.5 150 0.3 10 721.38
HPC 11 375 + 25 100 0 262.5 787.5 150 0.3 10 720.46
HPC 12 375 + 50 0 75 262.5 787.5 150 0.3 10 727.35
HPC 13 375 + 50 25 50 262.5 787.5 150 0.3 10 726.44
HPC 14 375 + 50 50 25 262.5 787.5 150 0.3 10 725.52
HPC 15 375 + 50 75 0 262.5 787.5 150 0.3 10 724.6
HPC 16 375 + 75 0 50 262.5 787.5 150 0.3 10 730.57
HPC 17 375 + 75 25 25 262.5 787.5 150 0.3 10 729.65
HPC 18 375 + 75 50 0 262.5 787.5 150 0.3 10 728.73
HPC 19 375 + 100 0 25 262.5 787.5 150 0.3 10 733.8
HPC 20 375 + 100 25 0 262.5 787.5 150 0.3 10 732.87
HPC 21 375 + 125 0 0 262.5 787.5 150 0.3 10 737

Table 6
Results of characterization tests. which the LF and SF enters in reaction, decreases, which was
resulting in a decrease in the amount of C–S–H. Moreover, the
Mixture no Slump flow (cm) Compressive strength (MPa)
amount of free Ca(OH)2 produced by the hydration of the cement
At 7 days At 28 days
(which contains 67.7% of CaO (Table 1)) is consumed by the reac-
Rf RC Rf RC tion with the SF (SiO2) resulting in the addition of more strength-
HPC 1 7.5 3.18 34.17 3.71 49.12 ening C–S–H phase [11]. Indeed, the main constituent of LS is
HPC 2 8 3.27 35.15 3.78 50.08 CaCO3 which has an accelerating effect on C3S and cement hydra-
HPC 3 7.5 3.21 35.04 3.7 44.95 tion leads to the precipitation of calcium carbosilicate hydrate (C–
HPC 4 8.5 3.14 33.49 3.68 46.04
HPC 5 9 3.11 31.87 3.74 48.41
S–H), which may also contribute to void filling [29–32]. The SF
HPC 6 8.5 3.09 32.22 3.66 47.49 therefore participates by SiO2. Compared to LF, the increase in
HPC 7 8.5 3.49 38.67 3.9 54.72 compressive strength at long term due to both the consumption
HPC 8 10 3.37 38.51 4 55.26 of Ca(OH)2 and the demonstrated optimum proportions of mineral
HPC 9 9 3.5 40.12 3.98 58.03
additions, suggest that both SF and LF can be utilized as partial ce-
HPC 10 9 3.3 36.25 3.81 52.01
HPC 11 10 3.40 37.52 3.87 56.94 ment replacement materials. Depending on the obtained results of
HPC 12 10.5 3.66 43.51 4.23 58.68 the compressive tests, we were able to build ternary graphs show-
HPC 13 12.5 3.67 44.25 4.23 64 ing the effects of the 03 constituents of the studied response
HPC 14 10.5 3.41 39.83 3.92 60.71 (Fig. 11). Concerning the flexural strength, exactly as in the com-
HPC 15 9.5 3.52 40.75 4.13 62.88
HPC 16 14 4.08 45.42 4.6 67.02
pression case, the factors obtained in models at 7 days and 28 days
HPC 17 12 4.22 49.51 4.71 73.19 are very significant and can be written as follow:
HPC 18 11 3.82 46.22 4.38 71.46
HPC 19 13 3.92 47.04 4.45 65.57
Rf 7 ¼ 3:71  C þ 3:08  SF þ 3:17  LF þ C  ðSF  1:08Þ þ C
HPC 20 10.5 3.76 41.67 4.31 68.43  ðLF  1:74Þ ð11Þ
HPC 21 9 3.53 42.18 4.15 61.17

Table 7
Model parameters estimates of studied responses.

