Materials and Design 52 (2013) 957–965

Contents lists available at SciVerse ScienceDirect

Materials and Design

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

Evaluation of ultra-high-performance-fiber reinforced concrete binder

content using the response surface method
M.A.A. Aldahdooh, N. Muhamad Bunnori ⇑, M.A. Megat Johari
School of Civil Engineering, Universiti Sains Malaysia (USM), Malaysia

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

Article history: One of the major disadvantages in ultra-high-performance-fiber reinforced concrete (UHP-FRC) is its high
Received 11 April 2013 ordinary Portland cement (OPC) content, which directly translates into an increase in OPC production.
Accepted 14 June 2013 More OPC production results in increased emission of greenhouse gases, as well increased electrical
Available online 27 June 2013
energy consumption and concrete price. This study is aimed at adjusting the binder content (OPC and sil-
ica fume (SF) contents) of UHP-FRC using the response surface method. The present investigation shows
Keywords: that, for a given water/binder and superplasticizer/OPC, the compressive strength is independent of the
Ultra-high-performance-fiber reinforced
binder content, whereas the flow depends on the binder content. Increasing the binder content does not
concrete
Mechanical properties
enhance the strength compared with the required design strength because the capillary porosity
Central composite design method increases with increasing OPC content; however, the workability increases. The final result is the produc-
Compressive strength tion of a UHP-FRC with an OPC content of 720.49 kg/m3, an SF content of 214.25 kg/m3, a compressive
Workability strength of 181.41 MPa, a direct tensile strength of 12.49 MPa, a bending tensile strength of
30.31 MPa, and a flow of 167 mm.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction into greater OPC production, which correspondingly increases the

emission of certain greenhouse gases, namely, carbon dioxide
To date, the worldwide use of ultra-high-performance-fiber [12], sulfur dioxide [13], and nitrogen dioxide [13], in addition to
reinforced concrete (UHP-FRC) in structural applications has signif- increasing the electrical energy consumption and the price of
icantly increased owing to its numerous advantages compared concrete [12]. Hence, UHP-FRC products can be considered as
with conventional concrete such as normal concrete (NC) and uneconomical construction materials and pose a threat to the
high-performance concrete (HPC) [1,2]. UHP-FRC has superior environment.
properties such as enhanced strength, durability, and long-term Several methods can be employed to engineer concrete with
stability [1,3]. Moreover, UHP-FRC is more efficient in producing equivalent properties but with lesser OPC or binder content. One
smaller, lighter, and thinner sections [4] and possesses substantial method could be realized via optimizing the mix design of concrete
resistance to harsh environments and permeability [5]. using mathematical or statistical methods. Gopinath et al. [14]
UHP-FRC is characterized as a composite that has a large con- optimized the mixing proportions of NC and HPC using the particle
tent of ordinary Portland cement (OPC) and silica fume (SF), large packing method. Simon et al. [15] used a statistically designed
volume of steel fiber, low water/binder ratio, and high microsilica, mixture experiment to identify the best factor settings to optimize
as well as the absence of coarse aggregates, that is, larger than the properties of HPC. Furthermore, the mix proportion of reactive
4 mm [6]. UHP-FRC is defined as a concrete matrix with typical powder concrete can be optimized using the group method of data
compression strength of 150 MPa to 200 MPa, uniaxial tension handling and genetic programming [16]. De Larrard and Sedran
strength of 7–15 MPa, and bending strength of 25–40 MPa [7–9]. [17] optimized the ultra-high performance concrete (UHPC) mix
Furthermore, it exhibits high energy absorption capacity [10] and proportion using the density packing model. Another method of
strain hardening under tension [8]. developing economical concrete mixes involves adding or partially
Generally, an increase in the OPC content implies increase in the replacing the OPC with external ultrafine materials such as fly ash
overall demand for OPC [11,12]. This increase could be translated [18], SF [19], palm oil fuel ash [20], or ground granulated blast-fur-
nace slag [4].
In the present study, a multi-objective simultaneous optimiza-
⇑ Corresponding author. Address: School of Civil Engineering, Universiti Sains tion technique was used to adjust the binder content (OPC and
Malaysia, Engineering Campus, Nibong Tebal, 14300 Penang, Malaysia. SF) in the UHP-FRC mixture to yield an acceptable mechanical
Tel.: +60 4 5996259; fax: +60 4 5941009. strength. The response surface method (RSM) and three-level full
E-mail address: cenorazura@eng.usm.my (N. Muhamad Bunnori).

