Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Influence of Filler Addition on Mechanical Behavior

of Cementitious Mortar-Rubber Aggregates:


Experimental Study and Modeling
Mohamed Turki1; Inès Zarrad2; Estelle Bretagne3; and Michèle Quéneudec4

Abstract: This work focuses on the feasibility of reusing and valorizing rubber aggregates from shredded worn tires in cement paste.
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

This kind of eco-material could be used, for example, in special buildings, roads or pavement, or recreational areas. Fractions of the volume
of sand were substituted by the same volume of rubber aggregates. The volume contains the ratio of rubber aggregates ranging from 0 to
50%. Nevertheless, compressive and flexural strengths decrease when the volume ratio of rubber aggregates increases in mortar-rubber
aggregate mixtures. A solution to the decrease in mechanical strengths consists of adding mineral particles, such as siliceous or limestone
fillers. Experimental characterization of the mechanical behavior of mortar-rubber aggregates containing either siliceous or limestone fillers is
developed in this paper. It is observed that the use of either siliceous filler or limestone filler enhances mechanical properties. Modeling
is developed to predict compressive and flexural strengths of mortar-rubber aggregate mixtures using the fuzzy logic method. Experi-
mental mechanical results are compared with modeling. DOI: 10.1061/(ASCE)MT.1943-5533.0000512. © 2012 American Society of Civil
Engineers.
CE Database subject headings: Rubber; Aggregates; Mortars; Fills; Mechanical properties; Fuzzy sets; Experimentation.
Author keywords: Rubber aggregates; Mortar; Fillers; Mechanical properties; Fuzzy logic methods.

Introduction temperature (−20, þ140°C) and have also shown that the presence
of rubber particles reduces the stresses in the matrix, thus alleviat-
Many works in the world have already been carried out on the val- ing the cracks generated by thermal expansion. An exposure to
orization of rubber aggregates of tires in construction materials. temperatures above 600°C showed that the risk of bursting of
Some studies are about the influence of the rubber aggregates at the composite decreases with the presence of rubber aggregates,
the fresh state, the rheology of ordinary concrete (Khaloo et al. which suggests better behavior under the effect of high tempera-
2008), and self-placing concrete (Garros 2007; Khatib and Bayony
tures. A study of delayed deformations under constant load has
1999). Fedroff et al. (1996) showed that the quantity of air content
identified the main trends of these delayed deformations related
increased with the volume of rubber embedded in concrete. At the
hardened state of concrete, other studies showed a lower compres- to percentage, size, and type of rubber aggregates. The analysis
sive strength (Khatib and Bayony 1999; Eldin and Senouci 1993) of the experimental results allows the writing of a simple model
when replacing gravel and sand with rubber aggregates. Kaloush giving the ultimate delayed deformations under constant load for
et al. (2005) revealed that the rise of rubber aggregate content different parameters and minimizing the duration of the creep tests.
in concrete leads to the splitting tensile decreasing. The surface In other works, Ganjian et al. (2009) studied the scrap-tire-
treatment of rubber aggregates has also been studied to improve rubber replacement both for aggregate and filler in concrete mainly
the adherence between rubber and cement paste (Segre and for durability; the mechanical tests were performed, and the results
Joekes 2000; Xi et al. 2004; Albano et al. 2005). The work of were analyzed. The mechanical tests included compressive, tensile,
Sukontasukkul and Chaikaew (2006) shows the deformation and flexural strengths and modulus of elasticity. For instance, the
capacity and the shrinkage cracking. The work of Topçu (1997) reduction in compressive strength at 28 days of age was approx-
concerns the fragility of the specimen. Moreover, some of these imately 10–23% for crumb rubber aggregates and approximately
authors have focused their studies on mechanical behavior and 20–40% for filler rubber replacement. Yilmaz and Degirmenci
(2009) used waste tire rubber and fly ash with Portland cement
1
Ecole Nationale d’Ingénieurs de Sfax, BP:1173 Sfax, 3038 Tunisia as construction materials. The reductions in compressive strength
(corresponding author). E-mail: mohamed.turki@gmail.com (28 days) were 15, 9, and 0.2% for 0.0–0.25, 0.25–0.50, and
2
Ecole Nationale d’Ingénieurs de Sfax, BP:1173 Sfax, 3038 Tunisia. 0.50–1.0 mm fractions, respectively, with an increase in rubber
3
Laboratoire des Technologies Innovantes (EA 3899), Université de content from 20 to 30%. This trend is slightly influenced by the
Picardie Jules Verne, IUT de l’Aisne, 15 Avenue François Mitterrand, fraction size. Comparisons among three rubber fraction sizes indi-
02880 Cuffies, France. cate that the rubber particles with 0.50–1.0-mm diameter seemed to
4
Université de Picardie Jules Verne, unité de recherche EPROAD, have a better compressive strength than the others. However, the
5–7 rue du Moulin Neuf, 80 000 Amiens, France.
investigation of the potential use of recycled tire rubber and silica
Note. This manuscript was submitted on March 9, 2011; approved on
March 7, 2012; published online on March 10, 2012. Discussion period fume in cement matrix was studied by Pelisser et al. (2011). Con-
open until April 1, 2013; separate discussions must be submitted for indi- crete formulations were produced with the replacement of 10%
vidual papers. This paper is part of the Journal of Materials in Civil En- sand aggregate by recycled tire rubber using conventional rubber
gineering, Vol. 24, No. 11, November 1, 2012. © ASCE, ISSN 0899-1561/ and rubber modified with alkaline activation and silica fume addi-
2012/11-1350-1358/$25.00. tion to improve the mechanical properties. There was a synergy

