You are on page 1of 14

Construction and Building Materials 189 (2018) 1200–1213

Contents lists available at ScienceDirect

Construction and Building Materials


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

RSM-based modeling and optimization of self-consolidating mortar to


predict acceptable ranges of rheological properties
Mehran Aziminezhad, Mahdi Mahdikhani ⇑, Mohammad Mahdi Memarpour
Department of Civil Engineering, Imam Khomeini International University, Qazvin, Iran

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 The ranges with sufficient stability


and low segregation in the fresh
phase are presented.
 Mixtures containing silica fume
decreased the rheological limitation
of EFNARC code.
 RSM is an efficient technique for
optimizing the properties of the SCM.

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

Article history: Using a high or low amount of cementitious material, water-cement ratio (W/C), and superplasticizer (SP)
Received 21 January 2018 cause instability and reduction in compressive strength of concrete according to their impact factor and
Received in revised form 19 May 2018 interaction. However, making a concrete with the optimum amount of variables could improve its prop-
Accepted 6 September 2018
erties. So, optimization of concrete ingredient due to absence of proper mix design code is required. The
remarkable features of response surface method (RSM) are modeling and optimization. This study was
carried out with the aim of investigating rheological and hardened properties of self-consolidating mortar
Keywords:
(SCM) using RSM with four factors (silica fume, slag, SP and W/C). Rheological criteria of EFNARC code
Response surface method
Self-consolidating mortar
were used to specify the optimum restriction. According to statistical analysis and validation of equa-
Optimization tions, it can be concluded that RSM is a confident and efficient method to evaluate the properties of
Silica fume SCM. In addition, according to model, the silica fume increased EFNARC code limitation ranges by
Slag improving the rheological properties, however, slag had an almost neutral role. In the best case, the addi-
Rheological properties tion of silica fume, reduced the segregation up to 5 times. Although silica fume increased the compressive
strength up to 45%, slag totally decreased the compressive strength.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction tionships between variables, creates an empirical model, and


finally, indicates the optimum response of the experimental data.
One way to study the results of experiments according to inde- Derived answer is displayed as a mathematical model which can
pendent variables or knobs is Design of Experiment (DOE). Design be used to predict results under the different settings of parame-
of experiment is considered as a proper method based on statistical ters [1–3]. One of the DOE branches is Response Surface Methodol-
analysis which reduces the number of experiments, evaluates rela- ogy (RSM). DOE branches use a collection of mathematical and
statistical techniques to produce a robust design of experiment.
It has been reported that RSM can be widely used in optimizing,
⇑ Corresponding author.
mostly for industrial processes [4–6]. There are several methods
E-mail addresses: mehranaziminezhad@edu.ikiu.ac.ir (M. Aziminezhad),
mahdikhani@aut.ac.ir (M. Mahdikhani), memarpour@eng.ikiu.ac.ir (M.M. Memarpour). for optimization and modeling in the field of concrete technology.

https://doi.org/10.1016/j.conbuildmat.2018.09.019
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1201

For instance, Yaman et al. [7] and Mashhadban et al. [8] studied on the experiment design tools. It should be noted that there is a
mechanical parameters of concrete using artificial neural network, strong relationship between the RSM and linear regression analy-
Yan et al. [9,10] and Gholampour et al. [11] studied mechanical sis. In RSM, attempts were made to find a way to estimate interac-
properties of concrete using genetic algorithm, also, Mehrinejad tions of quadratic effects and local form of the response surface
Khotbehsara et al. [12] worked on durability and microstructure using a good experimental program. If the applied surface gives a
of SCM using an adaptive neuro-fuzzy inference system (ANFIS). desirable approximation of actual response surface, the response
Many studies have been carried out on mechanical, rheological surface analysis will be compared with similar approximation of
and durability properties of SCC and SCM so far [13–18]. In addi- the actual system analysis. For a nonlinear and complex system,
tion, in order to study the effect of various pozzolans, admixtures the equation Y and variables of xi are calculated by Eq. (1) as
and aggregates on fresh and hardened properties of SCC and follows:
SCM, additional researches were conducted [19–23].
Alyamac et al. studied the effect of waste marble powder on
X
k X
k XX
k
fresh and hardened properties of SCC by RSM in order to achieve Y ¼ b0 þ bi xi þ bii x2i þ bij xi xj þ e ð1Þ
an eco-efficient SCC mixture [24]. A study was carried out by Gha- i¼1 i¼1 i<j
fari et al. to predict the performance of self- consolidating ultra-
high performance concrete reinforced with hybrid steel fibers by
using Central Composite Design (CCD) model based on RSM. They where, Y is the predicted response (i.e., mini slump flow, mini V-
could obtain the goal mixture with the highest flexural strength funnel, mini J-ring, mini column segregation and the compressive
and minimum content of steel fibers according to optimum results strength), b0 represents the intercept, bi the first-order (linear) coef-
[3]. Mehdipour et al. conducted a study aimed to evaluate statical ficient, bii the second-order (quadratic) coefficient, bij the coefficient
and dynamical stability by observation of hardened visual stability of interaction effect, xi and xj are independent variables, and e the
index (HVSI) and visual stability index (VSI). Contour graphs associated random error. One of the advantages of the second-
describe the CCD modeled results. According to the results, the order model is flexibility. This model can cover a wide type of
highest performance among the tested ternary and binary cementi- functions.
tious systems is a ternary combination with the amount of 25% MK The central composite design (CCD) is a type of RSM, which is
and 25% FA [25]. Soto-Pérez et al. obtained the equations of com- the most well-known fitting method. This design includes 2 k fac-
pressive strengths and spread percentage of the fresh paste accord- torial points with nF factorial runs, 2 k axial (star) runs and nc cen-
ing to fly ash-binder, nano iron oxide-binder and water-binder tral runs. The factorial runs are used for fitting the first-order
ratios by using RSM. Afterward, changes of variables were reported model and the axial runs for the fitting of second-order sentences.
and then chosen targets were used to reach optimum values of fly The value of the alpha determines the distance between the axis
ash-binder, nano iron oxide-binder and water-binder ratios [26]. points and the center point. Rotationality is a reasonable basis
Rezaifar et al.’s study was proposed to optimize the concrete made for choosing the design of response surface. Since the aim of the
with crumb rubber (CR) as waste material in combination with RSM is optimization, and while the location of optimal point is
metakaolin (MK). An optimization is made for the highest strength unknown prior to tests, using a design that delivers equal accuracy
and the lowest water absorption. A CCD model was used to design for estimation in all directions is beneficial. The CCD can be rotated
experimental program. The equations were obtained by regression by selecting alpha. The alpha value depends on the number of
analysis and then verified. RSM was utilized to design the experi- points on the factorial design part. Different formulas have been
ments [27]. Mermerdas et al. in the first step of the paper, by using proposed for obtaining alpha, in some cases they are also experi-
a full factorial experimental program investigated results of binder mentally selected. However, the most used is a = (nF)1/4, where
content, curing time and curing temperature on the compressive nF is the number of factorial points of the design [2,4–6,29].
strength of light-weight geopolymer mortar (LWGM). The base
materials used for LWGM are fly ash (FA) and ground granulated
blast furnace slag (GGBFS). The second step of the study was to opti-
mize the experimental parameters through RSM. The proposed 3. Material and test method
optimum parameters were tested by experimental verifications.
The experimental verifications showed good accordance with the 3.1. Materials
optimized results. Finally, results indicated that the compressive
strength increases with the increase of curing temperature, curing In this study commercial type II Portland cement according to
period and binder content of LWGM [28]. ASTM C150 [30] was used. The chemical composition of Portland
Several papers have been published on RSM and fresh properties cement, silica fume (SF) and slag are tabulated in Table1. Particle
of self-consolidating concrete, however, none of them provides an size distribution (PSD) of pozzolans and cement are shown in
optimal range of variables containing pozzolans that present an Fig. 1. Polycarboxylate ether superplasticizer with a specific weight
appropriate rheology. This study was carried out to model and opti- of 1.08 has been used. Aggregate sand was provided with maxi-
mize the rheological characteristics and compressive strength with mum size of aggregate (MSA) 4.1 mm. The gradient diagram is
four factors: water-cement ratio, silica fume, slag and superplasti- given in Fig. 2. For all designs, limestone powder is used as filler.
cizer. For this purpose, the RSM method has been used. To ensure
obtained equations, 24 concrete mixtures have been implemented Table 1
and the results of this method have been compared with the results Chemical composition of cement, silica fume and slag.
of the equations. After ensuring the results of the equations, the dia-
Items Cement (%) Silica fume (%) Slag (%)
grams of experiments are plotted, analyzed and interpreted.
SiO2 20.38 85.19 34.5
Al2O3 4.13 0.31 7.58
Fe2O3 3.82 3.42 0.65
2. Response surface method CaO 62.96 0.75 32.78
MgO 3.5 2.06 7.56
In 1951, the surface response method (RSM) was introduced for SO3 2.87 _ _
LOI 0.98 3.8 7
the first time by Box and Wilson [4,29], and is still used as one of
1202 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