Slump flow Rf7 Rf28 Rc7 Rc28

R2 = 0.78 R2 = 0.81 R2 = 0.84 R2 = 0.89 R2 = 0.91
P = 0.0001 P < 0.0001 P < 0.0001 P < 0.0001 P < 0.0001
Coeff. p-value Coeff. p-value Coeff. p-value Coeff. p-value Coeff. p-value
a1: (C) 10.20 <0.001⁄ 3.70 <0.001⁄ 4.29 <0.001⁄ 43.12 <0.001⁄ 63.93 <0.001⁄
a2: (SF) 8.65 <0.001⁄ 3.05 <0.001⁄ 3.64 <0.001⁄ 31.30 <0.001⁄ 46.25 <0.001⁄
a3: (LF) 6.69 <0.001⁄ 3.15 <0.001⁄ 3.60 <0.001⁄ 33.81 <0.001⁄ 48.41 <0.001⁄
a4: (LF⁄SF) 3.42 0.3035 1.10 0.0592 0.91 0.0871 21.99 0.0046 48.90 0.0002⁄
a5: (C⁄FC) 15.92 0.0002⁄ 1.76 0.0052⁄ 1.81 0.0025⁄ 27.13 0.0009⁄ 30.30 0.0088⁄
a6: (SF⁄FC) 1.48 0.6498 0.11 0.8347 0.07 0.8881 4.92 0.4686 3.45 0.7299

R2: correlation coefficient.

P: probability.
p-value: is small enough to indicate very convincing significance. P-values are the probability of getting an even more extreme statistic given the true value being tested is at
the hypothesized value, usually at zero.
Coeff.: the estimated coefficient in the linear model.
*
marks the probability values validated by the used software.
14 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16

Fig. 11. Iso-response curves of the compressive strength.

Fig. 12. Iso-response curves of the flexural strength.

Rf 28 ¼ 4:30  C þ 3:65  SF þ 3:62  LF þ C  ðSF  0:90Þ þ C

 ðLF  1:80Þ ð12Þ

where Rf7 and Rf28 are the flexural strengths at 7 and 28 days
respectively.
Similarly, as for compressive strength, the flexural strength is
also influenced by the substitution percentage of SF and LF. The de-
rived statistical model of the flexural strength shows clearly that
the latter is governed by the cement content which plays an impor-
tant role in moisturizing and has a positive effect on improving the
flexural strength, followed by the used additives which signifi-
cantly lead to an increase in flexural strength. This can be ex-
plained by the fact that the fineness promotes the physical effect
of limestone fines and the addition of grinded dune sand leads to
the formation of an additional C–S–H, thus making a further contri-
bution to the resistance of the substituted mixtures. The obtained
model also shows that beyond 5% SF and 5% of LF (in total volume
of cement), the increase of the dosages of SF and LF is harmful to
the flexural strength. It shows that couples C/LF have a great influ- Fig. 13. Iso-response curves of the slump flow as function of cement, SF and LF
ence on the studied response in comparison with the C/SF couples. proportions.
 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16
Fig. 14. Forecasting of slump flow according to the proportions of cement, SF and LF.
At 28 days, the concrete with 5% of SF and 5% of LF (HPC17)
develops the highest flexural strength (4.71 MPa), followed by
HPC16 concrete with 10% LF (4.7 MPa) and HPC21 control concrete
(4.15 MPa). And consequently, the flexural strength decreases with
the increase of the percentage of SF and LF. The lowest value is
developed by the concrete with 25% of SF by cement weight and
it is about 3.66 MPa. Ternary graphs showing the effect of the 03
constituents of the studied responses is presented the in Fig. 12.
3.2. Slump flow modelling
The workability is the essential quality of concrete that permits
to evaluate the ease of implementation of concrete. There are sev-
eral tests of workability (as the slump test using Abram’s cone),
that can evaluate the plasticity of concrete which is a prerequisite
to assess the manoeuvrability. The mathematical model used is:
SlumpðcmÞ ¼ 10:20  C þ 8:85  SF þ 6:89  LF þ C  ðSF
 3:18Þ þ C  ðLF  15:68Þ ð13Þ
According the derived statistical model of workability, it is clear
that the studied parameters have an influence on the slump value
as it is indicated by the coefficients of each parameter and give sat-
isfactory values for workability. The values measured by the
Abram’s cone slump for the different optimized concretes are given
in Table 6 and plotted in the form of ternary diagrams in Fig. 13
which illustrates better the effects of the three studied factors.
According to the profiler of slump prediction (Fig. 14) which has al-