0261-3069/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.matdes.2013.06.034
958 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965

factorial experimental design were applied to satisfy these condi- An optimal predictor quadratic model [Eq. (1)] was used to
tions. The central composite design (CCD) method, considered as determine the optimal condition of the responses (compressive
one of the most common design methods under RSM [21], was also strength and flow) [23].
used. RSM has been widely employed in various fields such as in
the chemical industry for optimization and processing purposes. X
k X
k X
k X
k
Y ¼ bo þ bii X i þ bii X 2i þ bij X i X j þ ei ð1Þ
Nonetheless, the application of RSM in the concrete industry has
i¼1 i¼1 iij j
been lacking [4]. The limited studies include the work of Simon
[15], who employed RSM to optimize the HPC mixtures. Further- where Y is the predicted response, Xi and Xj are the coded values of
more, a more ductile concrete behavior could be obtained using the preparation variables, i is the linear coefficient, j is the quadratic
statistical RSM [4]. coefficient, b is the regression coefficient, k is the number of factors
studied and optimized in the experiment, and e is the random error.
1.1. Research significance

2.3. Mix proportions

Improving the mechanical properties of UHP-FRC relative to the
OPC and SF contents requires an ideal strategy. The effects of OPC
To identify the best factor setting for adjusting the binder con-
and SF on the compressive strength and workability of UHP-FRC
tent of UHP-FRC, a statistically designed mixture experiment was
were investigated to determine the ideal values of the design
conducted, as described in Section 2.2. The volume theory of con-
parameters using RSM. The results showed that this strategy can
crete mix design was employed based on the method proposed by
be used to adjust the binder content in the UHP-FRC mixture to
the American Concrete Institute Committee [24], where the sum of
yield acceptable mechanical strength using RSM.
the volume fractions is constrained to unity, that is, the total
amount (i.e., volume or mass) of any concrete mixture is fixed,
2. Experimental details and the factors are proportional to the total amount in the mixture
experiment [15].
2.1. Material properties Taking into consideration the effect of the OPC and SF contents
on the mechanical strength, RSM was applied to adjust the amount
The constituent materials include the following: OPC (American of the OPC and SF components based on their reasonable ranges.
Society for Testing and Materials [ASTM] Type I, 42.5R); densified The remaining components such as the water/binder (W/B)
micro SF with a particle size of 0.1–1 lm, 92% silicon dioxide, (0.18) [7,19], superplasticizer/cement ratio (SP/OPC) (0.07) [7,8],
and surface area of 23.7 m2/g; mining sand with particle size of and brass-coated steel fiber (6% total volume) [7,8] were deter-
100–1180 lm and specific gravity of 2.65 [22]; and two short mined from the preliminary study and extant literature, as shown
brass-coated microsteel fibers (6 and 13 mm) with diameters of in Table 2. A high volume fraction of the steel fiber was used so that
0.16 mm and tensile strength of up to 2850 MPa. The short fiber the resulting material (in its hardened state) will be very ductile
was used to enhance the flexural and tensile strength, whereas with a relatively high tensile strength [19].
the long fiber was used to increase the toughness [19]. To achieve In the final stage, the following equation was used to determine
a flow of 200 mm, a polycarboxylic ether-based superplasticizer the fine aggregate according to the American Concrete Institute
was used according to ASTM: C230M. Committee [24]:
  Xn
2.2. Design of experiments using statistical RSM kg
Sand 3
¼1 ðVÞk xðcÞk ð2Þ
m k¼1
Generally, RSM is a mathematical and statistical model used in
detecting the factor settings to optimize the average response where Vk is the volume fraction of each component k, (c)k is the
values. The most commonly used design method in finding the density of each component k, and n is the number of mixture com-
functional relationship between the response and the factors using ponents excluding the sand.
RSM is CCD [21]. In the current study, the Design-ExpertÒ 6.0.7 The final mix compositions of the UHP-FRC were determined
software (Sat-Ease Inc., Minneapolis, USA) was employed for the from the preliminary study, research literature, and statistical
design, mathematical modeling, statistical analysis, and optimiza- RSM, as presented in Table 2. Moreover, Table 2 shows that the
tion of the process variables. The interaction among variables amount of sand decreased as the total binder increased because
and the influence of each parameter were investigated using of the used of the volume theory of concrete mix design to deter-
ANOVA. mine the sand quantities for all mixes. The amount of sand de-
In the present study, the compressive strength (y1) and flow (y2) pends on the residual volume fractions of all components in each
denote the responses. Two factors were selected to describe this mix based on Eq. (2).
model: OPC content (x1) and SF content (x2). Based on the prelimin-
ary study and extant literature, the ranges of the selected factors
2.4. Specimen preparation
(parameters) are as follows: the OPC content is from 700 kg/m3
to 1000 kg/m3 [7,19], and the SF content is from 175 kg/m3 to
In the mixing procedure employed by Benson and Karihaloo
250 kg/m3 [7,16,19]. The coded and actual points and their
[19], OPC, mining sand, and SF were blended first. Thereafter, the
corresponding values are described in Table 1.
6 and 13 mm-long short steel fibers were added to the dry constit-
uents using vibrating aperture surfaces (sieves) of 5 and 12 mm,
Table 1 respectively, to achieve a uniform distribution. Proportional
Coded and actual values of independent variables. amount of water and superplasticizer was added to ensure mix
Variables Code Units Coded variable levels workability. The mixture was shaken until the constituents were
thoroughly mixed and fluid. All specimens were cast and com-
a 1 0 +1 +a
pacted on a vibration table. The samples were de-molded after
OPC X1 kg/m3 637.87 700 850 1000 1062.13
24 h. Thereafter, the recommended curing treatment for the
SF X2 kg/m3 159.47 175 212.50 250 268.53
UHP-FRC (steaming at 90 °C and 100% relative humidity for 48 h)
 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965 959

Table 2
UHP-FRC mix proportions.