100.0

90.0

80.0
percentage passing (%)

70.0

60.0

50.0

40.0

30.0 Slag

20.0 Silica Fume

10.0 Cement

0.0
0.1 1.0 10.0 100.0

Particles Size Log (µm)

Fig. 1. Particle size distribution (PSD) of cement, silica fume and slag.

3.2. Preparation of samples and test procedure

The aggregates and cement are mixed for 1 min, then the water
and the SP are gradually added to the mixture, and afterward slag
and SF are added in the form of sludge. The mixture is mixed for
5 min after adding water. Immediately after preparing the mortar,
the mini slump flow test and mini V-funnel are performed accord-
Fig. 3. Details of mini J-ring.
ing to EFNARC; also mini J-ring, mini column segregation tests are
performed as suggested by researchers [31–34]. Then the samples
of 100  100  100 mm are filled up for compressive strength with SI = segregation index (in percent), CAT = mass of aggregates in
self-consolidating mortar [35]. In the mini slump flow test, frustum the top section of the mini column, CAB = mass of aggregates in
cone, located on a leveled surface was filled up with SCM; then by the bottom section of the mini column.
lifting the cone and after the spreading of SCM, two perpendicular
diameters of the mortar are measured as mini slump flow. For mini
V-funnel test, the funnel is filled with SCM and mortar is held for 4. Results and discussion
10 s; then, the time interval between the trap door opening and
seeing the first light from above through the funnel, is recorded 4.1. Model fitting and validation
as discharge time of mini V-funnel. The mini J-ring test is carried
out according to Mahdikhani and Ramezanianpour’s study [33]. The first step to create RSM design is to specify the number of
In mini J-ring, the ability to pass the SCM through the bars is exam- factors and levels. The number of factors is selected based on the
ined. The proposed detail for this test is shown in Fig. 3. In order to variables of the proposed design and the level interval based on
obtain the segregation percentage of the SCM, after 15 min of fill- the applicable test range or suggested values by researches. After
ing the column with the SCM, the mortar of the upper and lower determining the codes and obtaining the test program, the tests
parts of the column are washed on #50 sieve, and then the remain- are carried out and responses are recorded to obtain the relation-
ing aggregates on the sieve are dried. The segregation percentage is ships between the variables and the results. This study was con-
calculated according to the Eq. (2). This test is carried out according ducted in form of 4 factors and 5 levels, with 6 midpoints, 8 axis
to the Libre et al’s study [32]. points and alpha value of 2. Responses include mini slump flow,
  discharge time of mini V-funnel, diameter of mini J-ring, segrega-
ðCAB  CAT Þ tion index and compressive strength. The levels of coded and
SI ¼ 2  100 ð2Þ
ðCAB þ CAT Þ actual variables are shown in Table 2. Design and analysis of
RSM is carried out using the Minitab software. The desired CCD
with 4 factors includes 30 runs, and the results of the 30 mixture
designs are presented in Table 3.

4.2. Modification of equations according to statistical criteria

In CCD procedures, an iterative model fitting process is used in


order to achieve an accurate model for each response. First, in
order to estimate the relationship between the variables and the
response based on experimental results of CCD, a second-order
polynomial model (Eq. (2)) is chosen. Then, the process is followed
by removing insignificant terms from the model by performing a t-
test. After removing each term, the fitting process is repeated until
all the insignificant terms are removed from the model. Afterward,
a proper transformation could be selected for normalization of
equations.
The next step after generating the model is to evaluate the accu-
Fig. 2. Grading diagram of aggregates. racy of equation by performing residual analysis. It includes statis-
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1203

Table 2
Variables and range of variation.

Variable Levels
Actual Coded 2 1 0 1 2
W/C X1 0.38 0.41 0.44 0.47 0.50
Silica Fume (SF) X2 0 1.5 3 4.5 6
Slag X3 0 3 6 9 12
Super Plasticizer (SP) X4 0.50 0.55 0.60 0.65 0.70

Table 3
Design points and corresponding responses.

