Construction and Building Materials 159 (2018) 234–251

Contents lists available at ScienceDirect

Construction and Building Materials

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

Mechanical performance of roller compacted concrete pavement

containing crumb rubber and nano silica
Bashar S. Mohammed ⇑, Musa Adamu
Department of Civil and Environmental Engineering, Universiti Teknologi PETRONAS, 32610 Bandar Seri Iskandar, Malaysia

h i g h l i g h t s

Nano silica lead to improve properties of RCR.

Nano silica densifies the microstructure of RCR.

Models have been developed using RSM to predict the strengths of RCR.

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

Article history: In this study, roller compacted rubbercrete (RCR) was developed by partially replacing fine aggregate
Received 10 August 2017 with crumb rubber (CR) in roller compacted concrete pavement (RCCP) to improve its flexural resistance,
Received in revised form 10 October 2017 ductile behavior and reduces the environmental effect of waste tires. Due to the fact that crumb rubber
Accepted 23 October 2017
reduces strength in concrete, the addition of nano silica (NS) to mitigate the loss in strength of RCR is a
possible solution. Sixteen mixtures with four levels of CR (0%, 10%, 20%, and 30%) by volume replacement
to fine aggregate and four levels of NS (0%, 1%, 2% and 3%) addition by weight of cementitious materials
Keywords:
were prepared and tested. Fresh properties including density and Vebe consistency time have been deter-
Crumb rubber
Nano silica
mined. Hardened properties including unit weight, compressive, flexural and splitting tensile strength,
Roller compacted rubbercrete elastic modulus, abrasion resistance and water absorption have been investigated. The findings show that
Response surface methodology CR increases the consistency of RCR. The compressive strength and abrasion resistance of RCR increases
Compressive strength with 10% CR and the flexural strength also increases with replacement of up to 20% fine aggregate with
Flexural strength CR. Addition of NS was successful in improving the performance of RCR due to its physicochemical effects
which have been verified by carrying out microstructural analysis. Mathematical models developed using
Response surface methodology for predicting the strengths and water absorption using CR and NS as the
input variables shows a high degree of correlation and predictability. The results of multi-objective opti-
mization showed that, an optimum mixture can be achieved with a 10% volume replacement of fine
aggregate and 1.13% NS addition by weight of cementitious materials to have a high strength, durable
and ductile RCR for pavement applications.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction volume poses serious environmental and health hazards if they are
not properly disposed or recycled [19]. Tires ingredients contain
The needs in transportation continue to escalate globally due to certain chemicals that help in making their service life to last
the rapid increase in human populations. This in conjunction with longer. However, these chemicals used make the waste tires gener-
advancement in technology results to increase in the quantity of ated to be non-biodegradable, and if disposed into landfills leaches
vehicle production and usage which consequently results in the toxic substances to the ground thereby causing land pollution, and
rapid generation of waste tires [19]. About 1 billion waste tires water pollution in the presence of a nearby water source [47].
are generated annually in more than half of it disposed of without Another problem related to waste tires are they are highly flam-
pretreatment, and the number is expected to reach 1.2 billion by mable with higher calorific properties making them very difficult
2030 [53]. These waste tires which are mostly disposed of in high to extinguish if burnt, with the production of black fume and high
temperature causing air pollution and increasing global warming
by emitting more CO2 to the environment [13]. However, if not
⇑ Corresponding author.
burnt disposed waste tires provides breeding grounds for all sort
E-mail address: bashar.mohammed@utp.edu.my (B.S. Mohammed).

https://doi.org/10.1016/j.conbuildmat.2017.10.098
0950-0618/Ó 2017 Elsevier Ltd. All rights reserved.
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 235

Table 1 Table 2
Chemical properties of materials. Properties of nano silica.

Oxides Composition (%) Cement Fly ash Item Quality

SiO2 20.76 57.06 Appearance High-dispersive white powder
Al2O3 5.54 20.96 Hear reduction (%) (105 °C 2 h)  3
Fe2O3 3.35 4.15 Loss of ignition (%) (950 °C 2 h)  6
MnO – 0.033 SiO2 content (dry base) (%)  92
CaO 61.4 9.79 SiO2 content (%) (950 °C 2 h)  99.8
MgO 2.48 1.75 Specific surface area (m2/g) 100 ± 25
Na2O 0.19 2.23 PH value 6.5–7.5
K2O 0.78 1.53 Surface density (g/ml)  0.15
TiO2 – 0.68 Dispensability (%) (%) (CCl4)  80
Loss of ignition 2.2 1.25 Oil-absorbed value (ml/100 g)  250
Specific gravity 3.15 2.3 Average particle size (nm) 10–25
Blaine fineness (m2/kg) 325 290 Hydrophobicity Strong

of rodents, mosquitoes, snakes etc. thereby increasing the risk of

transferring and spreading deadly diseases such as Zika virus, den-
gue fever, malaria fever [13]. In order to effectively utilized and
manage waste tires, it is mostly used in the production of tire
derived fuel (pyrolysis), production of carbon black, fuel for a
cement kiln. However, the main disadvantage of tire-derived fuel
is that its production involves emission of large amount of CO2 to
the environment and it is less economical compared to fuel from
petroleum products [53]. It is also used for ground applications
such as; asphalt for paving streets and highways, sports surfacing,
animal bedding, fill material for turf grasses [51,9]. The civil engi-
neering industry utilizes a little amount of waste tire generated in
the US where it stands at just 7% as of 2015 [48]. With the con-
struction industry been one of the sectors that affect sustainability
negatively by consuming a lot of natural resources such as aggre-
gate [8], where river sand which is the commonest material used Fig. 2. XRD pattern of nano silica.
for fine aggregate is rapidly depleting and becoming scarce and
more expensive due to its high demand [6]. In order to remedy
the challenges facing the construction industry in terms of sustain- Lower Limit (ACI) Combined Aggregate Upper Limit (ACI)
ability of materials and issues related to waste tire disposal prob- 100
lems, several researchers incorporated the waste tires as partial
90
replacement to aggregate in concrete by grinding the tires into
80
Percentage Passing (%)

smaller sizes in form of crumb rubber (CR) with the fibers and steel
70
removed [56].
60
Crumb rubber (CR) when used in concrete has been found to
reduce the durability performance and mechanical properties of 50
concrete [59]. The major cause for strength reduction is attributed 40
to the poor adhesion/bonding the hardened cement matrix and CR 30
which is caused by the smooth surface and lower hydraulic con- 20
ductivity CR. The poor bonding can also be due to the presence 10
of zinc stearate during the production of the tire which conse- 0
quently diffuses to the rubber surface during mixing in concrete 0.01 0.1 1 10 100
Sieve Sizes (mm)

Fig. 3. Combined aggregate gradation.

7.0
6.8
Flexural Strength (MPa)

6.6
6.4
6.2
6.0
5.8
5.6
5.4
5.2
5.0
11 12 13 14 15 16
Cement Content (%)

Fig. 1. Particle size analysis of aggregate. Fig. 4. Relationship between cement content and flexural strength of RCCP.
236 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