1350 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012

J. Mater. Civ. Eng. 2012.24:1350-1358.


between the combinations of the chemical treatment by sodium
hydroxide followed by the addition of 15% silica fume. This could
result in a resistance recovery specifically for a compressive resis-
tance of 50 MPa (the same resistance as the reference concrete).
The objective of this study is to investigate the effect of fillers
(siliceous and limestone) in mortar-rubber aggregates mixtures.
Also this paper describes experimental tests that provide mechani-
cal properties such as flexural and compressive strengths. A corre-
lation between mechanical strengths and matrix porosity allows
studying the links between them.
In modeling studies, Topçu and Sandemir (2008) predicted
mechanical behavior of mortar containing rubber particles with
fuzzy logic model. In this paper, a similar fuzzy logic method
was used to predict mechanical properties for mortar-rubber aggre-
Fig. 1. Grain-size distribution of sand and rubber aggregates
gates materials containing fillers on the basis of some experimental
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

results. In testing this model, mass compositions of samples


(cement, sand, filler, rubber, and water) were introduced as inputs.
Flexural and compressive strengths were used as outputs. Sub-
sequently, tests were performed to obtain a similar value to the
experimental results.

Experimental Program

Materials
Rubber aggregates used in this study were obtained by shredding
worn tires. Rubber particles were introduced in the mortar mixture
by partial volume substitution of sand (0 to 50%). For example, the
quantity of sand substituted with rubber aggregates is approxi-
mately 61.13 kg=m3 for 10% of substitution (Table 1). Fillers
were also introduced in the mortar mixture by the mass substitution Fig. 2. Grain-size distribution of siliceous and limestone fillers
of sand. The volume substitution represents 5% of the reference
material sand mass composition; siliceous fillers are approxi
mately 67.5 kg=m3, and limestone fillers are approximately
34.39 kg=m3 . The rubber-filler-free material was taken as reference substitution of sand. The volume substitution represents 5% of
material. reference material sand mass composition (Table 1).
The apparent dry bulk density of rubber was 1200 kg=m3 , and
the apparent dry bulk density both of sand and siliceous filler used Tests
in the composites was 2650 kg=m3 . The apparent dry bulk density
of limestone filler was 1350 kg=m3 . The gap-grading analysis of The tests were carried out on three prismatic 40-by-40-by-160-mm
rubber aggregates, fillers, and sand are displayed in Figs. 1 and 2; test samples that had been cured for 28 days at a constant temper-
the sand had 0–2-mm-size grading. The rubber aggregates had ature (20°C) and a constant relative humidity (100%). Distributions
1–4-mm-size grading. The hydraulic binder was an artificial of particle size of sand and rubber had been determined by sieve-
Portland cement CPA CEM I 42.5 (AFNOR 1995, EN 196-1) with grading analysis. Distributions of siliceous and limestone fillers
a bulk density of 3100 kg=m3 . Some authors, such as Turatsinze had been determined by X-ray SediGraph analysis. Dry bulk den-
et al. (2005), used additives to control workability, but they can sities for different mortar–rubber aggregates mixtures were deter-
affect mechanical strength. In the present work, there is no addition mined by means of geometrical measurement and weighting.
of plasticiser. Starting water-to-cement ratios (W/C) was equal to Compressive and flexural strengths were evaluated on several pris-
0.5, with varied proportions of rubber aggregates ranging to 0 from matic 40-by-40-by-160-mm test samples in compliance with the
50%. Fillers were introduced in the mortar mixture by the mass operating methods specified in standard EN 196-1 (AFNOR 1995).
Each value of compressive (Rc ) and flexural (Rf ) strengths was
the average of three tests. The microstructure porosity of various
samples, determined by image analysis, was studied by means
Table 1. Mass Composition of Prismatic Test Samples
of scanning electron microscopy (SEM).
Volume ratio
of rubber Rubber
aggregates Cement Sand Water aggregates
Results and Discussion
(percentage) (kg=m3 ) (kg=m3 ) (kg=m3 ) (kg=m3 )
0 1,350 0
10 1,215 61.13 Flexural and Compressive Strengths
20 1,080 122.26 The addition of rubber aggregates in mortar decreases considerably
30 945 183.4 the mechanical behavior of the specimens (Table 2), which is why
40 810 244.53 adding fine mineral fillers was proposed to restore the loss of com-
50 450 675 225 305.66
pressive and flexural strengths. Improvements cannot be made to

JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012 / 1351

J. Mater. Civ. Eng. 2012.24:1350-1358.


Table 2. Evolution of Flexural and Compressive Strengths Volume Ratio Percentage of Rubber Aggregates
Strength (MPa) 0 10 20 30 40 50
Flexural strength Rf0 (mixture without filler) 3.940 1.807 1.450 1.007 1.020 1.090
Flexural strength Rf1 (mixture with calcitic filler) 3.898 3.047 2.209 1.703 1.331 0.961
Flexural strength Rf2 (mixture with siliceous filler) 4.550 3.220 2.260 1.310 1.135 0.960
Compressive strength Rc0 (mixture without filler) 44.690 21.870 16.560 8.430 7.250 6.870
Compressive strength Rc1 (mixture with calcitic filler) 49.175 32.578 24.467 14.159 10.960 7.332
Compressive strength Rc2 (mixture with siliceous filler) 47.830 37.490 26.500 15.510 11.315 7.120

match the control mortar because the effect of rubber aggregates in respectively, in comparison with filler-free mortar–rubber aggre-
matrix is more significant than fillers, i.e., the low density of rubber gates mixtures with the same volume ratio of rubber aggregates.
in comparison with the sand substituted is the major reason to lose The increase of flexural strength is higher with the siliceous filler
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