Run No W/C Silica Fume (%) Slag (%) SP (%) Mini Slump Flow(cm) Mini J-Ring (cm) Mini V-Funnel (sec) SI (%) 28-day CS(Mpa)
1 0.38 3.0 6 0.60 16.90 15.05 11.58 1.20 60.00
2 0.41 4.5 3 0.55 16.25 15.05 8.00 2.80 59.00
3 0.41 1.5 9 0.55 17.20 15.30 8.10 3.14 51.50
4 0.41 1.5 3 0.55 17.25 14.75 8.67 0.50 53.50
5 0.41 4.5 9 0.65 22.50 21.20 4.00 1.50 56.00
6 0.41 1.5 9 0.65 23.50 22.00 5.20 3.84 52.00
7 0.41 4.5 9 0.55 17.85 16.60 5.80 1.75 57.00
8 0.41 4.5 3 0.65 25.70 24.10 5.00 4.65 58.80
9 0.41 1.5 3 0.65 24.80 23.90 5.50 3.44 54.00
10 0.44 3.0 6 0.60 24.20 21.30 3.60 4.84 53.50
11 0.44 3.0 6 0.60 24.15 21.35 3.64 4.40 53.00
12 0.44 3.0 6 0.60 24.65 21.60 3.54 4.22 52.45
13 0.44 3.0 6 0.60 24.50 21.60 3.60 10.00 53.50
14 0.44 3.0 6 0.50 17.40 16.10 4.35 4.67 58.00
15 0.44 3.0 12 0.60 24.85 22.75 2.90 9.10 51.00
16 0.44 6.0 6 0.60 21.95 19.75 3.73 8.23 61.00
17 0.44 3.0 6 0.70 30.65 30.50 2.10 8.82 54.00
18 0.44 3.0 6 0.60 24.65 21.60 3.52 4.40 54.00
19 0.44 0.0 6 0.60 26.00 24.00 3.80 12.26 52.00
20 0.44 3.0 0 0.60 24.30 22.40 3.25 3.46 56.30
21 0.44 3.0 6 0.60 24.55 21.55 3.62 8.37 52.00
22 0.47 4.5 9 0.65 29.30 29.50 2.26 5.48 54.00
23 0.47 1.5 9 0.65 30.45 31.80 2.30 19.50 46.00
24 0.47 4.5 3 0.65 30.45 30.60 2.31 16.10 58.00
25 0.47 4.5 3 0.55 25.65 24.50 2.50 8.03 57.00
26 0.47 1.5 9 0.55 30.90 29.60 2.70 15.67 47.00
27 0.47 1.5 3 0.65 33.25 32.65 2.30 26.98 47.00
28 0.47 1.5 3 0.55 29.60 29.00 2.50 16.25 50.00
29 0.47 4.5 9 0.55 24.65 21.40 3.15 6.29 51.00
30 0.50 3.0 6 0.60 31.15 30.10 2.18 13.20 46.00

tical calculation such as residual standard deviation as well as error and have no influence on the equation. In addition, if an inde-
residual plots, in which the adequacy of the selected model can pendent variable was deleted from the equation, the whole of its
be graphically evaluated. Finally, Correlation coefficients and P- interaction should be omitted from the equation. There is no accu-
values of obtained model could be examined. According to the rate formal statistical rule to confirm for this criteria. However,
mentioned statistical process, four steps are applied in this section: variables with P-Value lower than 0.1 are effective also variable
with P-value lower than 0.05 and 0.001 could be determined as
4.2.1. Checking the P-Value significant and highly significant parameter, respectively.
At first, the equation of the results is obtained based on second- According to Table 4, the coefficients and P-values are specified
order polynomial model. Then, variables with P-value higher than from the variables of water-cement ratio, silica fume (SF), slag, and
0.1 must be removed; the remaining variables just increase the superplasticizer (SP) and their interactions. The results related to

Table 4
ANOVA response for proposed models.

Mini Slump Flow Mini V-Funnel SI Mini J-Ring 28-day CS


Term Coef P-Value Coef P-Value Coef P-Value Coef P-Value Coef P-Value
Constant 3.1849 <0.0001** 1.8704 <0.0001** 2.436 <0.0001** 3.1108 <0.0001** 3.9708 <0.0001**
X1 0.3428 <0.0001** 0.9254 <0.0001** 1.808 <0.0001** 0.3928 <0.0001** 0.0956 <0.0001**
X2 0.0721 0.001** 0.0715 0.099 0.591 0.008* 0.0791 0.012* 0.1051 <0.0001**
X3 0.0519 0.001**
X4 0.2453 <0.0001** 0.3302 <0.0001** 0.497 0.023* 0.2977 <0.0001**
X1X1 0.5961 <0.0001**
X2X2 0.748 0.057 0.0412 0.094
X1X2 0.1163 0.018* 0.2660 0.015* 1.496 0.006* 0.1765 0.021* 0.0562 0.089
X1X4 0.2298 <0.0001** 0.4100 0.001**

Note: It is noteworthy to mention that since an independent variable was deleted from equation whole of its interaction should be omitted from equation.
**
Highly significant parameter (P-Value < 0.001).
*
Significant parameter (P-Value < 0.05).
1204 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

the mini slump flow, mini V-funnel, mini J-ring, mini column seg- tests, the SP has very little impact on 28-day compressive strength
regation and 28-day compressive strength are presented in the results. Therefore, it is omitted from the equation. Now, after iden-
Table 4. tifying the effective parameters, the p-value of the models is given
As shown in Table 4, slag has insignificant effect on the mini in Table 6.
slump flow. Also, according to the coefficient, it can be determined It should be noted that almost no pozzolans participate in the
that the ratio of water-cement has the greatest impact on the chemical reaction at the fresh phase of SCM (when the tests was
spread of the mini slump. It should be noted that when a main carried out) because the effect of pozzolan on the rheological prop-
parameter is removed, all related interactions and second-order erties of the SCM depends on their porosity, shape and the size of
are also deleted. their constituent particles [36].
According to the results of the mini V-funnel, slag has a little
effect on the time of discharge. Therefore, the slag due to P-value 4.2.2. Obtaining and normalization of the equations
higher than 0.1 is removed from the equations. After water- Box and Hunter [29], believed that the second-order response
cement ratio, SP is the most effective parameter on discharge time surface design should be rotational. It means that V ¼ ½y ^ðxÞ has
of SCM. the same value at all points of (x) which have equal distance from
In the mini column segregation parameters, slag has a neutral each other from the center of the design, i.e. the variance of
roll and has no effect on segregation, contrary to the water- response predicted in the spheres is constant.
cement ratio which is the most effective factor. The existence of inconstant variance shows that there is devia-
According to the coefficient results of the mini J-ring, both tion from the analysis of the variance assumption criteria that was
water-cement ratio and SF have a considerable effect on the spread mentioned earlier. In this section, various transformations are
diameter. examined using the Box-Cox method. The transformations are
In addition to the water-cement ratio and SF, slag has also been selected based on trial and error method (Table 5). According to
effective on the compressive strength equations. Unlike rheological the formula y ¼ yk , k is the parameter of the transformation.
The mini slump flow, mini V-Funnel, mini column segregation,
Table 5 mini j-ring, and 28-day compressive strength (28-day CS) equa-
Transformation of responses. tions are shown as follows:
Responses Transformation lnðMini Slump FlowÞ ¼ 11:69 þ 30:63 W=C þ 0:260 SF
Mini Slump Flow (cm) k = 0 (natural log)
Mini V-Funnel (sec) k = 0.5 (square root)
þ 19:30 SP  0:646 W=C  SF
SI (%) k = 0.5 (square root)  38:29 W=C  SP ð3Þ
Mini J-Ring (cm) k = 0 (natural log)
28-day CS (MPa) k = 0 (natural log)

Table 6
The results for full regression models.