Components (kg/m3) Mix1 Mix2 Mix3 Mix4 Mix5(i:v) Mix6 Mix7 Mix8 Mix9
OPC 637.9 700 700 850 850 850 1000 1000 1062.1
SF 212.5 175.0 250.0 265.5 212.5 159.5 250.0 175.0 212.5
M.S. 1184 1154 1028 778 868 957 581 708 552
W 153 158 171 201 191 182 225 212 229
SP 45 49 49 60 60 60 70 70 74
Steel fiber
lf1 = 6 mm 390 390 390 390 390 390 390 390 390
lf2 = 13 mm 78 78 78 78 78 78 78 78 78
W/OPC 0.24 0.23 0.24 0.24 0.23 0.21 0.23 0.21 0.22
W/B 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18 0.18
SP/OPC 0.070 0.070 0.070 0.070 0.070 0.070 0.070 0.070 0.070
SF/OPC 0.33 0.25 0.36 0.31 0.25 0.19 0.25 0.18 0.20
P
Vk 1 1 1 1 1 1 1 1 1

Cement and silica fume quantities were determined by RSM; sand quantity was determined by Eq. (2); the others components were determined from preliminary study and
literature, as a percentage from cement amount; steel fiber were selected according to [8].
Mix5(i:v) is referred to the number of repeat Mix 5 according to RSM.
M.S. = mining sand; W = water; SP = superplasticizer; B = binder content.

was applied [25]. The samples were then cured in water at direct tensile performance characteristics of the final mix were also
27 °C ± 2 °C until the testing day. investigated.

2.5. Test procedures

3.1. Workability
A compressive strength test was conducted for all mixes on the
In the mixture design strategy, we observed that the inclusion
4th, 14th, and 28th day. The other tests were conducted only for
of SF slightly decreased the workability of UHP-FRCs. Moreover,
the final mix on the 7th and 28th day to study the mechanical
the increase in the sand content decreased the workability, as
behavior of the mix.
shown Fig. 1. Higher workability is always associated with a higher
OPC content and lower SF (Fig. 1) and sand contents, as shown in
2.5.1. Compressive test
Fig. 2b. The reduction in the amount of sand is due to the use of
To determine the compressive strength of UHP-FRC, 100 mm
the volume theory of concrete mix design to determine the sand
concrete cubes were tested using a 3000-kN concrete compression
quantity for all mixes. The amount of sand depends on the residual
machine according to BS EN 12390-3 [26]. At least three samples
volume fractions of all components in each mix based on Eq. (2), as
from each mix (aged 4, 14, and 28 days) were tested.
described in Section 2.3.
The effect of SF was clear in Mix Nos. 3, 4, and 7, as shown in
2.5.2. Flow table test Fig. 1. The workability decreased slightly compared with those of
A flow table test was used to measure the rheology of UHP-FRC the others mixes with lower SF content. Jolicoeur and Simard
according to ASTM: C1437. The flow table was dropped 20 times [28] stated that SF has a high water demand and could adsorb a
for approximately 20 s. Once the concrete reached a steady state, low amount of superplasticizer because of the ‘‘surface area effect
the average diameter was determined by measuring the concrete phenomenon’’ and the ‘‘dilution effect phenomenon’’ that result
diameter at three points. from the larger surface area of SF and the formation of the aphtha-
lene sulfonate molecule adsorption, respectively. Therefore, the
2.5.3. Four-point bending test workability decreased.
For the bending tensile strength measurement, four-point The effect of the sand content on the reduction of UHP-FRC mix-
bending tests were conducted. The beams used to measure the ture workability is shown in Figs. 1 and 2b. The increase in the sand
bending tensile strength had a length of 279 mm, cross section of content notably decreased the workability of the mixes. The lowest
51 mm  51 mm, upper loading span of 76 mm, lower support flow value (<160 mm) was measured for Mix No. 1, which had the
span length of 229 mm [16], and crosshead loading rates of highest sand content. The highest flow value was more than
0.1 mm/min [27]. The test was performed using a 100-kN AG-X 250 mm (Mix No. 9), which had the lowest sand content and the
Shimadzu Universal Testing Machine. highest cement content. This result is in good agreement with
those of Yurdakul [29] and Shilstone and Shilstone [30], which re-
2.5.4. Direct tensile test vealed that insufficient amount of sand induces segregation of
To measure the direct tensile strength, dog bone-shaped speci- mixture. On the other hand, increasing the sand content cause
mens with a testing section length of 80 mm and cross section of the mixture to be stiff because the sand has a high water require-
16 mm  30 mm were used. The test was conducted using the ment owing to its high specific surface area. Moreover, the work-
same 100-kN AG-X Shimadzu Universal Testing Machine. The spec- ability decreased as the OPC content decreased for a given W/
imens were tested at a displacement rate of 0.2 mm/min [21]. OPC and aggregate content, because of the inadequate paste to
lubricate the aggregate [31,33].
3. Experimental results Based on the above results, increasing the workability above the
required design workability is normal because the W/B and SP/OPC
The tests were performed to investigate the effectiveness of the are fixed for all mixes, and the sand proportion decreased as the
OPC and SF contents in improving the compressive strength. The OPC content increased, as shown in Table 2 and Figs. 1 and 2b.
final mix was selected based on the highest compressive strength On the other hand, the workability increased as the OPC content
using statistical RSM. Then, the compressive, flow, flexural, and increased, as shown in Fig. 1. Mix Nos. 7–9 had the highest flow
960 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965