on mechanical properties. addition in the case of 0, 10, and 20% rubber aggregates volume
The evolution of flexural and compressive strengths of mortar- ratio. Otherwise, the increase of flexural strength is better with the
rubber aggregates mixtures containing siliceous or limestone filler limestone filler addition in the case of 30 and 40% rubber aggre-
is given in Table 2. Each value of compressive and flexural gates volume ratio. The increase of compressive strength is higher
strengths (Rc and Rf ) was the average of three tests. The standard with the siliceous filler addition in the case of 10, 20, 30, and 40%
deviation values of Rc and Rf without fillers ranged between 0.02 of rubber aggregates volume ratio. Otherwise, the increase of flexu-
and 0.47 and 0.11 and 0.58, respectively. With siliceous and lime- ral strength is slightly higher with the limestone filler addition in
stone fillers, the standard deviation values were between 0.15 and the case of 0 and 50% rubber aggregates volume ratio.
0.80 for Rc and 0.03 and 0.19 for Rf. The reduction factors for flexural and compressive strengths
The addition of 1% both for limestone and siliceous fillers are lower for filler-free mixtures than for mixtures with siliceous
in mortar mixture is used according to previous studies (Turki or limestone filler (Figs. 6 and 7), except in the cases of 50% rub-
et al. 2010), which explained the negative effect of the fine aggre- ber aggregates volume ratio and 40% for reduction factors for
gates addition in cementitious materials more to 5%, on mechanical flexural strength. Filler presence limits the relative effect of incor-
properties. So the main objective of this paper is to substitute differ- porating rubber aggregates for mechanical strengths. Results are
ent percentages of rubber (0 to 50% with sand) and not the filler. better with limestone filler in the case of flexural strengths and with
Flexural and compressive strengths decrease significantly when siliceous filler in the case of compressive effect.
the percentage of rubber increases in the mixture. Flexural and
compressive strengths are lower for filler-free mixtures than for
mixtures with siliceous or limestone filler (Figs. 3 and 4), except Matrix Porosity
for flexural strength in the case of the 50% rubber aggregates Matrix porosity of filler-free materials was first estimated by an
volume ratio. image analysis and the use of a computer program by Turki et al.
Fig. 5 shows the relationship between the reduction factor for (2009). The average value of total porosity was obtained by soft-
flexural strengths (RFFS) and the reduction factor for compressive ware analysis of scanning electron microphotographs taken in a
strengths (RFCS) depending on volume ratio of rubber aggregates. section plane. This porosity can be correlated with global porosity
Fig. 5 reveals the variation of flexural and compressive strengths, values calculated by the pycnometer method (Table 3 and Fig. 8).

Fig. 3. Relationship between flexural strength and volume ratio of rubber aggregates

1352 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012

J. Mater. Civ. Eng. 2012.24:1350-1358.


Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Relationship between compressive strength and volume ratio of rubber aggregates

Fig. 5. Variation in comparison with filler-free mortar-rubber: (a) flexural; (b) compressive strengths aggregates mixtures with the same volume ratio
of rubber aggregates