making the CR entrap air and repel water [58]. This result in a and yielded so many positive effects such as reduction in traffic
higher amount of air inside the fresh mix thus leading to increase noise, increasing lifespan of the pavement and lowering cost of
void content and porosity in the hardened concrete thereby maintenance [13].
reduces its durability performance and strengths [31,58]. In addi- Fakhri [13] has used silica fume as cement replacement to mit-
tion, the entrapped air on the CR surface increases the interfacial igate the loss in strength in roller compacted concrete made by
transition zone (ITZ) thickness, and causes poor bonding with the partially replacing fine aggregate with crumb rubber. He has
hardened matrix as the ITZ itself is a weak porous zone between reported a reduction in fresh density and water absorption. This
cement paste and aggregate [28], with the weak bonding causing reduction has been found to be even higher when silica fume par-
micro cracks formation leading to reduced strength and lead to tially replaced cement. He has also reported increasing in the com-
premature failure [49,54]. Therefore, for CR to be effectively used pressive strength by 7% and 9% for 5% and 10% of CR replacement
in concrete the strength loss need to be mitigated. To improve respectively and increase in flexural strength by 9% for 5% of CR.
adhesion between CR particles and hardened cement paste silica The increasing in strength is higher when 10% silica fume has been
fume or mortar to coat the CR has been used [35,38]. Youssf [59] used.
have treated crumb particles with sodium hydroxide solutions to Several studies have been carried out on utilizing CR to partially
make its surface rougher [35,22,59]. While, Ossola and Wojcik replace fine aggregate in producing concrete. Similarly, strength
[40] have changed the surface energy of CR particles through ultra- and durability of RCC pavement are affected negatively by increas-
violet by water retention method to prevent the crumb rubber ing CR content. By capitalizing on the advantages of inclusion CR in
from entrapping air. Other researchers have used pozzolanic mate- the RCC and to mitigate the side effects, silica fume has been used
rials such as, metakaolin, fly ash, and silica fume as partial replace- to produce RCR with good properties, however, the results was not
ment to cement to improve chemical and pozzolanic reaction encouraging. Therefore, in this paper, nano silica has been utilized
thereby increasing strength [41,42,28]. Mohammed et al. [28] have as an addition to cement to mitigate the loss in strengths of RCC
utilized nano-silica in the dry mixture of concrete containing pavement where CR partially replaced fine aggregate.
crumb. They have reported that nano silica has densified the inter-
facial tranzission zone between crumb rubber particles and the
hardened cement paste which resulting in higher compressive
strength. 2. Materials and methods
Roller compacted concrete pavement (RCCP) is a sustainable
and economical material used for construction of rigid pavement. 2.1. Materials
The cost of pavement construction when RCCP is used can be
15%–30% lower than if conventional rigid pavement or ashpaltic In this study, ordinary Portland Type I cement with chemical
pavement [43]. In terms of environmental sustainability RCCP properties shown in Table 1 and conforming to the requirements
requires lower cement content to achieve desired cement which of ASTM C150M-15 has been used. Natural sand having a specific
greatly reduced cement consumption, CO2 emission during gravity of 2.65, water absorption of 1.24% and fineness modulus
cement production and heat of hydration during mixing [7]. On of 2.86 has been used. Two nominal maximum size aggregates
the other hand, some of the deficiencies of RCCP include lower ten- (NMSA) which are firstly 19 mm size having 2.66 a specific gravity
sile strength, high rigidity, possibility of cracking due to drying and 0.48% water absorption, and secondly 6.35 mm size with a
shrinkage [16]. In addition due to no reinforcing bar used, all specific gravity of 2.55 and absorption of 1.05%. To obtain the sizes
applied loads on the RCCP are transferred through aggregate inter- of CR comparable to that of fine aggregate, three different sizes of
lock to the lower courses, with the corresponding effects such as crumb rubber have been used. The sieve analysis has been con-
tensile cracks, fatigue cracks, fatigue damage, expansion and con- ducted in accordance to the requirements of ASTM D5644, in
traction, thermal and shrinkage cracks are all resisted by the RCC which several trials of combinations in sizes have been performed
concrete strength alone [2]. In order to reduce the brittleness nat- until a gradation curve similar to that of fine aggregate has been
ure of RCCP and improve its ductility and bending resistance and to obtained. The percentage proportions of the three different crumb
delay crack propagation, materials such as fibers, polymers and rubber sizes are 40% of 0.595 mm (mesh 30), 40% of 1–3 mm size,
crumb rubber can be added to absorb deformation and strain and of 20% 3–5 mm size. Fig. 1 shows the particle gradation of
energy [24]. Crumb rubber has been used in asphaltic pavement crumb rubber.

Table 3
Constituent materials for RCR mixtures.

Mixture Quantities for 1 kg/m3 of RCR

Cement Nano silica Filler Fine aggregate Coarse aggregate 19 mm Coarse aggregate 6.35 mm Water CR
R0-0 268.69 0 103.76 1148.05 415.03 416.85 98.24 0
R0-1 268.69 2.69 103.76 1148.05 415.03 416.85 98.24 0
R0-2 268.69 5.37 103.76 1148.05 415.03 416.85 98.24 0
R0-3 268.69 8.06 103.76 1148.05 415.03 416.85 98.24 0
R10-0 268.69 0 103.76 1033.25 415.03 416.85 98.24 114.89
R10-1 268.69 2.69 103.76 1033.25 415.03 416.85 98.24 114.89
R10-2 268.69 5.37 103.76 1033.25 415.03 416.85 98.24 114.89
R10-3 268.69 8.06 103.76 1033.25 415.03 416.85 98.24 114.89
R20-0 268.69 0 103.76 918.44 415.03 416.85 98.24 229.78
R20-1 268.69 2.69 103.76 918.44 415.03 416.85 98.24 229.78
R20-2 268.69 5.37 103.76 918.44 415.03 416.85 98.24 229.78
R20-3 268.69 8.06 103.76 918.44 415.03 416.85 98.24 229.78
R30-0 268.69 0 103.76 803.64 415.03 416.85 98.24 344.67
R30-1 268.69 2.69 103.76 803.64 415.03 416.85 98.24 344.67
R30-2 268.69 5.37 103.76 803.64 415.03 416.85 98.24 344.67
R30-3 268.69 8.06 103.76 803.64 415.03 416.85 98.24 344.67
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 237

40 211.3R-02 [1]. The first method is to obtain a combined grading

34 35
35 33 33 32 32 of fine and coarse aggregate to be within the limits recommended
30 31 30
30 27 28 27 27 by ACI 325.10R and ACI 211.3R. A combination of 55% fine aggre-
Vebe Time (s)

25 24 25
25 gate, 20% of 19 mm maximum sized coarse aggregate, 20% of 6.3
20 mm coarse aggregate, and 5% mineral filler has been selected and
15 the combined aggregate gradation is plotted as shown in Fig. 3.
10 The optimum moisture contents and maximum dry densities for
5 RCC mixes with 12%, 13%, 14% and 15% by weight of dry aggregate
0 have been determined according to the requirements of ASTM D
1557-12e. The optimum moisture content for RCC with 12%, 13%,
14% and 15% cement has been found to be 5.46%, 5.56%, 5.92%
Mixture and 6.09%, respectively. Subsequently, the relationship between
cement content and flexural strength was determined as presented
Fig. 5. Vebe time for RCR mixes. in Fig. 4, this has been done by producing four RCC mixes with 12%,
13%, 14% and 15% cement content and water content equal to their
OMC and then tested for flexural strengths at age of 28 days. Based
2500
on target flexural strength of 4.8 MPa, 13% cement content has
2450 2427
2410 been selected to be used in this study. A water/cement ratio of
2400 2386
Fresh Density (kg/m3)

2376 2363 2355 0.42 has been established based on the quantities of the selected
2350 2333 2333
2304 materials. To increase reduce the water content and increase con-
2300 2261
2265 2232 2244 sistency, 1% superplasticizer by weight of cementitious materials
2250 2227 2223 2232
was added and the water content has been reduced by 12%, bring-
2200 ing the water/cement ratio down to 0.37.
2150 In order to study the effect of hybrid of CR and NS on the perfor-
2100 mance of RCC pavement, sixteen mixes have been prepared with
2050 different ratios of CR (0%, 10%, 20% and 30%) as partial replacement
to fine aggregate by volume, while NS has been added at varying
percentages (0%, 1%, 2% and 3%) by weight of cementitious materi-
Mixture
als. Each mixture is given a unique ID as shown in Table 3. For
Fig. 6. Fresh density of RCR mixes. instance, mixture R20-3 is an RCR with 20% of CR and 3% of NS.

One of the requirements for RCC pavement mixture production

is using materials of sizes lower than 75 mm (No 200 sieve) to
achieve a more cohesive paste with reduced void volume. The rec- 2.3. Samples preparations and test methods
ommended amount should be between 2% and 8% of the total
aggregates, and materials such as naturally occurring non-plastic In this study, the Bosch vibration hammer was used to simulate
silt, fine sand or Pozzolanas can be used [10,1]. Therefore, fly ash the compaction of the samples in the laboratory in accordance to
which conforms to the requirements of ASTM C612 and ASTM the requirements of ASTM C1435. The consistency and fresh den-
C311 has been used as mineral filler with properties presented in sity of the RCR mixes have been determined according to ASTM
Table 1. Commercially available nano silica having size of 10–25 C1170 [5].
nm with strongly hydrophobicity has been used as addition to The compressive strength test was carried out on 100 mm
cementitious materials addition; its properties are presented in cubes according to BS 12390-3:2009, for each mix a total of 12
Table 2. Fig. 2 shows that the nano silica is amorphous in nature cubes were produced and tested at 3, 7, 28 and 90 days; in which
with a high percentage of SiO2. for each curing period three samples have been tested and the
mean value has been recorded.
The indirect tensile strength of RCR has been measured using
2.2. Mix proportions splitting tensile test according to the requirements of BS EN
12390-6:2009 in which total of 12 cylinders of 100 mm diameter
Proportioning of RCR mixes has been carried out according to by 200 mm heights have been prepared for each mixture and
the geotechnical approach conforms to the requirements of ACI tested at curing age of 3, 7, 28 and 90 days.