300 1200

1000

1000
1062
250 1000

850
850 850 850 850 850 850

Components (Kg/m3)
700
200 800

700

638
Flow (mm))
150 600

250.0
100

265.5
400

250.0
212.5

212.5

212.5

212.5

212.5

212.5

212.5
175.0

175.0
159.5
50 200

0 0
3 2 5iv 1 5i 5iii 5v 5ii 6 4 8 7 9
Number of mixes
cement silica fume Flow

Fig. 1. Flow of UHP-FRCCs mixes.

(a) 200 1200

1000

1000
1062
180
Compressive strength (MPa)

1000
160
850

Components (Kg/m3)
850 850 850 850 850 850
140
800
700
700

638

120
100 600
80
400
265.5
250.0

250.0
60
212.5

212.5

212.5

212.5

212.5

212.5
212.5
175.0

175.0
159.5

40
200
20
0 0
3 2 5iv 1 5i 5iii 5v 5ii 6 4 8 7 9
Number of mixes
Cement Silica fume 4-days 14-days 28-days

(b) 200 1400

1250

1275
1184
1154

1175
1116

180
1200
1028
Compressive strength (MPa)

160 1063 1063 1063 1063 1063

950

957

1009
Components (Kg/m3)

140 1000
868

868

868
875

850
868

868

778

120
800
708

100
581

552

600
80

60 400
40
200
20

0 0
3 2 5iv 1 5i 5iii 5v 5ii 6 4 8 7 9
Number of mixes
Sand Total binder 4-days 14-days 28-days

Fig. 2. Compressive strength development of UHP-FRC at the 4th, 7th, and 28th day: (a) effect of OPC and SF content and (b) effect of total binder and sand content.

(P220 mm) because of the high OPC and lower sand contents. The (sand and steel fibers) to move during the flow test. Thus, the
moderate flow values (Mix Nos. 4–6) were in the range of 185– workability increased.
195 mm, which had a moderate OPC content of 850 kg/m3 The steel fiber reinforced concrete with high workability has a
(Fig. 1) and sand content of 700–960 kg/m3 [Fig. 2b]. The lowest high probability of experiencing steel fiber segregation, which
flow value (Mix Nos. 1–3) were in the range of 158–170 mm at automatically reflects on the mechanical properties and perfor-
an OPC content of 6700 kg/m3 and sand content of P1027 kg/ mance of concrete [31]. Therefore, the OPC content should be
m3, as shown in Figs. 1 and 2b, respectively. Increasing the OPC balanced to achieve the highest performance with lower internal
content filled the gaps between the sand particles and steel fibers micro-cracks, shrinkage, (steel fiber [31], aggregate [29,32]) segre-
and provided lubrication to allow the mix composite materials gation, and pores [29,32].
 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965 961