Fig. 6. Relationship between reduction factor for flexural strength and Fig. 7. Relationship between reduction factor for compressive strength
volume ratio of rubber aggregates and volume ratio of rubber aggregates

JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012 / 1353

J. Mater. Civ. Eng. 2012.24:1350-1358.


Table 3. Matrix Porosity of Mortar–Rubber Aggregates Containing Siliceous and Limestone Fillers
Volume ratio of rubber aggregates (%) 0 10 30 50
Porosity (%) without filler obtained by image analysis 14.35 16.2 17.84 26.86
Porosity (%) without filler obtained by toluene pycnometer method 5.24 7.86 16.06 24.58
Porosity (%) with siliceous filler obtained by toluene pycnometer method 10.49 13.11 15.35 18.48
Porosity (%) with limestone filler obtained by toluene pycnometer method 5.86 10.15 15.22 20.11
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Matrix porosity of mortar-rubber aggregates mixtures (toluene pycnometer method)

The standard deviation values ranged between 0.08 and 1.15 for filler-free mortar-rubber aggregates mixtures. Indeed, more detailed
the different methods used in this case. An observed difference considerations of mechanical properties require the analysis of
between porosities was obtained by an image analysis and by the the interface between aggregate and matrix (Ramesh et al. 1996).
pycnometer method. Nevertheless, numerical results obtained by The conclusion was that at less two types of porosity exist in
image analysis are considered to be qualitative information cementitious matrix and around rubber aggregates. In the case
(increase or decrease of matrix porosity). of filler mortar–rubber aggregates composites, microstructural ob-
In this case, the comparison between image analysis and pyc- servations of interfacial transition zone (ITZ) revealed the compac-
nometer methods is qualitative and not quantitative. But the trend tion effect of fine particles in the matrix that improves adherence
of the rise in porosity is observed when the percentage of rubber with rubber aggregates. So even if matrix global porosity increases
aggregates increases (for 0 to 50%). According to the Table 3, it
was concluded that the difference in results between the two
methods is significant both for mortar–rubber aggregates mixture
containing 10% and for the reference materials (0%). Specimens
containing 0, 10, 30, and 50% of rubber aggregates were used be-
cause only the representative images with high resolution taken by
SEM, respectively, were available to make an image analysis and
obtain suitable results of matrix porosity.
Matrix porosity of mortar-rubber aggregates mixtures increased
in comparison with the reference material when the percentage of
rubber rose in the mixture. The effect of siliceous and limestone
filler was decreasing porosity, in comparison with specimens con-
taining rubber aggregates, only for the specimens of 30 and 50%,
which is in the range of 15.35–18.48% and 15.22–20.11%, respec-
tively. The decrease of matrix porosity improves the adherence of
cement paste and the mechanical behavior of specimens. Whereas,
at 0 and 10% of substitution, an inverse proportionality was ob-
served between the rise of porosity and the improvement of
mechanical properties, in this case, a correlation between porosity
and mechanical strengths cannot be made.
A previous study (Turki et al. 2007) focused on the strong
Fig. 9. Structure of the fuzzy controller
relation between microstructure and physical properties in