3 Days 7 Days 28 Days 90 Days

90
3 Days 7 Days 28 Days 90 Days
Compressive Strength (MPa)

80
Splitting tensile strength (MPa)

7
70
6
60
5
50
4
40
3
30
2
20
1
10
0
0
R0-0

R0-1

R0-2

R0-3

R10-0

R10-1

R10-2

R10-3

R20-0

R20-1

R20-2

R20-3

R30-0

R30-1

R30-2

R30-3

Mixture Mixture

Fig. 7. Compressive Strength of RCR. Fig. 8. Splitting tensile strength of RCR.

238 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

7 Days 28 Days 90 Days R20-1 R20-2 R20-3

10 R30-1 R30-2 R30-3
Flexural strength (MPa)

9 45
8
7 40
6
5 35
4 30
3

Stress (MPa)
2 25
1
0 20
15

Mixtures 10
5
Fig. 9. Flexural strengths of RCR.
0
0 1000 2000 3000 4000 5000
Strain x 10-6 mm/m
40
Modulus of elasticity (GPa)

35 Fig. 11b. Stress-strain curve for RCR mixtures with 20%–30% CR.
30 7 Days 28 Days
25
2.5
20

Water absorption (%)

15 2
10
5 1.5
0
1

0.5
Mixture
0
Fig. 10. Modulus of elasticity of RCR.

Mixture

Fig. 12. Water absorption of RCR.

R0-0 R10-1 R10-2 R10-3
70
60 10
50 9
Stress (MPa)

100 Rev 200 Rev 300 Rev

8
40
Cantabro Loss (%)

7
30
6
20 5
10 4
0 3
0 500 1000 1500 2000 2500 3000 2
Strain x 10-6 mm/m 1
0
Fig. 11a. Stress-strain curve for RCR mixtures with 0%–10% CR.

Mixture Designation
Flexural strength was carried out in accordance with require-
ments outlined by ASTM C293M-10 with 9 beams produced for Fig. 13. Cantabro loss of RCR mixes.
each mix and tested at curing age of 7, 28 and 90 days.
The modulus of elasticity was determined using 150 mm diam- C642-13. For each mix, three samples have been tested and the
eter by 300 mm height cylinders according to ASTM C469 at 7 days mean value has been recorded.
and 28 days. All samples are tested in triplicate and the average Finally, abrasion resistances of the control and mixtures con-
value has been recorded. The modulus of elasticity is computed taining crumb rubber has been determined at curing age of 28 days
using Eq. (1). using Cantabro test method, in which Los Angeles abrasion
machine has been used with the steel balls removed. This method
Ec ¼ ðr2  r1 Þ=ðe2  0:00005Þ ð1Þ
is in accordance to the requirements of ASTM C1747 and also has
where Ec is the modulus of elasticity; r2 is the stress equivalent to been implemented by several researchers [11,44]. In this study,
40% of ultimate compressive force; r1 is the strain corresponding to three cylindrical samples of 100 mm heights and 150 mm diame-
a longitudinal strain of 50  106; and e2 is the longitudinal stress ters have been tested for each selected mixture and the average
corresponding to r2. value has been recorded. The abrasion resistance has been mea-
Water absorption has been measured using 100 mm cubes at sured at 100, 200 and 300 revolutions. The Cantabro loss (abrasion
age of 28 days curing according to the requirements of ASTM resistance) is calculated using Eq. (2).
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 239

W1  W2 The result shows a reduction by 0.96%, 6.66% and 6.45% for mixes
C L ð%Þ ¼  100 ð2Þ
W1 R10-0, R20-0 and R30-0, respectively in comparison R0-0 (control).
The reduction is mainly caused by the lower bulk density of CR in
where CL is the Cantabro abrasion loss; W1 is the initial weight of
comparison to fine aggregate. The addition of Nano silica (NS)
sample (g); W2 is the final weight of sample (g).
however has slightly increased the fresh density in some mixes
and decreases in some mixes. For example, 1% NS addition to
3. Results and discussion 10%, 20%, and 30% CR changes the percentage decrease to 2.22%,
5.24% and 6.83%, respectively. The partial increment in density
3.1. Effect of crumb rubber and nano silica on fresh properties of RCR with NS addition can be attributed to the ability of the nanoparti-
cles to densify the pore structure at the fresh stage, while the
3.1.1. Vebe consistency time reduction in density might be due to agglomeration of nanoparti-
The effect of CR and NS addition on the Vebe time of RCR is cles which causes air to be entrapped within the agglomerate.
shown in Fig. 5. Increase in partial replacement of fine aggregate
with crumb resulted to decrease in Vebe time of RCR. There is a 3.2. Effect of crumb rubber and nano silica on hardened properties of
decrease in the Vebe time by 9.1%, 24.2% and 27.3% for R10-0, RCR
R20-0, and R30-0, respectively in comparison to R0-0 control). This
is attributed to the hydrophobic nature of CR as it repels water thus 3.2.1. Compressive strength
increase the quantity of free water in the mix, and consequently The result of compressive strength at all ages of curing and for
reducing the compaction time and effort needed [23]. However, all RCR mixes is shown in Fig. 7. The addition of crumb rubber (CR)
NS addition is slightly increasing the Vebe consistency time. For decreases compressive strength, for instance, mixtures R20-0 and
example, mix R20-1, R20-2 and R20-3; the Vebe consistency time R30-0 decreases by 16.3% and 23.2%, respectively as compared to
was increased by 8%, 12%, and 20%, respectively in comparison to R0-0. The loss in strength of RCR with incorporation of CR is attrib-
R0-0. The increase is attributed to the higher surface area of NS uted to the increased thickness of the interfacial transition zone
making it absorb part of the free water available in the mixture. (ITZ) between the hardened paste matrix and CR particles which
causing poor bonding between them and leading to micro cracks
3.1.2. Fresh density formation and consequently premature failure [28,54].
The result of fresh density for each RCR mixtures is shown in The compressive strength of RCR significantly increased with
Fig. 6 and it decreases with increasing crumb rubber (CR) content. the addition of NS. As presented in Fig. 7, compressive strength

a) R0-0 b) R10-0

c) R10-1 d) R10-2
Fig. 14. FESEM images for RCR mixtures.
240 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

e) R20-0 f) R20-2

g) R30-0 h) R30-2
Fig. 14 (continued)

increases significantly with the addition of 1% and 2% NS and caused by the poor adhesion (bond) between the rubber particles
decreases with 3% of NS addition. For example, the 28 days’ com- and hardened cement matrix causing micro cracks formation and
pressive strength of mixes R10-1 and R10-2 increases by 25% and sudden loss in strength [54].
11.5%, respectively, while R10-3 decreases by 16.7% compared to The addition of NS has been successfully mitigating the loss in
R0-0. The increment in strength with NS addition is attributed to split tensile strength in RCR. The amount of NS from 1% to 3%
the following reasons; NS has the ability to fill up nano voids and increases the split tensile strength with 1% NS being as optimum
also refining the pore system making the hardened micro structure amount for 0%, 10%, and 30% CR content. For example, the split ten-
of RCR mix denser. Furthermore, the high pozzolanic reactivity of sile strength of mix R30-1 and R30-2 increases by 21.4% and
NS enabling it reacts and consumes the surplus Portlandites (Ca 18.75% respectively compared to R30-0. This increase is due to
(OH)2) and produces extra calcium-silicate-hydrate (CASAH) gel nano silica enhancing the bond between the hardened cement
which is the main compound for strength development and densi- paste and aggregates, which is as result of reaction between the
fication of the interfacial transition zone [28]. While the reduction reaction silica and excess calcium hydroxide to produce calcium
in compressive strength is due to higher content of NS percentage. silicate hydrate [14].
This causes agglomeration of nanoparticles triggered by insuffi-
cient water available due to higher surface area of nano silica [37].
3.2.3. Flexural strength
The result of flexural strength of RCR is shown in Fig. 9. The 28
3.2.2. Splitting tensile strength days’ flexural strength of mixes R10-0 and R20-0 and R30-0 has
The result of split tensile strength for all RCR mixes is presented increased by 39.3% and 9.3% respectively compared to R0-0 control
in Fig. 8. The 28 days split tensile strength of mixture R10-0 mix). Similar findings were reported by Yilmaz and Degirmenci
increases by 18.7% while that of R20-0 and R30-0 decreases by [57]. The increase in flexural strength is mainly due to higher bend-
15% and 29.4%, respectively, compared to R0-0 control mix). The ing deformation and fiber nature of CR, which gives the RCR post-
increase in tensile strength is due to the ability of crumb rubber cracking behavior and allows it to resist some flexural load even
to elongate and act as micro fiber thereby bridging cracks in the after failure [52].
RCR hence inhibiting total disintegration from occurring [39]. The The addition of NS to RCR increases its flexural strength. The
reduction in split tensile strength with higher percentage of CR is optimum percentage of NS was found to be 1% for all CR contents.
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 241