Responses R2 (%) Adj-R2 (%) Pred-R2 (%) (Adj-R2) - (Pred-R)2 (%) P-Value
Mini Slump Flow (cm) 95.89 95.03 92.64 2.39 <0.0001
Mini V-Funnel (sec) 96.53 95.42 93.15 2.27 <0.0001
SI (%) 81.41 77.53 70.69 6.84 <0.0001
Mini J-Ring (cm) 92.27 91.03 87.74 3.29 <0.0001
28-day CS (MPa) 85.61 82.61 75.11 7.50 <0.0001

Table 7
Verification design points and corresponding responses.

Mix No W/C Silica Fume (%) Slag (%) SP (%) Mini Slump Flow (cm) Mini V-Funnel (sec) SI (%) Mini J-Ring (cm) 28-day CS (MPa)
1 0.38 1.5 3 0.55 15.00 14.00 1.00 15.00 54.00
2 0.38 3.0 6 0.60 16.60 9.91 1.00 15.05 55.00
3 0.38 4.5 9 0.65 23.60 6.30 0.50 20.80 58.40
4 0.38 6.0 12 0.70 27.05 4.70 0.50 25.90 61.00
5 0.41 0.0 6 0.60 21.00 4.22 13.00 20.00 54.40
6 0.41 1.5 9 0.65 27.00 3.68 4.47 30.40 57.00
7 0.41 3.0 12 0.70 28.35 3.95 4.05 26.00 58.00
8 0.41 4.5 0 0.50 17.20 6.08 3.50 16.35 58.10
9 0.41 6.0 3 0.55 16.50 6.15 3.50 15.35 60.60
10 0.44 0.0 12 0.70 34.80 2.50 40.00 36.20 47.00
11 0.44 1.5 0 0.50 23.00 3.50 0.50 21.25 57.70
12 0.44 3.0 3 0.55 25.90 3.03 0.50 24.20 53.20
13 0.44 4.5 6 0.60 27.60 3.54 0.50 26.25 56.80
14 0.44 6.0 9 0.65 29.40 2.68 12.70 28.45 55.40
15 0.47 0.0 3 0.55 29.55 1.78 40.00 34.55 47.30
16 0.47 1.5 6 0.60 31.15 2.10 21.00 32.00 49.40
17 0.47 3.0 9 0.65 32.60 2.43 30.00 34.10 54.40
18 0.47 4.5 12 0.70 32.60 2.07 27.00 32.35 55.60
19 0.47 6.0 0 0.50 23.20 1.83 0.50 21.50 57.30
20 0.50 0.0 9 0.65 36.00 1.36 58.00 37.00 36.00
21 0.50 1.5 12 0.70 37.60 1.81 45.00 40.65 37.00
22 0.50 3.0 0 0.50 29.10 1.95 16.90 29.60 47.00
23 0.50 4.5 3 0.55 28.40 1.61 18.17 28.75 50.05
24 0.50 6.0 6 0.60 31.70 1.80 30.00 28.50 47.00
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1205

^
ðMiniV  FunnelÞ 0:5 ¼ 62:78  206:6 W=C  0:675 SF lnðMini J  RingÞ ¼ 2:771 þ 9:49 W=C þ 0:405 SF
 33:40 SP þ 165:6 W=C  W=C þ 2:977 SP  0:980 W=C  SF ð6Þ
þ 1:479 W=C  SF þ 68:4 W=C  SP ð4Þ
lnð28  day CSÞ ¼ 5:072  2:531 W=C  0:1298 SF
^
ðSIÞ 0:5 ¼ 23:44 þ 55:08 W=C þ 2:96 SF þ 4:97 SP  0:00865 Slag þ 0:00457 SF  SF
þ 0:0831 SF  SF  8:31 W=C  SF ð5Þ þ 0:312 W=C  SF ð7Þ

Fig. 4. The 95% confidence and 95% prediction range for scattered results of a) mini slump flow b) mini V-funnel c) mini column segregation d) mini J-ring e) 28-day
compressive strength.
1206 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

4.2.3. Checking the residual results the results of the new mixtures design. After ensuring the accuracy
The scattering of residuals should be in irregular form and of the equations, contour graphs are plotted and the optimum
should not include any scattering modes of cluster, funnel shape, ranges are determined in the end step. At this stage, 24 mixture
strong curve and large residual. If the residual results tend to be designs are selected randomly, none of which are among the
in one of the mentioned modes, then there is an error in the data 30 runs of Table 3. After testing and recording the results of the
and it should be evaluated. In the following, the results of the test fresh and hardened phases of the SCM, the accuracy of the pre-
residues are shown in Supplementary Figures. These residuals sented equations is examined. The 24 design points and corre-
belong to mini slump flow, mini V-funnel, mini column segrega- sponding responses are presented in Table 7.
tion, mini J-ring, and 28-day compressive strength tests. As the The results of these 24 mixtures were compared with the
results of the figures show, the scattering modes are irregular results of the equations in order to validate the equations. Accord-
and are not included in the referred modes. ingly, the results that were predicted based on the equations ver-
sus the actual results were plotted. The probability of re-
occurrence intervals is classified according to two criteria of 95%
4.2.4. Examination of correlation coefficients and P-value model
confidence and 95% prediction. The Eqs. (8) and (9) were used for
Analysis of variance (ANOVA) is similar to regression, which is
mentioned criteria respectively:
used to examine and model the relationship between response
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and variables. As shown in Table 6, the correlation results of the  2
1 ðxp  xÞ
three tests are more than 90% and the other two tests are reported b
y  tða2;n 1Þ se þ ð8Þ
more than 80%, which is considered as an acceptable result in this n SXX
regard. sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2
According to the statistical suggestions, the highest value of the 1 ðxp  xÞ
difference Adj-R2 and Pred-R2 should be <20%, and higher value is a b
y  tða2;n 1Þ se 1 þ þ ð9Þ
n SXX
sign of error in the statistical data. According to Table 6, the most
reported difference is 7.5%, which corroborates the correlation where,
accuracy between the results and the predicted graph. n 
X 
 2
SXX ¼ xi  x
i¼1
4.3. Verification of models
 1X n