3.2. Compressive strength 250 kg/m3. Moreover, high compressive strength values
(>170 MPa) are always associated with the range of the total bin-
Approximately 117 samples of 100-mm concrete cubes were der (OPC and SF) contents (850–1060 kg/m3), sand (870–1030 kg/
tested for the compressive strength. The specimens were tested m3), SP (45–60 kg/m3), and water (153–191 kg/m3), as shown in
on the 4th, 14th, and 28th day (Fig. 2). Table 2 and Fig. 2.
Mix No. 3 had the highest compressive strength. The compres- Compared with the results obtained in previous studies [7,19],
sive strength obtained after 4 days reached 170.2 MPa. Even on the RSM can be used to adjust the UHP-FRC mixture binder contents.
14th and 28th day, the compressive strength still improved. From
the 4th to the 28th day, the compressive strength increased by an- 3.3. Mathematical modeling and statistical analysis
other 3.9% to 176.9 MPa. Mix No. 9 exhibited the lowest gain in
compressive strength, achieving 128.7, 132.0, and 134.0 MPa after The interaction and relationship between the process factors
4, 14, and 28 days, respectively. (OPC and SF) and the responses (compressive strength and flow)
In this study, the compressive strength is a function of W/B and were obtained by ANOVA. The quality of the quadratic prediction
SP/OPC, as described in Table 2. Based on these data, Fig. 2 shows model was determined by the coefficient of determination R2.
that, for a given W/B and SP/OPC, the compressive strength is inde- The probability (P-value) with 95% confidence level and the statis-
pendent of the binder content. Increasing the binder content did tical significance test (t-test) at a 5% significance level (P < 0.05)
not enhance the strength compared with the required design were used to evaluate and verify the model terms. The model
strength. was generated according to the actual values.
The results of this study are consistent with those obtained in To establish the prediction model of the compressive strength,
previous studies [29,32], which concluded that the compressive only the 28th day average values of the compressive strength re-
strength decreased as the OPC content increased for a given W/ sponses were used. The empirical relationship between the re-
OPC because the capillary porosity increased and pores occurred sponses (compressive strength and flow) and the variables are
for chloride penetration [29,32]. In addition, the workability in- expressed by the following second-order polynomial equations:
creased as the OPC content increased for a given W/OPC and aggre-
Comressiv e strength ðMPaÞ ¼ 229:2226 þ 0:6680x1
gate content because of inadequate lubrication of the aggregate
[29,32], as discussed in the previous section. Alsadey [33] con- þ 1:5600x2  0:0004x21
cluded that the compressive strength decreases if the applied SP  0:0030x22  0:0004x1 x2 ð3Þ
dosage is beyond the optimum dosage because segregation and
bleeding phenomena will occur, which can affect the uniformity
FlowðmmÞ ¼ 355:3339  0:5441x1  0:0899x2 þ 0:0004x21 ð4Þ
and cohesiveness of concrete. Furthermore, the steel fiber-rein-
3 3
forced concrete with high workability has a high probability of where x1 (kg/m ) and x2 (kg/m ) are the model terms that charac-
experiencing steel fiber segregation, which automatically reflect terized the OPC and SF contents, respectively.
on the mechanical properties and performance of concrete [31,34]. The parameters and the interactions significantly affected the
Chan and Chu [35] explained that the highest interfacial bond compressive strength and flow, as shown in Eqs. (3) and (4),
strength and pullout energy between the steel fiber and concrete respectively. The ANOVA of both response surface quadratic mod-
matrix can be obtained at an SF dosage of up to 30%. When the els for the compressive strength and flow are presented in Table 3.
SF content increases to 40%, the interfacial toughening effect in Both compressive strength and flow models were significant
the bond strength vanishes. Based on this fact, the highest com- according to the t-test at a 5% significance level (P < 0.05), as shown
pressive strength is at an SF dosage of up to 30% (Mix Nos. 1–3) in Table 3. Furthermore, the models’ F-values of 159.10 and 516.72
as shown in Table 2. and low probability values imply that the models were significant
Fig. 2 shows that the highest compressive strength values are for the compressive strength and flow, respectively. Only a 0.01%
always associated with an OPC content of more than 700 kg/m3 chance exists that the model F-value of this magnitude can occur be-
but less than 850 kg/m3 and an SF content between 175 and cause of the noise. The F-value of 2.91 and 0.39 for the compressive

Table 3
ANOVA results for response surface quadratic model parameters.

Responses Source SOS DOF MS F P>F R

Compressive strength (MPa) at 28 days Model 2583.61 5 516.72 159.10 <0.0001 Sign.
x1 1907.37 1 1907.37 587.28 <0.0001 Sign.
x2 20.73 1 20.73 6.38 0.0394 Sign.
x21 571.80 1 571.80 176.06 <0.0001 Sign.
x22 120.75 1 120.75 37.18 0.0005 Sign.
x1x2 20.70 1 20.70 6.37 0.0395 Sign.
Residual 22.73 7 3.25
Lack of fit 15.59 3 5.20 2.91 0.1646 Not-sign.
Pure error 7.15 4 1.79
Cor. total 2606.34 12
Flow (mm) Model 7731.16 3 2577.05 583.33 <0.0001 Sign.
x1 6959.97 1 6959.97 1575.43 <0.0001 Sign.
x2 90.93 1 90.93 20.58 0.0014 Sign.
x21 680.27 1 680.27 153.98 <0.0001 Sign.
Residual 39.76 9 4.42
Lack of fit 12.96 5 2.59 0.39 0.8371 Not-sign.
Pure error 26.80 4 6.70
Cor. total 7770.92 12

SOS: sum of squares; DOF: degree of freedom; MS: mean square; P: probability of error; R: remark; F: F-value; Sign.: significant.
962 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965

Table 4 3.4. Graphical interpretation of the interaction between variables

Model validation for both responses.