The 28 days’ flexural strength of R10-1, R20-1, and R30-1 increases at higher percentage levels above 10% decreases the brittleness
by 4.71%, 6.07%, and 3.90% respectively compared to R10-0, R20-0 nature and increases the flexibility of RCR hence reduction in
and R30-0 respectively. However, above 1% NS, the 28 days’ flexu- MOE. At 10% CR replacement the MOE has increased by 1.2% and
ral strength of R0-2 and R30-2 increased by 16.63% and 2.71% com- decreased by 41.9% and 45.3% for 20% and 30% CR, respectively.
pared to R0-0 and R30-0 respectively, and decreases for all other These results were similar to the findings by other researchers
mixtures. The increase in flexural strength of RCR with the addition [26,29,18]. The increment in MOE for 10% CR is due to the higher
of NS is due to strengthening and densification of the ITZ between compaction pressure used which reduces the effect of increased
CR-cement paste and aggregate due to increased pozzolanic reac- porosity caused by crumb rubber. While the decrease in MOE is
tion by NS [34]. The addition higher amount of NS increases the attributed to higher deformation and lower elastic modulus of CR
water demand to achieve proper consistency and hydration pro- in comparison to fine aggregate [18].
cess, due to its large surface area. With, RCR having low water con- The addition of NS increases the MOE for all CR replacement
tent, this resulted in reduction in hydration and pozzolanic levels as shown in Fig. 10. At 20% CR replacement level, the 28
reaction, and also causes agglomeration of nano particles and poor days MOE of RCR increases by 24.77%, 2.47% and 15.83% for 1%,
consistency, which consequently reduces the flexural strength of 2%, and 3% NS addition, respectively compared to 0% NS. Similar
RCR [25]. findings were reported by [4]. This increasing is due to the pore
filling ability, microstructural refinement and ability of NS to
3.2.4. Modulus of elasticity of RCR react with Ca(OH)2 from cement hydration to produce more
The results of modulus of elasticity (MOE) for all the RCR mixes calcium-silicate-hydrate which increases strength and conse-
are shown in Fig. 10. Partial replacement of fine aggregate with CR quently MOE [4].

a) R0-0 b) R10-0

c) R10-1 d) R10-2
Fig. 15. Pore size distribution for RCR. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
242 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

e) R20-0 f) R20-1

g) R20-2 h) R30-0
Fig. 15 (continued)

The stress-strain curves of selected RCR mixes are shown in 3.2.5. Water absorption
Fig. 11a and Fig. 11b, initially before reaching maximum stress, The water absorption is used to evaluate the durability perfor-
all the mixes have similar behave. After reaching the maximum mance of RCR and the results are shown in Fig. 12. The water
stress, mixes with lower CR contents suddenly reach their breaking abruption rate increases with increase in percentage replacement
points without much strain softening; this shows the brittleness of fine aggregate with crumb rubber. The water absorption of
nature of the RCR, with lower compressive toughness. While for R10-0, R20-0, and R30-0 increases by 12.42%, 20.92%, and 31.37%,
mixes with higher CR contents, they undergo prolonged strain soft- respectively compared to R0-0. This increase is caused by an
ening before failure with higher peak strain as shown by the increasing in pore volume of hardened RCR caused by air
extended parts of the curve after the maximum stress is reached. entrapped on crumb rubber surface during mixing [45]. The addi-
This further demonstrates how CR increases the ductility behavior tion of nano silica decreases the water absorption of RCR as shown
of RCR. in Fig. 11. For example, the water absorption of R20-1, R20-2, and
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 243

R20-1, R20-2, and R20-3 increases by 8.45%, 11.57%, and 7.26%,

respectively compared to R20-0. Similar findings were reported
by Nazari and Riahi [36]. This increasing is due to the ability of
NS to increase pozzolanic reaction, fill the pores in RCR and densify
the interfacial transition zone between crumb rubber particles and
cement matrix, hence resulting in increasing in strength and abra-
sion resistance [36].

3.3. Microstructural analysis

3.3.1. FESEM
The microstructural configurations for different RCR mixtures
were studied using field emission scanning electron microscope
(FESEM) as shown in Fig. 14. One of the factors that cause a reduc-
tion in strength when CR is utilized in RCR is the poor bonding
between CR and cement paste. Another factor is that increasing
in thickness of the ITZ between hardened cement paste and CR.
The ITZ is a porous boundary between cement paste and aggregate
mostly of thickness ranging from 10 to 50 mm, with the formation
of micro cracks in the ITZ, causing premature failure and conse-
quently reduction in strength [28]. Due to the high-pressure com-
i) R30-2 paction applied to RCR and low water/cement ratio, the thickness
of the ITZ and the porosity of the cement matrix are low. The addi-
Fig. 15 (continued) tion of NS densified the ITZ between CR and cement paste, it also
consumes the unreacted Ca(OH)2 and produces more CASAH gels,
and refined the microstructure of RCR. This is due to the high poz-
R20-3 decreases by 13.51%, 5.95%, and 20%, respectively compared zolanic reaction of NS [50]. As shown in Fig. 14a, so many excess Ca
to R20-0. This increase is caused by the ability of nano silica to (OH)2 on the surface, when NS is added it reacts with the excess Ca
refine the pore system of the hardened RCR microstructure [15]. (OH)2 and produces more CASAH gel as shown in Fig. 14c, which
consequently densified the ITZ between hardened cement paste
3.2.6. Cantabro abrasion resistance and CR and refined the pore structure. However, when a higher
The abrasion resistance of any RCC pavement is an important amount of NS was added above 1%, agglomeration of nanoparticles
parameter since its surface will always be exposed to wearing from occurs as shown in Fig. 14g and is caused by lack of sufficient water
vehicle tires. The abrasion resistance of RCR is determined using in the fresh RCR mix for the NS to properly react due to its high sur-
the Cantabro loss. The lower the percentage weight loss the higher face area. This causes poor bonding between the CR and cement
the abrasion resistance. As shown in Fig. 13, replacement of fine paste and makes the microstructure of the cement matrix more
aggregate with 10% crumb rubber enhances the abrasion resis- porous thus leading to drop in the strength. Increasing the CR con-
tances of RCR. The mechanism is that the crumb rubber projected tent increases the voids in hardened RCR and thickness of the ITZ
to the surface of the RCR and act like a brush on the surface of RCR as seen in Fig. 14e–i. These findings are in agreements with those
thereby restricting grinding/rubbing [55]. However, at 20% and 30% reported by Li et al. [20] and Mohammed et al. [28].
replacement levels, the abrasion resistance decreases significantly
and is mainly due to weak adhesion between crumb rubber parti- 3.3.2. Mercury intrusion porosimetry
cles and hardened cement paste even at the surface. The mercury intrusion porosimetry (MIP) test was used to study
The addition of nano silica decreases the weight loss in RCR the effect of physiochemical reaction of NS on RCR which resulted
hence increases the abrasion resistance for all number of revolu- to pore structure refinement in the hardened matrix of RCR by fill-
tions. For example, at 200 revolutions the abrasion resistance of ing the voids within the interconnected pores thereby breaking the

Table 4
Developed experimental design and responses.

Run Factors Responses

Coded values Actual values
A: Crumb B: Nano A: Crumb B: Nano Compressive Strength Flexural Strength Splitting Tensile Water
Rubber (%) Silica (%) Rubber (%) Silica (%) (MPa) (MPa) Strength (MPa) Absorption (%)
1 1 1 30 0 40.41 5.73 3.36 2.01
2 0 1 20 0 44.02 6.13 4.05 1.85
3 0 0 20 1 54.92 6.33 4.7 1.61
4 1 1 10 2 67.01 7.33 5.87 1.61
5 0 0 20 1 53.23 6.66 4.78 1.53
6 1 0 10 1 75.11 8.18 6.13 1.34
7 0 0 20 1 49.69 6.48 4.73 1.66
8 1 1 30 2 41.57 5.89 3.99 1.88
9 0 0 20 1 49.69 6.7 4.82 1.64
10 1 1 10 0 60.11 7.81 5.65 1.72
11 0 1 20 2 49.85 5.88 4.19 1.74
12 1 0 30 1 44.95 5.96 4.08 1.91
13 0 0 20 1 51.79 6.32 4.67 1.55
244 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

Table 5
ANOVA for response models.