lowed to establish the curves having as a role to facilitate the inter-
pretation and the analysis, we can see that the presence of
limestone improves the workability of concrete tested in the fresh
state (Abram’s cone), especially for the percentages of 40% of LF
and 0% of SF and that the slump of concrete with addition of lime-
stone filler (LF) and grinded dune sand significantly increases com-
pared to the reference concrete (HPC 21). This may be due to the
fact that the used additions are finely grinded and consequently fill
the micro-pores and release trapped water. But, at a high rate of
limestone filler, we can see that the latter has a negative influence
on the slump flow. HPC without additions presents a slump flow
higher than that of HPC with fillers additions. This finding proves
that the large quantity of filler rate significantly increases the de-
mand of water. Indeed, the higher the filler content, the higher is
the demand of water required to wet the entire surface area of
grains [28].
Finally, it should be noted that the obtained results confirmed
that grinded dune sand can be used as a partial cement replace-
ment in concrete production [33]. This finding has significant
implications for the sustainability and economy of concrete con-
15
struction in countries with abundant supplies of natural dune
sands.
4. Conclusions
The present work aims at the study of the effect of the partial
substitution of cement with two type of mineral additions, sili-
ceous fillers (dune sand finely grinded (SF)) and limestone fillers
(aggregate crushing waste (LF)) on the properties of HPC using
mixture design approach. The obtained results led to the following
main conclusions:
The mixture design approach developed in this experimental
work was found to be an effective technique to investigate the ef-
fects of the three types of mineral additions (Cement, SF sand and
LF) on fresh and hardened properties of HPC, in binary and ternary
systems.
Using only 21 HPC mixtures, this approach has allowed to pro-
vide mathematical models that led to present obtained results in
ternary contour plots. Such graphical representation offers a sim-
ple optimization visual tool of the proportions of cement, SF and
LF, in binary and ternary systems, in order to meet different perfor-
mances criteria of HPC such as flowability, passing ability, stability
and compressive strength. The obtained results show that, overall,
the introduction of limestone (LF) and the Siliceous (SF) in cement
(by substitution) plays a beneficial role in improving the mechan-
ical strength at both the medium and long terms. The latter in-
creases when the percentage of additions increases till an
optimum (5% SF and 5% LF), then decreases for larger percentages.
Compared to the reference concrete, the rate of increase in com-
pressive strength of the concrete made with 5% of SF and 5% of
LF is of about 47.8% at 28 days, while, in flexural strength, the in-
crease is of about 11.6%. Moreover, the obtained results confirm
that the grinded dune sand is not inert and participates in the for-
mation of new C–S–H that increases the density and the compact-
ness of the cement paste. This finding is fully consistent with the
results of literature. On the other hand, by the addition of LF and
SF, it was possible to develop high performance concrete having
good workability. The concrete with addition of 5% of limestone
presents an Abrams cone slump of 14 cm against 9 cm in the case
of control concrete, while the concrete with addition of both LF and
SF presents intermediate values (7.5–14 cm).
Finally, it should be noted that at economical level, the results
of HPC with 25% of additions of SF and LF are very encouraging.
These concretes are very economical and develop acceptable com-
pressive strengths at long-term. The used mineral additions can be
obtained at very low price: the dune sand are naturally available in
large quantities and requires only the grinding without any addi-
tional treatment and LF is obtained by sieving waste crushing.
16 R. Zaitri et al. / Construction and Building Materials 60 (2014) 8–16