Compressive strength (MPa) Flow (mm) The perturbation and the response surface plots for the
Santander deviation 1.80 2.10 compressive strength and flow are shown in Fig. 4A and B,
Mean 163.98 194.58 respectively.
R2 0.9913 0.9949 The comparative effect of the OPC content (kg/m3) (A) and SF
Predicted R2 0.9532 0.9902 content (kg/m3) (B) on enhancing the flow and compressive
Adjusted R2 0.9850 0.9932
Coefficient of variance 1.10 1.08
strength are clarified by the perturbation plots shown in
PRESS 122.01 76.07 Fig. 4A(a) and b, respectively. The curvature (A) is much sharper
Adequate precision 35.671 72.377 than the curve (B) in both tests (flow and compressive strength
tests), indicating that the flow and compressive strength were
PRESS: predicted residual error sum of squares.
more sensitive to the OPC content in this mixing procedure. The
OPC content increased, the compressive strength decreased, and
the workability increased.
The effects of the experimental conditions of the OPC and SF
strength and flow, respectively, implies that the lack of fit is not sig-
contents on the flow and compressive strength performance are
nificant relative to the pure error. The chances that the lack of fit of
demonstrated in the 3D surface response plots shown in
the F-value of this magnitude could occur because of the noise are
Fig. 4B(a) and b, respectively. Based on Fig. 4A and B it is noticed
16.46% and 83.71% for the compressive strength and flow,
that, that the highest compressive strength values are always asso-
respectively.
ciated with an OPC content of more than 700 kg/m3 but less than
The ANOVA revealed a reliable confidence in the estimation of
850 kg/m3 and an SF content between 175 and 250 kg/m3, while
the compressive strength and flow efficiency (R2 = 0.9913 and
the highest workability is always associated with an OPC content
0.9949, respectively). The predicted R2 of 0.9532 and 0.9902 are
of more than 1000 kg/m3and SF content of lower than 200 kg/m3.
in reasonable agreement with the adjusted R2 of 0.9850 and
0.9932 for both the compressive strength and flow, respectively
(Table 4). A high R2 coefficient indicates a satisfactory adjustment 3.5. Adjusting the binder content using the numerical optimization
of the quadratic models to the experimental data. process by RSM
The adequate precision of the signal-to-noise ratios of 35.671
and 72.377 for the compressive strength and flow, respectively, The results were evaluated using the Design-ExpertÒ software.
which are higher than 4, is desirable and confirms that the pre- In the numerical optimization, OPC and SF were set in the range
dicted models can be used to navigate the design space defined where the compressive strength is maximized, taking into consid-
by the CCD. eration the flow of >165 mm, as recommended by Graybeal [37]. At
Fig. 3 shows the normal plots of the residual values of the com- suitable conditions, the OPC content (720.49 kg/m3) and SF content
pressive strength and flow, which were used to help us determine (214.25 kg/m3) at a 167 mm flow and 179 MPa compressive
the model satisfactoriness [36]. Based on the adequacy of the mod- strength were predicted based on the desirability function of
els, the residuals from the least square fit are important, as shown 1.00, as shown in Fig. 5.
in Fig. 3. The constructed plot of the studentized residual versus Based on the mixing design procedure in Section 2, the amounts
the normal percentage of probability were satisfied because the of mining sand, steel fiber, water, and superplasticizer were deter-
compressive strength and the flow residual plots agreed well with mined and are presented in Table 5. Additional degradation test
the straight lines, as shown in Fig. 3. Consequently, this model can was performed under final binder content conditions based on the
be used to navigate the design space. results of the model to confirm the accuracy of the predicted model.

Fig. 3. Normal plots of residual values: (a) flow (mm) and (b) compressive strength for 28 days.
 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965 963

(a) (a)
250
231.215

Flow (mm)
214.781
228.75 A 198.348
181.914
Flow (mm)

165.481
207.5

B
186.25 B
250.00
1000.00
Si 231.25
A lic 925.00
af
um 212.50
165 850.00 3)
e(
kg 193.75 775.00 (kg/m
-1.000 -0.500 0.000 0.500 1.000
/m 3
m ent
) 175.00 700.00 Ce
Deviation from reference point

(b) A (b)
179

B
Compressive strength (MPa)

B
)

167
ength (MPa

156
Compressive str

A
145

)
/m 3
(kg
134

nt
me
Silic

Ce
-1.000 -0.500 0.000 0.500 1.000 a fu
me
Deviation from reference point (kg/
m 3)

(A) (B)
Fig. 4. (A) Perturbation plot for (a) compressive strength (MPa) at 28 days and (b) flow (mm); (B) response surface plots for (a) compressive strength (MPa) at 28 days and (b)
flow (mm).

The mechanical performance of the optimal UHP-FRC was At the 28th day, the average strength still increased by up to
tested by compressive, direct tensile, and four-bending tensile 12.49 MPa and 30.31 MPa, respectively (Table 5).
tests. The dimensions and sizes of the samples used to determine The results of the mechanical properties are consistent with
the mechanical properties of the final mix were described in Sec- those obtained in previous studies [7,8,19].
tion 2.5. The average value of the number of specimens, which var- Table 6 shows that the responses obtained from the model pre-
ied between three and four, was calculated at each testing day. The diction and laboratory experiment agree, thus verifying the accu-
results of the mechanical properties of the final mix are provided in racy of the predicted model.
Table 5. The compressive strength of the adjusted mix is higher than
For the compressive strength, the specimens were tested after that of the other mixes, as shown in Fig. 2, due to the SF effect,
7, 14, and 28 days. The other mechanical properties were tested where the optimal SF/C dosage reached 30% (Table 5). Fardis [5] re-
after 7 and 28 days (Table 5). After 7 days of hardening, the com- ported that the UHPC production requires the SF dosage to be at
pressive strength increased by up to 175.75 MPa. Even at the 10–30% of the OPC mass for two reasons: to achieve maximum
14th and 28th day, the compressive strength still increased. From packing density wherein the SF fills the voids among the OPC
the 7th to the 28th day, the compressive strength increased by an- particles, and to replace the calcium hydroxide (Ca[OH]2) with cal-
other 3.2% to 181.41 MPa. cium silicate hydrates (CSH), which has a higher strength. CSH is
For the direct and bending tensile strength, the average strength created during the chemical reaction between pure SiO2 of the SF
after 7 days reached up to 11.31 MPa and 28.80 MPa, respectively. and Ca(OH)2. This reaction automatically decreases the porosity
964 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965

Fig. 5. Ramp report on numerical optimization of compressive strength for 28 day and flow.