Variable Factors Sum of squares df Mean square F -Values P-Values

Compressive Strength (MPa) Model 1157.46 5 231.49 34.88 <.0001
A-Crumb rubber 945.02 1 945.02 142.40 <.0001
B-Nano silica 32.16 1 32.16 4.85 .0636
AB 8.24 1 8.24 1.24 .2030
A2 127.78 1 127.78 19.25 .0011
B2 109.39 1 16.48 16.48 .0011
Lack of Fit 25.79 3 1.66 1.66 .0566
Flexural Strength (MPa) Model 6.86 5 1.37 45.89 <.0001
A-Crumb rubber 5.49 1 5.49 183.64 <.0001
B-Nano silica 0.054 1 0.054 1.81 .2204
AB 0.10 1 0.10 3.42 .1067
A2 1.08 1 1.08 36.26 .0005
B2 0.53 1 0.53 17.76 .0040
Lack of Fit 0.082 3 0.027 0.86 .5308
Splitting Tensile Strength (MPa) Model 7.63 5 1.53 134.52 <.0001
A-Crumb rubber 6.45 1 6.45 568.73 <.0001
B-Nano silica 0.16 1 0.16 14.41 .0068
AB 0.042 1 0.042 3.71 .0956
A2 0.63 1 0.63 55.48 .0001
B2 0.71 1 0.71 62.81 <.0001
Lack of Fit 0.065 3 0.022 5.91 .0594
Water Absorption (%) Model 0.36 5 0.072 12.25 .0024
A-Crumb rubber 0.21 1 0.21 36.38 .0005
B-Nano silica 0.020 1 0.020 3.49 .1039
AB 0.0001 1 0.0001 0.017 .8997
A2 0.0001 1 0.0001 0.17 .6952
B2 0.098 1 0.098 16.83 .0046
Lack of fit 0.028 3 0.0009 2.97 .1600

*df: degree of freedom; F-values: Fisher statistical test values, P- values: Probability values

Table 6
Model validation for developed models.

Response Compressive strength (MPa) Flexural strength (MPa) Splitting tensile strength (MPa) Water absorption (%)
Mean 52.49 6.57 4.69 1.70
SD 2.58 0.17 0.11 0.076
CoV (%) 4.91 2.63 2.27 4.51
R2 0.9614 0.9704 0.9897 0.8975
Predicted R2 0.7722 0.8598 0.9279 0.2346
Adjusted R2 0.9339 0.9492 0.9823 0.8242
Adequate precision 18.441 20.572 39.379 12.102

interconnectivity and reducing its diameter. Nano silica can be able The addition of NS further reduced the total pore volumes of
to fill pores in rubbercrete up to nano sizes. Examples of these RCR mixes. The total pore volume of mixes R20-1, and R20-2 as
pores are the capillary pores and the gel pore. The capillary pores shown in Fig. 15f and Fig. 15g respectively were found to decrease
occur when excess water during mixing evaporates from the hard- compared to that of R20-0 shown in Fig. 15e. This is due to the abil-
ened RCR as it dries, and can be as large as 1 mm to as low as 100 ity of NS to densify and refine the pore structure of RCR up to nano
nm. The gel pores occurred on the hydration products (calcium sil- sizes bottom-up), which is as a result of increased pozzolanic reac-
icate hydrate gels) when water evaporates and has diameter lower tion between silicon dioxide and unreacted calcium hydroxide
than 3 nm [28]. from cement hydration products to produce more calcium silicate
Fig. 15 shows the plots of cumulative pore volumes against pore hydrate gels [28].
diameters for RCR mixtures. The blue curves represent values of
the cumulative pore volumes, while the red lines show the range 3.4. Mathematical and statistical models using response surface
of values for the pore diameters of the RCR. The porosity of RCR methodology
increases with increment CR content in RCR as shown in
Fig. 15a, b, e, and h. This findings is similar with the outcomes pre- 3.4.1. Model development
vious studies [28]. The increased pore volume is caused by the Response surface methodology (RSM) is commonly used statis-
non-polar nature of CR which makes it to repel water and entrap tical and mathematical technique used for analyzing and develop-
air during mixing, thereby increasing the porosity of hardened ing models between one or more independent variables and
RCR [28]. However, the use of high pressure compaction effort lim- responses [32]. In addition, RSM can be used for model multi-
its and restricts air entrainment in RCR; this makes the increment objective optimization by setting defined desirable goals based
in pore volume to be minimal. This also makes the total pore vol- on either responses or variables [32]. RSM have been used in con-
ume for all RCR mixtures to be low as shown in Fig. 15(b–i), as the crete for development of models and mixtures optimization.
high compaction effort helps in densifying and refining its micro Mohammed et al. [30] have utilized RSM to model the compressive
structure. Furthermore, the pore sizes distribution for all the RCR strength of concrete containing paper mill as additives. Mtarfi et al.
mixtures ranges from 10 to 100,000 nm. [33], have optimized and developed model for predicting mortar
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 245

Predicted vs. Actual Predicted vs. Actual

8.20
76.00

66.25 7.55

Predicted
Predicted

56.50 6.90

46.75 6.25

37.00 5.60

37.90 47.20 56.50 65.81 75.11 5.62 6.26 6.90 7.54 8.18

Actual Actual

a) Compressive strength b) Flexural strength

Predicted vs. Actual

6.30
Predicted vs. Actual
2.05

5.55
1.87
Predicted

Predicted

4.80 1.69

4.05 1.52

1.34
3.30

1.34 1.52 1.69 1.87 2.05

3.36 4.08 4.79 5.51 6.22
Actual
Actual
c) Splitting tensile strength d) Water absorption

Fig. 16. Predicted vs. actual plots for developed models.

compressive strength with RSM. Güneyisi et al. [17] have devel- between the variables and measured responses. Multi-objective
oped models and optimized high-performance concrete by mini- optimization was also carried out to determine the optimized
mizing the durability factors and maximizing compressive RCR mixtures by maximizing strengths and minimizing water
strength using metakoalin and fly ash as variables. Rezaifar et al. absorption.
[46] developed models for predicting the compressive strength, There are different models which can be used to derive mathe-
unit weight and water absorption of rubbercrete using RSM. They matical relationships between responses and independent vari-
have also optimized the rubbercrete mixture by maximizing the ables in the RSM analysis. The central composite design model is
compressive strength and minimizing water absorption. the most commonly used and more reliable model. This is because
Mohammed et al. [27] developed mix design model for self- of the choice of selection of a which is the distance from the axial
compacting engineered cementitious composites (SC-ECC) using run to the design center, with a depending on the number of points
RSM. They also optimized ECC mixtures by maximizing modulus in the factorial design portion [12].
of elasticity and energy absorption. As the mathematical relationships between the responses and
In this study, RSM was used to evaluate the combined effect of independent variables are unknown, the model can be formed as
both CR and NS on the properties of RCR, and develop relationship a linear model given by the first order function in Eq. (3) [12].
246 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

76
Compressive Strength

8.2

Flexural Strength
66.25
7.525

56.5 6.85

46.75 6.175

37 5.5

2
2

30
1 1 30

B: Nano Silica B: Nano Silica

20
20

0 10 0 10
A: Crumb Rubber
A: Crumb Rubber

a) Compressive strength b) Flexural strength

6.4 2.05
Splitting Tensile Strength

5.625 1.8725
Water Absorption

4.85
1.695

4.075
1.5175
3.3

1.34
2

2 30

1 30
1 20
B: Nano Silica
20
B: Nano Silica A: Crumb Rubber
0 10
0 10
A: Crumb Rubber

c) Splitting tensile strength d) Water absorption

Fig. 17. Response surface plots for developed response models.

y ¼ b0 þ b1 X 1 þ b2 X 2 þ . . . bn X n þ  ð3Þ coefficient, b is the regression constant, b0 is the y-intercept for

which Xi = Xj = 0, k is the number of variables used for the analysis,
where y is the modeled response, b0 is the y-intercept for which X1
and e is error [12].
= X2 = 0, b1, and b2 are the coefficients of the first and second inde-
There are different models which can be used in DOE to derive
pendent variables respectively, X1 is the first variable coefficient
mathematical relationships between responses and independent
and X2 the second variables and e is the error [12].
variables. The central composite design model is the most com-
However, if the data contains curvature, then the linear model
monly used and more reliable model this is because of the choice
will not be suitable for the response, then a polynomial model of
of selection of a which is the distance from the axial run to the
higher degrees should be used as given by the second order func-
design center, with the value of a depending on the number of
tion given in Eq. (4).
points in the factorial design portion [12].
X
k X
k XX In this study, a commercially available software package (Stat-
y ¼ b0 þ bi X i þ bii X 2i þ bij X i X j þ  ð4Þ
Ease Design Expert version10) was utilized for the RSM analysis.
i¼1 i¼1 i< j
The face centered central composite design (FCCCD) with a = 1
where y is the modeled response, xi and xj are the independent has been used to develop the mathematical models for 28 days’
variable coded values, i is the linear coefficient, j is the quadratic compressive strength, flexural strength, splitting tensile strength
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 247