If the statistical adequacy of the obtained model is confirmed, x¼ xi


n i¼1
the next step is to examine the correctness of the equations with

Fig. 5. Mini slump flow variations versus water-cement ratio and silica fume at level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1207

n is the number of points, xp is the characteristic value, t (a/2, n*1) is diameter of mini slump flow increased, too. The maximum effect
the critical t-value for a/2 and n*  1 degrees of freedom, and the of water-cement ratio on the mini slump flow spread diameter var-
(se) is the standard deviation. ied up to 100%, i.e. in the Fig. 5(a, b) spread diameter of mini slump
According to the results of the mini slump flow (Fig. 4a), it is flow grew from 20 to 40 cm. Unlike water-cement, the rise in SF
shown that more than 60% of the points are in the 95% confidence reduced the spread diameter. However, this effect is negligible in
range and all points are located in the 95% prediction range. This the low ratio of water-cement ratio, hence it is quite visible in
dispersion is desirable and confirms the model. the high water-cement ratio (0.475–0.5). Similar results have been
In Fig. 4b–e, which are the results of the mini V-funnel, mini reported by other researchers [33,37]. This is owing to the fact that
column segregation, mini J-ring and 28-day compressive strength, SF in the presence of high water content is capable of taking further
approximately 80, 70, 70 and 50 percent of the points are placed in effect. Higher volume of fine particles and greater specific surface
the zone of 95% confidence. Also, nearly 100% of the verification area (SSA) of the SF resulted in higher relative water demand
points are in the 95% prediction range. (RWD) to increase fluidity. The SSA and particle-size distribution
It can be said that the high scattering at the end of the funnel (PSD) of cementitious materials can have a substantial effect on
shape and the larger numbers of the axis in relation to the mini col- fresh and hardened properties of SCM which was confirmed by
umn segregation and the mini J-ring are due to the inaccuracy of previous studies [36,38,39]. According to Fig. 1, due to smaller par-
the evaluation of the two apparatuses at high amount of responses. ticle size of SF (which devoted more SSA) in comparison with
In other words, these two apparatuses have more tolerance in high cement, by increasing the amount of SF lower spread of SCM was
values and cannot predict the results exactly. The specified points indicated. With regard to close gradient diagram of cement and
can confirm the validity of this phenomena. slag, by increasing the volume of slag no significant effect was
shown.
The main effect of SF on mini slump flow test in the presence of
4.4. Discussion and interpretation of models
high water-cement ratio (0.475–0.5) was 53% and at the lowest
amount of water-cement ratio was 0%. The variations in the spread
4.4.1. Mini slump flow
diameter of the mini slump flow have a direct relationship with
The ranges of water-cement ratio (0.38–0.5), SF (0–6%) and SP
increasing the amount of SP from 0.5 to 0.7%.
in the three levels of 0.5, 0.6 and 0.7% were presented. The varia-
tion in the spread diameter of mini slump flow starts from
<20 cm and extends up to 40 cm with interval of 2 cm. Fig. 5 shows 4.4.2. Mini V- funnel
that the water-cement ratio, varies from 0.4 (blue color) to 0.5 Variation in discharge time of SCM in the mini v-funnel test
(green color), while the water-cement ratio was increased, the begins in <3 s, and continues with an interval of 1–13 s and further.

Fig. 6. Variation of the discharge time of the SCM versus the water-cement ratio and silica fume at level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.
1208 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

The increase of water in the SCM leads to reduction in the friction time was 8 s, while in two other levels (0.5% and 0.6%), values
between the constituent materials by lubricating them, such as greater than 13 s were observed.
aggregates and cement, and they can more easily slip over each
other. Therefore, as shown in Fig. 6, the blue color of the contours 4.4.3. Mini column segregation
varied from water-cement ratio of 0.5–0.4, meaning that increasing The results of the separation percentage of the SCM start at <4%
the water-cement ratio reduced the discharge time of SCM. and continue with intervals of 4–40% and more. As it is expected,
At the SF level of 0%, the ratio of minimum to maximum dis- since increases in water-cement ratio reduces the friction between
charge time was 4.33. Silica fume, due to generating cohesion aggregates, therefore smaller aggregates can slip out of the coarser
between the components of the SCM (which increased stability aggregate and move downward, which causes segregation. It
of the mortar), causes greater homogeneity of the SCM, which should be noted that an increase in water-cement ratio (mixture
decreases the discharge time. Although it should be noted that this without SF), led to increase the segregation up to 10 times, while
effect is negligible in comparison with effect of water-cement ratio this increase at the level of 6% SF was 3 times. As shown in
variations, especially in the amount of 0.5% SP. Since in a high Fig. 7, the contours also express this issue that the dark blue range
amount of water, the effect of water-cement ratio is greater than (water-cement ratio of 0.4) changes to green area, (water-cement
the effect of SF content. Therefore, in this condition the role of SF ratio of 0.5), meaning that increasing the water-cement ratio
is practically insignificant. In the best case, SF has reduced the dis- increased the segregation percentage of the SCM. As observed in
charge time up to 2 times. There is a direct relationship between the results of the mini slump flow, owing to the greater effect
plastic viscosity and the V-funnel times and also slump flow and and participation of SF in the presence of high water-cement ratios
yield stress. According to Zerbino et al. [40] and Zhang and Han (0.475–0.5), the increase in the amount of SF improved the proper-
[41] SCC mixtures with simultaneously low values of plastic vis- ties of the SCM and reduced the segregation. In the best case, the
cosity and yield stress showed satisfactory performance in practi- addition of SF up to 5 times has led to reduction in the percentage
cal applications. The viscosity and the yield stress of SCM were of segregation. Greater involvement and effect of SF on the rheo-
decreased clearly with the augmentation of the SF. logical properties at high water-cement ratios is due to the disper-
It is observed that an increase in the amount of SP has increased sion of its components by water. Therefore, greater participation
the flowability and workability of the SCM. So that in the level of leads to more adhesion to the SCM. A growth in SF in the range
0.7% of SP (Fig. 6c), discharge time is allocated more to the dark of low water-cement ratio had no significant effect. Due to the
blue zone (<3 s). In addition, the maximum recorded discharge increased workability caused by the SP, the stability of the SCM

Fig. 7. Variation of the segregation percentage versus the water-cement ratio and silica fume at level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1209

are reduced, hence increasing the amount of SP increases the seg- of SP is constant, the SP-cement ratio increased, which led to
regation percentages. growth in flowability and spread diameter.