Table 5
Details of the final UHP-FRC mix proportion and properties based on RSM.

OPC SF Sand W SP Steel fiber W/OPC W/B SF/OPC

l1 l2
Composition (kg/m3) 720.49 214.25 1057.31 168.30 50.43 390 78 0.23 0.18 0.30

Mechanical tests on optimal mix

Time Compressive strength (MPa) Direct tensile strength (MPa) Bending tensile strength (MPa)
7 days 175.75 11.31 28.80
14 days 179.53 – –
28 days 181.41 12.49 30.31

OPC: refereed to OPC content and SF: refereed to SF content were adjusted by RSM.
Flow (mm) = 167 mm.

Table 6
Final response results from the model prediction and laboratory tests.

Parameters Flow (mm) Compressive strength (MPa) Des.

OPC (kg/m3) SF (kg/m3) Obs. Pre. Error (%) Obs. Pre. Error (%)
720.49 214.25 167 170.32 1.94 181.41 179 1.35 1

Obs.: referred to observed value; Pre.: referred to predicted value; Des.: referred to desirability.

of the bulk concrete matrix, particularly at the interfacial zones. sand particles and steel fibers and provide lubrication to allow the
Chan and Chu [35] explained that the highest interfacial bond mixed materials (sand and steel fibers) to move during the flow
strength and pullout energy between the steel fiber and concrete test. Thus, the workability increased.
matrix can be obtained at an SF dosage of up to 30%. When the The adjusted values of the design factors when the compressive
SF content increases to 40%, the interfacial-toughening effect in strength was involved were as follows: OPC content of 720.49 kg/
the bond strength vanishes. m3 and SF content of 214.25 kg/m3.
In the mixing design method, the sand and binder contents had
4. Conclusions significant effects on the strength, with the highest compressive
strength always associated with the range of the total binder
The present investigation has shown that, for a given W/B and (OPC and SF) contents (850–1060 kg/m3) and sand (870–1030 kg/
SP/OPC, the compressive strength is independent of the binder m3). The adjusted SF/OPC ratio was approximately 30%.
content. Increasing the binder content did not enhance the For the adjusted mix, the average compressive strength, flow,
strength compared with the required design strength because the direct tensile strength, and bending tensile strength after 28 days
capillary porosity increases with increasing OPC content. More- were 181.41 MPa, 167 mm, 12.49 MPa, and 30.31 MPa,
over, for a given W/B and SP/OPC, the workability is dependent respectively.
on the OPC content. Increasing the OPC content increases the flow Although the results indicate that the prediction by RSM was
because the increase in the OPC content filled the gaps between the satisfactory in adjusting the amount of OPC and SF in the
 M.A.A. Aldahdooh et al. / Materials and Design 52 (2013) 957–965 965