Com press ive Strength Flexural Strength

2 2
43.5194

5.99838
B: Nano Silica

B: Nano Silica
66.1573 6.41285
7.2418
1 5 6.82732
54.8384 1 7.65627 5
60.4978 49.1789

43.5194
5.99838

0 0
10 20 30 10 20 30
A: Crumb Rubber A: Crumb Rubber

a) Compressive strength b) Flexural strength

Splitting Tensile Strength Water Absorption

2 2
1.83657
B: Nano Silica
B: Nano Silica

1.51892

5.74862
5.27355
1 5
4.79847 4.3234 1 5 1.6248

1.73069

1.83657
3.84832
1.94245

0 0
10 20 30 10 20 30

A: Crumb Rubber A: Crumb Rubber

c) Splitting tensile strength d) Water absorption

Fig. 18. Contour plots for developed models.

and water absorption, using crumb rubber (A) and nano silica (B) (A * B) were not significant for compressive and flexural strength
as the independent variables. A total of 13 combinations based models. For splitting tensile strength all the model terms were sig-
on the variables levels were developed as shown in Table 4. nificant except the interaction of crumb rubber and nano silica
(A * B). While for water absorption model, all the terms were
3.4.2. Statistical models and analysis of variance insignificant except A and B2. The lack of fit is also used to check
The relationships between the independent variables and each the goodness of fit for the models. As seen in Table 5,
response were achieved by ANOVA as shown in Table 5. The 5% sig- P-significance for the lack of fit for all the models was greater than
nificance level (P < .05) is used to check if a model is significant. .05, meaning it is not significant, and this implies all the response
From Table 5 the P-values for all the response models were less models has a good fitness. The relationships and interactions
than .05, therefore, they are all significant models at 95% confi- between the variables (CR and NS in terms of their actual values)
dence level. Similarly, the statistical significance of each of the and the responses as obtained by the ANOVA are given in Eqs.
independent variable in the model has been checked at 0.05 signif- (5a)c
icance level (P < .05). The Fisher statistical test values (F-values) of
F C ¼ 93:08  3:832  A þ 17:772  B  0:144  AB
34.88, 45.89, 134.52, and 36.38 for compressive strength, flexural
strength, splitting tensile strength, and water absorption models þ 0:068  A2  6:293  B2 ð5aÞ
further explains their significance, and the chance of an F-value
of this size to occur due to noise is 0.01%. As shown in Table 5 F F ¼ 10:879  0:362  A þ 0:462  B þ 0:016  AB
and Eqs. 5(ad), all the response models are the quadratic. The
model terms A, A2 and B2 were siginifant while B and interaction þ 0:0063  A2  0:438  B2 ð5bÞ
248 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

Perturbation
Perturbation
8.2
76 A

A
Compressive Strength

7.575

Flexural Strength
67

6.95
58

6.325
49 B B A
A
B
B
5.7
40
-1.000 -0.500 0.000 0.500 1.000
-1.000 -0.500 0.000 0.500 1.000
Deviation from Reference Point (Coded Units)
Deviation from Reference Point (Coded Units)

a) Compressive strength b) Flexural strength

Perturbation
Perturbation
6.3 A
2.01
Splitting Tensile Strength

5.55 B
1.8425
Water Absorption

A
B
4.8
1.675
B
A
4.05 B
1.5075

A
3.3
1.34
-1.000 -0.500 0.000 0.500 1.000
-1.000 -0.500 0.000 0.500 1.000
Deviation from Reference Point (Coded Units)
Deviation from Reference Point (Coded Units)

c) Splitting tensile strength d) Water absorption

Fig. 19. Perturbation plots for developed models.

Table 7
Optimization criteria.

Variables & Responses Goal Lower limit Upper limit

Crumb rubber (%) In range 10 30
Nano Silica (%) In range 0 2
Compressive strength (MPa) Maximize 40.41 75.11
Flexural strength Maximize 5.73 8.18 0.94
Splitting tensile strength (MPa) Maximize 3.36 6.13
Water absorption (%) Minimize 1.34 2.01 0.705
Desirability

0.47

Table 8 0.235
Results of optimization.
0
Variables & Responses Optimum ratio and predicted responses
Crumb rubber (%) 10 10
0
Nano silica (%) 1.13
Compressive strength (MPa) 71.80 1 20
Flexural strength (MPa) 8.04 A: Crumb Rubber
Splitting tensile strength (MPa) 6.22
B: Nano Silica
2 30
Water absorption (%) 1.41
Desirability (%) 94.3
Fig. 20. Desirable combination of crumb rubber and nano silica.
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 249

Table 9
Model validation.

Response Solutions CR (%) NS (%) Predicted result Experimental result Error (%)
Compressive strength (MPa) 1 10 1.13 71.80 69.46 3.37
2 15 1.5 60.17 62.97 4.45
3 20 1.5 51.84 52.78 1.78
Flexural strength (MPa) 1 10 1.13 8.04 7.89 1.90
2 15 1.5 6.92 6.63 4.37
3 20 1.5 6.33 6.11 3.60
Split tensile strength (MPa) 1 10 1.13 6.22 5.94 4.71
2 15 1.5 5.28 5.07 3.55
3 20 1.5 4.66 4.47 4.25
Water absorption (%) 1 10 1.13 1.41 1.47 4.08
2 15 1.5 1.53 1.60 4.38
3 20 1.5 1.62 1.69 4.14

Table 10
ANOVA for compressive strength model.

Response Factors F -Values P-Values R2 Adjusted R2 Predicted R2 S.D CoV (%)

Compressive Strength (MPa) Model 67.67 <.0001 0.956 0.943 0.897 2.38 4.54
FF 4.21 .0703
FS 26.45 .0006
FF*FS 10.72 .0096

agreement with their corresponding adjusted R2 as the difference

between the two is less than 0.2, except for water absorption
model whose difference is greater than 0.2 as shown in Table 6.
Therefore, model reduction was done where the insignificant terms
were removed using backward elimination regression. The vari-
ability of the models with reference to the experimental data
was also checked using their standard deviations (SD) and coeffi-
cient of variations (CoV). The low SD and CoV of the models show
that the experimental data fitted the developed models with
higher correlation degree. Furthermore, for all the developed mod-
els their adequate precision values are desirable with adequate sig-
nal, as they are greater than 4, therefore models can be used in
navigating the design space
The final developed models with the insignificant terms remov-
ing using backward elimination regression, with the hierarchical
terms added afterwards are shown in Eqs. 6(ad).

Fig. 21. Relationship between compressive strength and flexural/split tensile F C ¼ 95:952  3:976  A þ 14:902  B
strength.
þ 0:068  A2  6:293  B2 ð6aÞ

F S ¼ 8:222  0:305  A þ 0:976  B þ 0:010  AB

F F ¼ 10:559  0:346  A þ 0:782  B
þ 0:0048  A2  0:508  B2 ð5cÞ
þ 0:0063  A2  0:438  B2 ð6bÞ

W ¼ 1:536 þ 0:012  A  0:426  B  0:0005  AB

F S ¼ 8:222  0:305  A þ 0:976  B þ 0:010  AB
þ 0:00019  A2 þ 0:189  B2 ð5dÞ
þ 0:005  A2  0:508  B2 ð6cÞ
where FC is the compressive strength; FF is the strength; FS is the
splitting tensile strength and W is the water absorption.
W ¼ 1:483 þ 0:019  A  0:450  B þ 0:196  B2 ð6dÞ
The adequacy and degree of correlation (determination) of each
response model can also be explained by the high regression coef- The predicted versus actual plots for the models are also impor-
ficient (R2) which also explains the fitness and quality of the mod- tant factors for checking the adequacy and fitness of the models
els as shown in Table 6. The compressive strength, flexural graphically as shown in Fig. 16(a–d), and they were found to align
strength, splitting tensile strength and water absorption have R2 and closely fit the straight line meaning the experimental data
values of 96.14%, 97.04%, 98.97%, and 89.75% respectively. This were in agreement and fit to the predicted results from models.
shows that the models have a higher degree of correlation between Therefore, the developed response models are applicable and
the predicted models and experimental data, and there is a close appropriate for predicting strengths and water absorption of RCR.
fitness between the data and their regression lines, with only The 3-Dimensional response surface plots for the developed
3.14%, 2.96%, 1.03% and 10.25% of the experimental data for com- models are shown in Fig. 17(a–d). It can be seen that crumb rubber
pressive strength, flexural strength, splitting tensile strength and exhibits higher and negative effects on the responses compared to
water absorption respectively cannot be correlated by the models. nano silica. The optimum combinations of 10% CR with 1% NS gave
In addition, for all response models, the predicted R2 were in best results for all the models.
250 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251