4.4.5. Compressive strength


4.4.4. Mini J-ring Contrary to the results of fresh phase of SCM, the SP had no
The overall trend of the mini J-ring variation according to Fig. 8 effect on compressive strength, but in the hardened phase, slag
is similar to the mini slump flow results, with the exception of a had a significant effect on compressive strength. Therefore, the
few cases that are being investigated. The first case is that the aver- contour graphs of Fig. 9 are based on variation in the water-
age diameter of the measured mini J-ring spread has been less than cement ratios and SF, which are presented at three levels of 0, 6
the mini slump flow. This can be achieved by comparing the rise of and 12% of slag. The range of results start at <44 MPa which is
the blue zone (<20 cm) in mini slump flow and mini J-ring at the described with steps of 2–64 MPa and more. Fig. 9 shows that
same amounts of the SP. However, the only opposite state to increasing the water-cement ratio reduced the compressive
reduce the diameter occurred at high flowability situation, i.e., strength. This decrease in compressive strength in the zero SF level
the water-cement ratio of 0.5, 0% SF and the SP level of 0.6 and is extremely sharp and varies greatly with changes from the green
0.7%. In general, the SCM has a relatively acceptable stability, by to blue range. For instance, at zero slag level, the compressive
passing the mortar from adjacent bars, the friction between the strength had shifted from 44 MPa to 62 MPa, which is a wide range
SCM and the bars prevents the SCM spread, and the diameter of and includes 44% reduction. But in the SF level of 6% variations are
spread decreases in comparison with the mini slump flow. How- about 2 MPa, i.e. only 0.03% decrease in compressive strength. It
ever, in the high-flowability range, the SCM has a low stability. could be noted that SF have a remarkable role in augmenting the
In this case, the presence of bars led to increase in the diameter compressive strength.
of the spread by increasing bleeding. Based on Mehdipour and Khayat [38]. The dispersion of binders
The second case is the reverse trend of the SF, the effect of has significant influence on compressive strength. In a well-
which is at a level of 0.7% SP. Silica fume at the low cement- dispersed system, the filling effect of the SF can be effectively pre-
water ratio has similar trend to the mini slump flow but this phe- pared to fill voids among cement grains, hence leading to signifi-
nomenon is more pronounced in the mini J-ring. Because the cant enhancement in packing density of the binder. Similar to
amount of SF is replaced by cement, and also there is little effect the results of mini slump flow, it is also observed in hardened
of SF on the fresh phase of the SCM in presence of low amount of phase, in which SF had a sharp variation in the presence of more
water, increasing the amount of SF, caused a greater effect of SP water content, due to greater involvement and participation of SF
on the low amount of cement. In this case, although the amount in the presence of high water content. This phenomenon was also

Fig. 8. Mini J-ring variation versus the ratio of water-cement ratio and silica fume at level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.
1210 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

Fig. 9. The 28-day compressive strength variation versus water-cement ratio and silica fume at the level of (a) 0% slag, (a) 6% slag, (a) 12% slag.

observed by Rao [42]. The highest variation of compressive feature has led to the control of some critical conditions for the
strength influenced by SF was about 20 MPa (Fig. 9a). This SCM, including bleeding and segregation when the unstable factors
increased compressive strength was nearly 45% in high water- play an impressive role. In fact, the threshold of qualified condi-
cement ratio. Although SF does not change compressive strength tions for the stability of the SCM has increased by using SF
significantly in the low water-cement ratio, due to sufficient stabil- (Fig. 11). Therefore, the previously proposed ranges of the EFNARC
ity conditions in the fresh phase, the high amount of compressive code, could be promoted.
strength was recorded. The range of discharge time and spread diameter of SCM, in
Based on Fig. 9, it is indicated that increasing the amount of slag mini V-funnel and mini slump flow rose as follows; its lower limit
has reduced compressive strength. By comparing the graphs of of discharge time has changed from 7 to 3 s, and its upper limit
Fig. 9a to Fig. 9c which is related to the values of 0 to 12% of slag, remains the same as 11 s. Also, the mini slump flow spread diam-
it is clear that the blue zone (which represents lower compressive eter has changed from 24–26 cm to 20–30 cm. The 4 cm is added to
strength) has increased. the value of upper limit and 4 cm is subtracted from the lower limit
of the EFNARC code. Obviously, controlling the implementation of
the proposed 10 cm range is easier than 2 cm range which is pro-
5. Multi-Responses optimization of SCM vided by the Code. The segregation percentage is considered to
be <15%. With respect to the Fig. 11, the proposed limits are
The EFNARC code [31] provides acceptance criteria for tests; as applied on all three conditions of the mentioned experiments.
a result of that a SCM with appropriate ranges of rheology is pre- Therefore, by selecting a mixture in the white range of the graphs,
sented. According to this code, the spread diameter of the SCM the SCM has favorable results in terms of rheology.
for the mini slump flow is 24–26 and the optimal time for the dis- According to the three graphs shown in Fig. 11, the largest
charge time of the SCM from the mini V-funnel is 7 to 11 s. Also, white area is related to graph of 0.6% SP with a water-cement ratio
according to other researches, the highest acceptable segregation of 0.425%. Proper amount of SP and water-cement ratio must be
for self-consolidating mortar is 15 percent [32,34,43]. In Fig. 10 used to achieve required amount of flowability for SCM. White
these restrictions have been applied at the 0.5, 0.6 and 0.7% SP zones provide low viscosity and yield stress, though they are
levels. As can be seen, owing to the effects of SF, there is no zone placed in a range <15% of segregation. In the other hand, the segre-
for overlaying of these three restrictions. gation of specified ranges (20–30 cm for mini slump flow and 3–
The use of SF has improved the rheological properties of SCM by 11 s for mini V-funnel) for fresh state of SCM is nearly lower than
increasing adhesion in the SCM components and stability [33]. This 15%. It was proved that this range of flowability did not make any
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1211

Fig. 10. Applied EFNARC code ranges on water-cement ratio versus silica fume at Level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.

Fig. 11. The applied proposed ranges of this study on water-cement ratio versus silica fume at level of (a) 0.5% SP, (b) 0.6% SP, (c) 0.7% SP.
1212 M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213

segregation problem for SCM even with increasing the amount of


Conflict of interest
SP. Most mixtures with mini slump flow higher than 30 cm and
mini V-funnel lower than 3 s were placed in a zone covered with
None.
segregation higher than 15%, even by using different dosage of SF.
As mentioned in results of fresh phase, although SF have a great
influence on improving the fresh properties of SCM in high water- Acknowledgment
cement ratio, high content of water-cement ratio made the SCM
totally unstable due to dominant effect on mini slump flow, mini This research did not receive any specific grant from funding
V-funnel discharge time and segregation. So, content of SF was agencies in the public, commercial, or not-for-profit sectors.
restricted to lower ratio of water-cement (lower than 0.475–0.5
ratio). In the other hand, mixtures containing up to 6% SF could Appendix A. Supplementary data
not control the adequate stability of SCM in high water-cement
ratio. Supplementary data associated with this article can be found, in
Mixture with mini slump flow lower than 20 cm and mini V- the online version, at https://doi.org/10.1016/j.conbuildmat.2018.
funnel higher than 11 s (with high viscosity and yield stress) 09.019.
increase the problem of fluidity.