References sand on the durability of mortars exposed to chemical solutions. Constr Build
Mater 2013;47:146–58.
[19] Bayramov F, Tas_demir C, Tas_demir MA. Optimization of steel fiber reinforced
[1] Russell HG. ACI defines high-performance concrete. Concr Int
concretes by means of statistical response surface method. Cem Concr Compos
1999;21(2):56–7.
2004;26:665–75.
[2] Forster SW. High-performance concrete – stretching the paradigm. Concr Int
[20] Goupy J, Creighton L. Introduction to design of experiments with JMP
Oct 1994;16(10):33–4.
examples». 3rd ed.; 2007. p. 439 [sas].
[3] Goodspeed CH, Vanikar S, Cook R. High-performance concrete defined for
[21] Goupy Jacques (2000). Design of experiments; Revue MODULAD; 2006 [in
highway structures. Concr Int February 1996;18(2):7.
French].
[4] Georges D, Jean F. New guide of concrete and its constituents. 8th ed. Edition
[22] Goupy J. Experience plans: the mixtures. Dunod. Paris. p. 285. ISBN 2 10
Eyrolles; 1998 [in French].
004218 1 [in French].
[5] Neville Adam M. Properties of concrete. 5th ed. p. cm; 2010. p. 799.
[23] Bouziani T. Assessment of fresh properties and compressive strength of self-
[6] Moore JA. High-performance concrete for bridge decks. Concr Int
compacting concrete made with different sand types by mixture design
1999;21(2):58.
modelling approach. Constr Build Mater 2013;49:308–14.
[7] Seyyed MM, Pejman A, Amir Hossein G, Amir Hossein AHB. A new predictive
[24] Aitcin CP. High performances concrete. Paris: Edition Eyrolles; 2001 [in
model for compressive strength of HPC using gene expression programming.
French].
Adv Eng Softw 2012;45(1):105–14.
[25] ACI 211. 1-91. Standard practice for selecting proportions for normal,
[8] Bederina M, Khenfer MM, Dheilly RM, Queneudec M. Reuse of local sand: effect
heavyweight, and mass concrete. American Concrete Institute, Farmington
of lime stone filler proportion on the rheological and mechanical properties of
Hills, Michigan; 1991.
different concrete sand. Cem Concr Res 2005;35(6):1172–9.
[26] Senhadji Y, Escadeillas G, Mouli M, Khelafi H, Benosman AS. Influence of
[9] Bederina M, Otteicha M, Belhadj B, Dheily RM, Khenfer MM, Queneudec M.
natural pozzolan, silica fume and limestonefine on strength, acid resistance
Drying shrinkage studies of wood sand concrete – effect of different wood
and microstructure of mortar. Powder Technol 2014;254:314–23.
treatments. Constr Build Mater 2012;36:1066–75.
[27] I_lker Bekir Topcßu, Ali Uğurlu. Effect of the use of mineral filler on the
[10] Guettala S, Mezghiche B. Compressive strength and hydration with age of
properties of concrete. Cem Concr Res 2003;33:1071-1075.
cement pastes containing dune sand powder. Constr Build Mater
[28] Bederina M, Makhloufi Z, Bouziani T. Effect of limestone fillers the physic-
2011;25(3):1263–9.
mechanical properties of limestone concrete. Phys Proc 2011;21:28–34.
[11] Alhozaimy A, Jaafar MS, Al-Negheimish A, Abdullah A, Taufiq-Yap YH, Noorzaei
[29] Makhloufi Z, Kadri EH, Bouhicha M, Benaissa A. Resistance of limestone
J, et al. Properties of high strength concrete using white and dune sands under
mortars with quaternary binders to sulfuric acid solution. Constr Build Mater
normal and autoclaved curing. Constr Build Mater 2012;27(1):218–22.
2012;26(1):497–504.
[12] Goupy J. The method experiment plans. Dunod, Paris; 1988 [in French].
[30] Makhloufi Z, Kadri EH, Bouhicha M, Benaissa A, Bennacer R. The strength of
[13] Yahia A, Khayat KH. Experiment design to evaluate interaction of high range
limestone mortars with quaternary binders: leaching effect by demineralized
water – reducer and antiwashout admixture in high-performance cement
water. Constr Build Mater 2012;36:171–81.
grout. Cem Concr Res 2001;31:749–57.
[31] Sua-iam G, Makul N. Utilization of limestone powder to improve the
[14] Nunes S, Oliveira PM, Coutinho JS, Figueiras J. Interaction diagrams to assess
properties of self-compacting concrete incorporating high volumes of
SCC mortars for different cement types. Constr Build Mater 2009;23:1401–12.
untreated rice husk ash as fine aggregate. Constr Build Mater 2013;38:455–64.
[15] Yahia A, Khayat KH, Sayed M. Statistical modelling of the coupled effect of mix
[32] Sua-iam Gritsada, Makul Natt. Use of increasing amounts of bagasse ash waste
design and rebar spacing on restricted flow characteristics of SCC. Constr Build
to produce self-compacting concrete by adding limestone powder waste. J
Mater 2012;37:699–706.
Cleaner Prod 2013;57:308–19.
[16] Goupy J, Creighton L. Introduction to design of experiments with JMP
[33] Alhozaimy A, AL-Negheimish A, Alawad OA, Jaafar MS, Noorzaei J. Binary and
examples. 3rd ed. Cary, NC: SAS Institute Inc.; 2007.
ternary effects of ground dune sand and blast furnace slag on the compressive
[17] NA 442. Algerian Standards, hydraulic binders – common Cements,
strength of mortar. Cem Concr Compos 2012;34:734–8.
Composition, Specification, IANOR, Algiers; 2003 [in French].
[18] Bederina M, Makhloufi Z, Bounoua A, Bouziani T, Queneudec M. Effect of
partial and total replacement of siliceous river sand with limestone crushed