The 2-dimensional contour plots the developed models are that the compressive strength of RCR is directly proportional to
shown in Fig. 18(a–d). It can be see that all the contour lines are its flexural and tensile strengths.
oval in shape meaning there is a perfect interaction between CR
and NS for all the responses, with water absorption response hav- F C ¼ 188:144  33:348  F F  17:088  F S þ 5:279  F F  F S ð8Þ
ing the best interaction as shown in Fig. 18d [21]. The reddish
regions on the contour plots for strengths shows the best combina-
tions to give the optimum values, while the bluish region on the 4. Conclusions
plots for water absorption shows the optimum. In this case combi-
nation of 10% CR with 1% NS gave the optimum combinations Based on the experimental works and results analysis, the fol-
The perturbation plots for the developed models are shown in lowing conclusions were drawn
Fig. 19(a–d). These plots help to compare the effect of crumb rub-
ber and nano silica on the responses. The inclined curve (steep 1) The Vebe consistency time and fresh density decreases with
slopes) shows that the compressive strength, flexural strength, increment in CR and increases with the addition of NS. NS
splitting tensile strength, and water absorption of RCR are sensitive increases water demand in RCR.
to crumb rubber and nano silica incorporations. 2) The compressive strength and splitting tensile strength of
RCR decreases with replacing fine aggregate with CR above
10%, while flexural strength increases for up to 20% CR
3.4.3. Multi-objective optimization
content.
A multi-objective optimization has been carried to using RSM to
3) The elastic modulus of RCR decreases with increment in CR
determine the best optimum hybrid of CR and NS that could yield
content and increases with the addition of NS, while water
maximum compressive, flexural strengths and splitting tensile
absorption increases with increment in CR and decreases
strength and minimum water absorption of RCR as they are the
with NS addition.
major properties for design and construction of RCR pavement.
4) The Cantabro loss (abrasion resistance) of RCR increases
The optimization goals are summarized in Table 7.
with incorporation of 10% CR as replacement to fine aggre-
The results for the multi-objective optimization are presented
gate and addition of up to 3% NS by weight of cementitious
in Table 8. Based on the optimization goals, the best optimal mix-
materials.
ture proportions selected by the RSM software is obtained by com-
5) The addition of up to 2% NS improves both compressive,
bining 10% of CR as a fine aggregate replacement with 1.13% of NS
splitting and flexural strengths of RCR at all age of curing,
as addition to cement to give the maximize responses with com-
with 1% NS is being the optimum value and is successful in
bined desirability of 94.3% as shown in Table 8 and Fig. 20.
mitigating the loss in 28 days compressive and splitting ten-
In order to validate the optimization results and the overall
sile strengths for up to 20% CR. However higher NS above 2%
response models, sequence of supplementary experiments were
shows a negative effect on strengths.
carried out in the laboratory using the optimized mixture propor-
6) Addition of NS to RCR resulted in refining the pore structure
tions and selecting some percentage variables. All other fixed con-
and densification of both ITZ and micro structure.
stituent materials were the same as for the overall mixtures as
7) The results of the optimization process using RSM shows
given in Table 3. The compressive strength, flexural strength, split-
that the combination of 10% CR as a fine aggregate replace-
ting tensile strength and water absorption were then tested exper-
ment by volume and addition of 1.13% NS by weight of
imentally and compared with the theoretical predicted). The
cement will be yielding the best compressive strength, flex-
percentage error or variations between the predicted responses
ural strength, split tensile strength and water absorption for
and the experimental results were calculated using Eq. (7) [3].
RCR.
The summary of the predicted and experimental results is pre-
8) The results of the RSM analysis shows a good correlation
sented in Table 9.
exists between the developed models and their correspond-
Experimental  Model ing experimental data as all exhibited quadratic relation-
Errorð%Þ ¼  100 ð7Þ ships, with a higher degree of correlations.
Experimental
9) The predicted values are fitted to the actual values, and
The experimental and predicted results were presented in model validation shows that there is less percentage mean
Table 9 and it can be observed that the experimental results are error (3.20%, 3.29%, 4.17% and 4.20% for compressive
in agreement with the theoretical values with average mean errors strength, flexural strength, splitting tensile strength and
of 3.20%, 3.29%, 4.17% and 4.20% for compressive strength, flexural water absorption models respectively). Therefore the devel-
strength, splitting tensile strength and water absorption oped models can be used to predict strengths and water
respectively. absorption of RCR using CR and NS as the independent
variables.
3.4.4. Relationship between compressive strength, flexural strength
and splitting tensile strength of RCR using RSM
The developed model which can be used for predicting the com-
Acknowledgement
pressive strength (FC) of RCR using its corresponding flexural
strength (FF) and splitting tensile strengths (FS) as the variables is
The authors would like to thank the Ministry of Education
given in Eq. (8). As seen, from Table 10, the model is significant
(MOE) of Malaysia for granting the project under code
having P-value less than .05, with a high F-value of 67.67. Further-
PRGS/1/13/TK03/UTP/02/02.
more, the adequacy of the model can be verified by the high degree
of determination (R2 = 95.65), and the predicted R2 was in agree-
ment with the adjusted R2 as their differences is less than 0.2. References
The low standard deviation (SD) and coefficient of variation
(CoV) shows that the developed model fitted the experimental data [1] ACI 211.3R-02, A., Guide for Selecting Proportions for No-Slump Concrete,
American Concrete Institute, Michigan, USA, ACI, 2002.
with lower variation. The relationship between the developed [2] ACI 325-10R (2001). Report on Roller-Compacted Concrete Pavements,
model and the variables is presented in Fig. 21, and can be seen American Concrete Institute.
 B.S. Mohammed, M. Adamu / Construction and Building Materials 159 (2018) 234–251 251