References
6. Conclusion [1] M. Romagnoli, C. Leonelli, E. Kamse, M. Lassinantti Gualtieri, Rheology of
geopolymer by DOE approach, Constr. Build. Mater. 36 (2012) 251–258,
In this study, rheological and hardened properties of SCM with https://doi.org/10.1016/j.conbuildmat.2012.04.122.
[2] P.G. Mathews, Design of experiments with MINITAB, ASQ Quality Press (2005).
four factors (water-cement ratio, silica fume, slag and superplasti- [3] E. Ghafari, H. Costa, E. Júlio, RSM-based model to predict the performance of
cizer) were modeled through RSM. To ensure the model, 24 con- self-compacting UHPC reinforced with hybrid steel micro-fibers, Constr. Build.
crete mixtures have been performed and the results have been Mater. 66 (2014) 375–383, https://doi.org/10.1016/
j.conbuildmat.2014.05.064.
compared with the results of the equations. After ensuring the
[4] Wiley: Response Surface Methodology: Process and Product Optimization
accuracy results of the equations, the diagrams of experiments Using Designed Experiments, 3rd Edition – Raymond H. Myers, Douglas C.
were plotted, optimized, and interpreted. Montgomery, Christine M. Anderson-Cook, 2009. http://eu.wiley.com/
WileyCDA/WileyTitle/productCd-0470174463.html.
[5] M.A. Bezerra, R.E. Santelli, E.P. Oliveira, L.S. Villar, L.A. Escaleira, Response
 Regarding the result of statistical analysis and validation of surface methodology (RSM) as a tool for optimization in analytical chemistry,
equations, it can be concluded that the RSM is an advantageous Talanta 76 (2008) 965–977, https://doi.org/10.1016/j.talanta.2008.05.019.
and efficient method for modeling and evaluating the rheologi- [6] I. Ferdosian, A. Camões, Eco-efficient ultra-high performance concrete
development by means of response surface methodology, Cem. Concr.
cal and hardened behavior of the SCM. Moreover, RSM can Compos. 84 (2017) 146–156, https://doi.org/10.1016/j.
reduce the number of experiments in comparison with similar cemconcomp.2017.08.019.
methods, such as the factorial, to obtain equations of responses. [7] M. Abu Yaman, M. Abd Elaty, M. Taman, Predicting the ingredients of self
compacting concrete using artificial neural network, Alexandria Eng. J. 56
 As a benefit of RSM, an optimum range was presented with (2017) 523–532, https://doi.org/10.1016/j.aej.2017.04.007.
respect to the integration of the accepted rheological criteria [8] H. Mashhadban, S.S. Kutanaei, M.A. Sayarinejad, Prediction and modeling of
for SCM (based on EFNARC code). The mixture design achieved mechanical properties in fiber reinforced self-compacting concrete using
particle swarm optimization algorithm and artificial neural network, Constr.
in this range includes a SCM with sufficient stability and low Build. Mater. 119 (2016) 277–287, https://doi.org/10.1016/
segregation in the fresh phase. j.conbuildmat.2016.05.034.
 According to the results, slag had no significant effect on the [9] F. Yan, Z. Lin, X. Wang, F. Azarmi, K. Sobolev, Evaluation and prediction of bond
strength of GFRP-bar reinforced concrete using artificial neural network
workability of SCM, but decreases the compressive strength. optimized with genetic algorithm, Compos. Struct. 161 (2017) 441–452,
Unlike slag, SF greatly affected the workability of the SCM and https://doi.org/10.1016/j.compstruct.2016.11.068.
improved compressive strength. The maximum effect of SF on [10] F. Yan, Z. Lin, New strategy for anchorage reliability assessment of GFRP bars to
concrete using hybrid artificial neural network with genetic algorithm,
the mini slump flow spread diameter varied up to 53% in the
Compos. B Eng. 92 (2016) 420–433, https://doi.org/10.1016/
presence of the high water-cement ratio. It should be noted j.compositesb.2016.02.008.
that, at the best state, SF could raise the discharge time of mini [11] A. Gholampour, A.H. Gandomi, T. Ozbakkaloglu, New formulations for
V-funnel up to 2 times. SF has enhanced the compressive mechanical properties of recycled aggregate concrete using gene expression
programming, Constr. Build. Mater. 130 (2017) 122–145, https://doi.org/
strength up to 45%. 10.1016/j.conbuildmat.2016.10.114.
 The increase in water-cement ratio without silica fume has led [12] M.M. Khotbehsara, B.M. Miyandehi, F. Naseri, T. Ozbakkaloglu, F. Jafari, E.
to growth of the SCMs segregation up to 10 times, while this Mohseni, Effect of SnO 2, ZrO 2, and CaCO 3 nanoparticles on water transport
and durability properties of self-compacting mortar containing fly ash:
increase at the level of 6% silica fume increased the segregation experimental observations and ANFIS predictions, Constr. Build. Mater. 158
up to 3 times. In the best case, the addition of silica fume (2018) 823–834, https://doi.org/10.1016/j.conbuildmat.2017.10.067.
reduced the segregation up to 5 times. [13] K. Samimi, S. Kamali-Bernard, A. Akbar Maghsoudi, M. Maghsoudi, H. Siad,
Influence of pumice and zeolite on compressive strength, transport properties
 It is noteworthy to mention that although both SP and water and resistance to chloride penetration of high strength self-compacting
could increase the flowability, SF is considerably more efficient concretes, Constr. Build. Mater. 151 (2017) 292–311, https://doi.org/
in the presence of high water. 10.1016/j.conbuildmat.2017.06.071.
[14] T.M. Grabois, G.C. Cordeiro, R.D. Toledo Filho, Fresh and hardened-state
 Silica fume, due to improved adhesive in the SCM, reduced seg- properties of self-compacting lightweight concrete reinforced with steel fibers,
regation and increased the stability. Using high amounts of SP Constr. Build. Mater. 104 (2016) 284–292, https://doi.org/10.1016/
and water causes instability in the SCM and their low amounts j.conbuildmat.2015.12.060.
[15] H. Fathi, T. Lameie, M. Maleki, R. Yazdani, Simultaneous effects of fiber and
also cause fluidity problems; so their consumption must be
glass on the mechanical properties of self-compacting concrete, Constr. Build.
more restricted. In this study, the appropriate ranges have been Mater. 133 (2017) 443–449, https://doi.org/10.1016/
presented by correcting these problems via the silica fume. The j.conbuildmat.2016.12.097.
best optimal mixture in terms of overlaying of the fresh results [16] F. Van Der Vurst, S. Grünewald, D. Feys, K. Lesage, L. Vandewalle, J. Vantomme,
G. De Schutter, Effect of the mix design on the robustness of fresh self-
belongs to the 0.6% superplasticizer diagram and the water- compacting concrete, Cem. Concr. Compos. 82 (2017) 190–201, https://doi.
cement ratio of 0.425% for different amounts of silica fume. org/10.1016/j.cemconcomp.2017.06.005.
M. Aziminezhad et al. / Construction and Building Materials 189 (2018) 1200–1213 1213