production of UHP-FRC materials, these values still need to be [16] Haghighi A, Koohkan MR, Shekarchizadeh M. Optimizing mix proportions of
reactive powder concrete using group method of data handling and genetic
reduced further, meaning that another method could be used to
programming. In: 32nd Conf Our World Concr Struct. Singapore; 2007.
reduce the OPC and SF contents through partial replacement of [17] De Larrard F, Sedran T. Optimization of ultra-high-performance concrete by
OPC by external ultrafine (or by-product) materials. the use of a packing model. Cem Concr Res 1994;24(6):997–1009.
[18] Mehta PK. High-performance, high-volume fly ash concrete for sustainable
development. In: Uo California, editor. International workshop on sustainable
Acknowledgments development and concrete technology. California, Berkeley, USA: Kejin Wang
Beijing; 2004. p. 3–14.
The authors wish to acknowledge the Universiti Sains Malaysia [19] Benson S, Karihaloo B. CARDIFRCÒ-development and mechanical properties.
Part I: Development and workability. Mag Concr Res 2005;57(6):347–52.
(USM) for providing the financial support through the USM fellow- [20] Sata V, Jaturapitakkul C, Kiattikomol K. Influence of pozzolan from various by-
ship scheme [APEX (1002/JHEA/ATSG/4001)], Short Term Grant product materials on mechanical properties of high-strength concrete. Constr
(304/PAWAM/6039047), and Postgraduate Research Grant Build Mater 2007;21(7):1589–98.
[21] Montgomery DC. Design and analysis of experiments. Wiley; 2008.
Scheme (1001/PAWAM/8045050). [22] Tayeh BA, Abu Bakar B, Megat Johari M, Voo YL. Mechanical and permeability
properties of the interface between normal concrete substrate and ultra high
References performance fiber concrete overlay. Constr Building Mater 2012;36:538–48.
[23] Auta M, Hameed BH. Optimized waste tea activated carbon for adsorption of
Methylene Blue and Acid Blue 29 dyes using response surface methodology.
[1] Brühwiler E, Denarié E. Rehabilitation of concrete structures using ultra-high
Chem Eng J 2011;175:233–43.
performance fibre reinforced concrete. In: Proc, Second Int Symp Ultra-high
[24] Committee A, Institute AC, Standardization IOf. Building code requirements for
Perform Concr; 2008.
structural concrete (ACI 318-08) and commentary. American Concrete
[2] Iskhakov I, Ribakov Y, Holschemacher K, Mueller T. High performance
Institute; 2008.
repairing of reinforced concrete structures. Mater Des 2013;44:216–22.
[25] Toledo Filho RD, Koenders EAB, Formagini S, Fairbairn EMR. Performance
[3] Diao B, Yang S. Influence of curing regime on the ductility of ultra-high
assessment of ultra high performance fiber reinforced cementitious
performance fiber reinforced concrete (UHPFRC). ICCTP 2009:1–7.
composites in view of sustainability. Mater Des 2012;36(0):880–8.
[4] Bayramov F, Tasßdemir C, Tasßdemir M. Optimisation of steel fibre reinforced
[26] British Standard Institution. BSI EN 12390-3, Testing hardened concrete.
concretes by means of statistical response surface method. Cement Concr
Compressive strength of test specimens. London: BSI; 2002.
Compos 2004;26(6):665–75.
[27] Li VC, Horikoshi T, Ogawa A, Torigoe S, Saito T. Micromechanics-based
[5] Fardis MN. Innovative materials and techniques in concrete construction: ACES
durability study of polyvinyl alcohol-engineered cementitious composite. ACI
Workshop. University of Patras: Springer Dordrecht Heidelberg London New
Mater J 2004;101(3):242–8.
York; 2012.
[28] Jolicoeur C, Simard MA. Chemical admixture-cement interactions:
[6] Wille K, Naaman AE, Parra-Montesinos GJ. Ultra-high performance concrete
phenomenology and physico-chemical concepts. Cement Concr Compos
with compressive strength exceeding 150 MPa (22 ksi): a simpler way. ACI
1998;20(2):87–101.
Mater J 2011;108(1):46–54.
[29] Yurdakul E. Optimizing concrete mixtures with minimum cement content for
[7] Spasojević A. Structural implications of ultra-high performance fibre-
performance and sustainability. Ph.D Thesis, Iowa State University; 2010.
reinforced concrete in bridge design. Ph.D. thesis, EPFL; 2008.
[30] Shilstone JM, Shilstone J. Performance-based concrete mixtures and
[8] Benson S, Karihaloo B. CARDIFRCÒ-development and mechanical properties.
specifications for today. Concr Int 2002;24(2):80–3.
Part III: Uniaxial tensile response and other mechanical properties. Mag Concr
[31] Schnütgen B, Vandewalle L. In: Proc 31: international RILEM workshop on test
Res 2005;57(8):433–43.
and design methods for steel fibre reinforced concrete-background and
[9] Richard P, Cheyrezy M. Composition of reactive powder concretes. Cem Concr
experiences. RILEM Publications; 2003.
Res 1995;25(7):1501–11.
[32] Wassermann R, Katz A, Bentur A. Minimum cement content requirements: a
[10] Habel K, Gauvreau P. Response of ultra-high performance fiber reinforced
must or a myth? Mater Struct 2009;42(7):973–82.
concrete (UHPFRC) to impact and static loading. Cement Concr Compos
[33] Alsadey Salahaldein. Influence of superplasticizer on strength of concrete. Int J
2008;30(10):938–46.
Res Eng Technol 2012;1(3).
[11] Worrell E, Price L, Martin N, Hendriks C, Meida LO. Carbon dioxide emissions
[34] lucic vandewalle. Design with r–e method, In: Proc 31: international RILEM
from the global cement industry 1. Annu Rev Energy Environ
workshop on test and design methods for steel fibre reinforced concrete-
2001;26(1):303–29.
background and experiences. RILEM Publications; 2003.
[12] Arshad MF. Influence of multiple blended binders on engineering properties
[35] Chan Y-W, Chu S-H. Effect of silica fume on steel fiber bond characteristics in
and durability of concrete Ph.D thesis. Malaysia: Universiti Sains Malaysia;
reactive powder concrete. Cement and Concrete Research.
2010.
2004;34(7):1167–72.
[13] Thomas HB. Particle packing; an effective approach to optimized design of
[36] Ghafari S, Aziz HA, Isa MH, Zinatizadeh AA. Application of response surface
ultra high strength and self compacting concretes. Australia: University of
methodology (RSM) to optimize coagulation–flocculation treatment of
New South Wales; 2011.
leachate using poly-aluminum chloride (PAC) and alum. J Hazard Mater
[14] Gopinath S, Murthy AR, Ramya D, Iyer NR. Optimised mix design for normal
2009;163(2):650–6.
strength and high performance concrete using particle packing method. Arch
[37] Graybeal BA. Material property characterization of ultra-high performance
Civ Eng 2011;4:357–71.
concrete. Ph.D thesis. McLean; 2007.
[15] Simon MJ, Lagergren ES, Snyder KA. Concrete mixture optimization using
statistical mixture design methods. In: Proc PCI/FHWA Int Symp High Perform,
Concr; 1997. p. 230–44.