[3] K.E. Alyamac, E. Ghafari, R. Ince, Development of eco-efficient self-compacting [31] B.S. Mohammed, K.M.A. Hossain, J.T.E. Swee, G. Wong, M. Abdullahi, Properties
concrete with waste marble powder using the response surface method, J. of crumb rubber hollow concrete block, J. Cleaner Prod. 23 (1) (2012) 57–67.
Cleaner Prod. 144 (2017) 192–202. [32] D.C. Montgomery, Design and Analysis of Experiments, Arizona State
[4] M. Amin, K. Abu el-Hassan,, Effect of using different types of nano materials on University, John Wiley & Sons, 2008.
mechanical properties of high strength concrete, Constr. Build. Mater. 80 [33] N. Mtarfi, Z. Rais, M. Taleb, K. Kada, Effect of fly ash and grading agent on the
(2015) 116–124. properties of mortar using response surface methodology, J. Build. Eng. (2016).
[5] ASTM C1170, Standard Test Method for Determining Consistency and Density [34] B.B. Mukharjee, S.V. Barai, Influence of nano-silica on the properties of
of Roller-Compacted Concrete Using a Vibrating Table 2014 West recycled aggregate concrete, Constr. Build. Mater. 55 (2014) 29–37.
Conshohocken Pennsylvania, United States, ASTM International. [35] K.B. Najim, M.R. Hall, Crumb rubber aggregate coatings/pre-treatments and
[6] M. Bederina, Z. Makhloufi, A. Bounoua, T. Bouziani, M. Queneudec, Effect of their effects on interfacial bonding, air entrapment and fracture toughness in
partial and total replacement of siliceous river sand with limestone crushed self-compacting rubberised concrete (SCRC), Mater. Struct. 46 (12) (2013)
sand on the durability of mortars exposed to chemical solutions, Constr. Build. 2029–2043.
Mater. 47 (2013) 146–158. [36] A. Nazari, S. Riahi, Abrasion resistance of concrete containing SiO2 and Al2O3
[7] C. Chhorn, S.J. Hong, S.-W. Lee, A study on performance of roller-compacted nanoparticles in different curing media, Energy Build. 43 (10) (2011) 2939–
concrete for pavement, Constr. Build. Mater. 153 (2017) 535–543. 2946.
[8] J. Chu, Solutions to sustainability in construction: some examples, Procedia [37] M.M. Norhasri, M. Hamidah, A.M. Fadzil, Applications of using nano material in
Eng. 145 (2016) 1127–1134. concrete: A review, Constr. Build. Mater. 133 (2017) 91–97.
[9] K. Connor, S. Cortesa, S. Issagaliyeva, A. Meunier, O. Bijaisoradat, N. [38] O. Onuaguluchi, Effects of surface pre-coating and silica fume on crumb
Kongkatigumjorn, K. Wattanavit, Developing a sustainable waste tire rubber-cement matrix interface and cement mortar properties, J. Cleaner Prod.
management strategy for Thailand, Worcester Polytechnic Institute, 104 (2015) 339–345.
Worcester, Massachusetts, 2013. [39] O. Onuaguluchi, D.K. Panesar, Hardened properties of concrete mixtures
[10] CRD-C 162,, Standard Practice for Selecting Proportions for Roller Compacted containing pre-coated crumb rubber and silica fume, J. Cleaner Prod. 82 (2014)
Concrete (RCC) Pavement Mixtures Using Soil Compaction Concepts, 125–131.
Department of the Army, Corps of Engineers, Washington, DC, USA, 1992. [40] G. Ossola, A. Wojcik, UV modification of tire rubber for use in cementitious
[11] Q. Dong, H. Wu, B. Huang, X. Shu, K. Wang, Development of a simple and fast composites, Cement Concrete Comp. 52 (2014) 34–41.
test method for measuring the durability of Portland cementpervious concrete, [41] E. Ozbay, M. Lachemi, U.K. Sevim, Compressive strength, abrasion resistance
Portland Cement Association, 2010. and energy absorption capacity of rubberized concretes with and without slag,
[12] C.M. Douglas (2008). Design and analysis of experiments. Mater. Struct. 44 (7) (2011) 1297–1307.
[13] M. Fakhri, The effect of waste rubber particles and silica fume on the [42] F. Pelisser, A. Barcelos, D. Santos, M. Peterson, A.M. Bernardin, Lightweight
mechanical properties of Roller Compacted Concrete Pavement, J. Cleaner concrete production with low Portland cement consumption, J. Cleaner Prod.
Prod. 129 (2016) 521–530. 23 (1) (2012) 68–74.
[14] S. Fallah, M. Nematzadeh, Mechanical properties and durability of high- [43] A.A. Ramezanianpour, A. Mohammadi, E.R. Dehkordi, Q.B. Chenar, Mechanical
strength concrete containing macro-polymeric and polypropylene fibers with properties and durability of roller compacted concrete pavements in cold
nano-silica and silica fume, Constr. Build. Mater. 132 (2017) 170–187. regions, Constr. Build. Mater. 146 (2017) 260–266.
[15] E. Ghafari, H. Costa, E. Júlio, A. Portugal, L. Durães, The effect of nanosilica [44] S.K. Rao, P. Sravana, T.C. Rao, Abrasion resistance and mechanical properties of
addition on flowability, strength and transport properties of ultra high Roller Compacted Concrete with GGBS, Constr. Build. Mater. 114 (2016) 925–
performance concrete, Mater. Des. 59 (2014) 1–9. 933.
[16] S. Ghahari, A. Mohammadi, A. Ramezanianpour (2017). Performance [45] A.M. Rashad, A comprehensive overview about recycling rubber as fine
assessment of natural pozzolan roller compacted concrete pavements. Case aggregate replacement in traditional cementitious materials, Int. J. Sustainable
Studies in Construction Materials. Built Environ. 5 (1) (2016) 46–82.
[17] E. Güneyisi, M. Gesoğlu, Z. Algın, K. Mermerdasß, Optimization of concrete [46] O. Rezaifar, M. Hasanzadeh, M. Gholhaki, Concrete made with hybrid blends of
mixture with hybrid blends of metakaolin and fly ash using response surface crumb rubber and metakaolin: optimization using Response Surface Method,
method, Comp. Part B Eng. 60 (2014) 707–715. Constr. Build. Mater. 123 (2016) 59–68.
[18] T. Gupta, S. Chaudhary, R.K. Sharma, Mechanical and durability properties of [47] A. Richardson, K. Coventry, V. Edmondson, E. Dias, Crumb rubber used in
waste rubber fiber concrete with and without silica fume, J. Cleaner Prod. 112 concrete to provide freeze–thaw protection (optimal particle size), J. Cleaner
(2016) 702–711. Prod. 112 (2016) 599–606.
[19] S. Hesami, I.S. Hikouei, S.A.A. Emadi, Mechanical behavior of self-compacting [48] RMA, Scrap tire markets in the United States, USA, Rubber Manufacturers
concrete pavements incorporating recycled tire rubber crumb and reinforced Association, Washington DC, 2016.
with polypropylene fiber, J. Cleaner Prod. 133 (2016) 228–234. [49] D.M. Sadek, M.M. El-Attar, Structural behavior of rubberized masonry walls, J.
[20] G. Li, Z. Wang, C.K. Leung, S. Tang, J. Pan, W. Huang, E. Chen, Properties of Cleaner Prod. 89 (2015) 174–186.
rubberized concrete modified by using silane coupling agent and carboxylated [50] A.M. Said, M.S. Zeidan, M. Bassuoni, Y. Tian, Properties of concrete
SBR, J. Cleaner Prod. 112 (2016) 797–807. incorporating nano-silica, Constr. Build. Mater. 36 (2012) 838–844.
[21] Q. Li, L. Cai, Y. Fu, H. Wang, Y. Zou, Fracture properties and response surface [51] R. Siddique, Waste Materials and By-Products in Concrete, Springer Science &
methodology model of alkali-slag concrete under freeze–thaw cycles, Constr. Business Media, 2007.
Build. Mater. 93 (2015) 620–626. [52] B.S. Thomas, R.C. Gupta, Properties of high strength concrete containing scrap
[22] R. Liu, L. Zhang, Utilization of waste tire rubber powder in concrete, Composite tire rubber, J. Cleaner Prod. (2015).
Interfaces 22 (9) (2015) 823–835. [53] B.S. Thomas, R.C. Gupta, V.J. Panicker, Recycling of waste tire rubber as
[23] A. Meddah, M. Beddar, A. Bali, Use of shredded rubber tire aggregates for roller aggregate in concrete: durability-related performance, J. Cleaner Prod. (2015).
compacted concrete pavement, J. Cleaner Prod. 72 (2014) 187–192. [54] B.S. Thomas, R.C. Gupta, V.J. Panicker, Recycling of waste tire rubber as
[24] T.B. Moghaddam, M.R. Karim, M. Abdelaziz, A review on fatigue and rutting aggregate in concrete: durability-related performance, J. Cleaner Prod. 112
performance of asphalt mixes, Sci. Res. Essays 6 (4) (2011) 670–682. (2016) 504–513.
[25] A.M. Mohamed, Influence of nano materials on flexural behavior and [55] B.S. Thomas, S. Kumar, P. Mehra, R.C. Gupta, M. Joseph, L.J. Csetenyi, Abrasion
compressive strength of concrete, HBRC J. (2015). resistance of sustainable green concrete containing waste tire rubber particles,
[26] B. Mohammed, N.J. Azmi, Failure mode and modulus elasticity of concrete Constr. Build. Mater. 124 (2016) 906–909.
containing recycled tire rubber, J. Solid Waste Technol. Manage. 37 (1) (2011) [56] B. Yesilata, H. Bulut, P. Turgut, Experimental study on thermal behavior of a
16–24. building structure using rubberized exterior-walls, Energy Build. 43 (2) (2011)
[27] B.S. Mohammed, B.E. Achara, M.F. Nuruddin, Y. Melvin, M.Z. Zulkefli, 393–399.
Properties of nano-silica-modified self-compacting engineered cementitious [57] A. Yilmaz, N. Degirmenci, Possibility of using waste tire rubber and fly ash with
composites, J. Cleaner Prod. (2017). Portland cement as construction materials, Waste Manage. 29 (5) (2009)
[28] B.S. Mohammed, A.B. Awang, S. San Wong, C.P. Nhavene, Properties of nano 1541–1546.
silica modified rubbercrete, J. Cleaner Prod. 119 (2016) 66–75. [58] O. Youssf, M.A. ElGawady, J.E. Mills, X. Ma, An experimental investigation of
[29] B.S. Mohammed, N. Azmi, Strength reduction factors for structural crumb rubber concrete confined by fibre reinforced polymer tubes, Constr.
rubbercrete, Front. Struct. Civil Eng. 8 (3) (2014) 270–281. Build. Mater. 53 (2014) 522–532.
[30] B.S. Mohammed, O.C. Fang, K.M.A. Hossain, M. Lachemi, Mix proportioning of [59] O. Youssf, J.E. Mills, R. Hassanli, Assessment of the mechanical performance of
concrete containing paper mill residuals using response surface methodology, crumb rubber concrete, Constr. Build. Mater. 125 (2016) 175–183.
Constr. Build. Mater. 35 (2012) 63–68.