[17] S. Nunes, C. Costa, Numerical optimization of self-compacting mortar mixture [30] ASTM C150/C150M – Standard Specification for Portland Cement, in: ASTM
containing spent equilibrium catalyst from oil refinery, J. Cleaner Prod. 158 International, 2016.
(2017) 109–121, https://doi.org/10.1016/j.jclepro.2017.04.161. [31] EFNARC-Specification and Guidelines for self-compacting concrete, 2002.
[18] B. Benabed, E.-H. Kadri, L. Azzouz, S. Kenai, Properties of self-compacting [32] N.A. Libre, R. Khoshnazar, M. Shekarchi, Relationship between fluidity and
mortar made with various types of sand, Cem. Concr. Compos. 34 (2012) stability of self-consolidating mortar incorporating chemical and mineral
1167–1173, https://doi.org/10.1016/j.cemconcomp.2012.07.007. admixtures, Constr. Build. Mater. 24 (2010) 1262–1271, https://doi.org/
[19] W. Wongkeo, P. Thongsanitgarn, A. Ngamjarurojana, A. Chaipanich, 10.1016/j.conbuildmat.2009.12.009.
Compressive strength and chloride resistance of self-compacting concrete [33] M. Mahdikhani, A.A. Ramezanianpour, New methods development for
containing high level fly ash and silica fume, Mater. Des. 64 (2014) 261–269, evaluation rheological properties of self-consolidating mortars, Constr. Build.
https://doi.org/10.1016/j.matdes.2014.07.042. Mater. 75 (2015) 136–143, https://doi.org/10.1016/
[20] W. Schmidt, H.J.H. Brouwers, H.-C. Kühne, B. Meng, Influences of superplasticizer j.conbuildmat.2014.09.094.
modification and mixture composition on the performance of self-compacting [34] I. Mehdipour, M.S. Razzaghi, K. Amini, M. Shekarchi, Effect of mineral
concrete at varied ambient temperatures, Cem. Concr. Compos. 49 (2014) 111– admixtures on fluidity and stability of self-consolidating mortar subjected to
126, https://doi.org/10.1016/j.cemconcomp.2013.12.004. prolonged mixing time, Constr. Build. Mater. 40 (2013) 1029–1037, https://
[21] M. Omrane, S. Kenai, E.-H. Kadri, A. Aït-Mokhtar, Performance and durability of doi.org/10.1016/j.conbuildmat.2012.11.108.
self compacting concrete using recycled concrete aggregates and natural [35] BS 1881-116 – Method for determination of compressive strength of concrete
pozzolan, J. Cleaner Prod. 165 (2017) 415–430, https://doi.org/10.1016/j. cubes. British Standard, 2003.
jclepro.2017.07.139. [36] S. Yaseri, G. Hajiaghaei, F. Mohammadi, M. Mahdikhani, R. Farokhzad, The role
[22] R. Sharma, R.A. Khan, Sustainable use of copper slag in self compacting of synthesis parameters on the workability, setting and strength properties of
concrete containing supplementary cementitious materials, J. Cleaner Prod. binary binder based geopolymer paste, Constr. Build. Mater. 157 (2017) 534–
151 (2017) 179–192, https://doi.org/10.1016/j.jclepro.2017.03.031. 545, https://doi.org/10.1016/j.conbuildmat.2017.09.102.
[23] J. Bernal, E. Reyes, J. Massana, N. León, E. Sánchez, Fresh and mechanical [37] F. Moshiri, A. Atashnama, Mohammadreza Panahi, Rasoul Gharebaghi, J.
behavior of a self-compacting concrete with additions of nano-silica, silica Mahlouji, AN INVESTIGATION ON THE EFFECT OF CEMENT AND SILICA FUME
fume and ternary mixtures, Constr. Build. Mater. 160 (2018) 196–210, https:// ON PROPERTIES OF SELF-COMPACTING CONCRETE, Unpublished, 2013
doi.org/10.1016/j.conbuildmat.2017.11.048. [38] I. Mehdipour, K.H. Khayat, Effect of particle-size distribution and specific
[24] K.E. Alyamac, E. Ghafari, R. Ince, Development of eco-efficient self-compacting surface area of different binder systems on packing density and flow
concrete with waste marble powder using the response surface method, J. characteristics of cement paste, Cem. Concr. Compos. 78 (2017) 120–131,
Cleaner Prod. 144 (2017) 192–202, https://doi.org/10.1016/j. https://doi.org/10.1016/j.cemconcomp.2017.01.005.
jclepro.2016.12.156. [39] I. Mehdipour, K.H. Khayat, Understanding the role of particle packing
[25] I. Mehdipour, M. Vahdani, K. Amini, M. Shekarchi, Linking stability characteristics in rheo-physical properties of cementitious suspensions: a
characteristics to material performance of self-consolidating concrete- literature review, Constr. Build. Mater. 161 (2018) 340–353, https://doi.org/
equivalent-mortar incorporating fly ash and metakaolin, Constr. Build. Mater. 10.1016/j.conbuildmat.2017.11.147.
105 (2016) 206–217, https://doi.org/10.1016/j.conbuildmat.2015.12.090. [40] R. Zerbino, B. Barragán, T. Garcia, L. Agulló, R. Gettu, Workability tests and
[26] L. Soto-Pérez, V. López, S.S. Hwang, Response Surface Methodology to optimize rheological parameters in self-compacting concrete, Mater. Struct. 42 (2009)
the cement paste mix design: time-dependent contribution of fly ash and 947–960, https://doi.org/10.1617/s11527-008-9434-2.
nano-iron oxide as admixtures, Mater. Des. 86 (2015) 22–29, https://doi.org/ [41] X. Zhang, J. Han, The effect of ultra-fine admixture on the rheological property
10.1016/j.matdes.2015.07.049. of cement paste, Cem. Concr. Res. 30 (2000) 827–830, https://doi.org/10.1016/
[27] O. Rezaifar, M. Hasanzadeh, M. Gholhaki, Concrete made with hybrid blends of S0008-8846(00)00236-2.
crumb rubber and metakaolin: optimization using response surface method, [42] G.A. Rao, Role of water-binder ratio on the strength development in mortars
Constr. Build. Mater. 123 (2016) 59–68, https://doi.org/10.1016/ incorporated with silica fume, Cem. Concr. Res. 31 (2001) 443–447, https://
j.conbuildmat.2016.06.047. doi.org/10.1016/S0008-8846(00)00500-7.
[28] K. Mermerdasß, Z. Algın, S.M. Oleiwi, D.E. Nassani, Optimization of lightweight [43] K. Amini, I. Mehdipour, S.D. Hwang, M. Shekarchi, Effect of binder composition
GGBFS and FA geopolymer mortars by response surface method, Constr. Build. on time-dependent stability and robustness characteristics of self-
Mater. 139 (2017) 159–171, https://doi.org/10.1016/j.conbuildmat.2017. consolidating mortar subjected to prolonged agitation, Constr. Build. Mater.
02.050. 112 (2016) 654–665, https://doi.org/10.1016/j.conbuildmat.2016.02.226.
[29] Douglas C. Montgomery, Wiley: Design and Analysis of Experiments, 8th
Edition, 2012. http://eu.wiley.com/WileyCDA/WileyTitle/productCd-
EHEP002024.html (accessed July 17, 2016).

You might also like