1354 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012

J. Mater. Civ. Eng. 2012.24:1350-1358.


Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Membership functions of input variables for Rf

Fig. 11. Membership functions of input variables for Rc

JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012 / 1355

J. Mater. Civ. Eng. 2012.24:1350-1358.


in case of 0 and 10%, rubber aggregates-matrix ITZ is improved Table 4. Comparison of Experimental Results with Training Results
by filler presence. This effect on the rubber-matrix ITZ permits Obtained from Fuzzy Logic Model
mechanical properties of mortar-rubber aggregates composites to Training results obtained
improve significantly. Experimental result from fuzzy logic model
Number Rc Rf Rc Rf
Fuzzy Logic Inference System: Flexural and 1 24.37 4 23.4476 4.05
Compressive Strengths 2 17.45 3.7 16.5023 3.6
3 13.71 3 16.1492 3.3
Zadeh has been studying the fuzzy logic concept since 1967. 4 32.4 6.5 22.4719 4.65
He suggested that the application of such fuzzy thinking would 5 22.02 5.1 22.0172 5.1
yield great benefits across a wide range of engineering and con- 6 15.42 4.5 15.4209 4.5
trol problems. Fuzzy inference systems are powerful tools for 7 14.06 4.2 10.5978 4.2
simulation of nonlinear behaviors (Inan et al. 2007). A fuzzy in- 8 11.81 3.9 14.6021 3.8987
ference system is presented in Fig. 9. It is usually constitutes three 9 8.17 3.1 17.7678 3.101
10 31.5 4.5 20.0787 4.4996
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

components: fuzzification interface, inference mechanism, and


defuzzifaction interface. Five properties were introduced as input 11 19.63 4.3 22.0859 4.3
12 12.87 3 13.0383 3.002
layers: mass composition of cement, sand, (limestone or siliceous)
13 12.4 3.35 12.077 3.349
fillers, rubber aggregates, and water in mortar-rubber aggregates 14 10.85 3 11.1017 3.0001
mixtures. For this first step, limestone or siliceous nature of fillers 15 9.78 1.9 9.7315 1.9
was not taken into account. So the term filler indicates, in a generic 16 20.68 3.6 16.4821 3.5977
way, the siliceous or limestone fillers for this fourth part. Flexural 17 17.1 3.3 22.0735 3.3016
strength Rf and compressive strength Rc were outputs of this 18 25.89 5.3 18.0397 5.2993
model. 19 10.7 3.7 13.3798 3.7
Fuzzification is the process that converts crisp, nonfuzzy 20 7.82 3.1 10.2609 3.1026
input values into their fuzzy representation. Inference involves 21 24.5 3.9 22.541 3.8978
taking the fuzzified inputs and mapping them to the rule base 22 10.72 2.7 16.6326 2.7006
23 9.11 2.3 7.1095 2.3
and producing a fuzzified output for each rule. Defuzzifaction inter-
24 22.53 4.1 23.4476 4.05
face is the opposite of fuzzification. Defuzzification converts the 25 15.55 3.5 16.5023 3.6
resulting fuzzy output(s) from the fuzzy inference mechanism to 26 18.59 3.6 16.1492 3.3
a number (Topçu and Sandemir 2008; Zadeh 1999). There are 27 12.54 2.8 22.4719 4.65
two kinds of fuzzy base rules: Mamdani and Sugeno types. In this 28 32.577 3.047 32.5605 3.0467
study, the fuzzy logic-based algorithm model was obtained by us- 29 15.51 1.702 15.5167 1.7021
ing the fuzzy logic toolbox in Matlab, and Sugeno fuzzy base rules 30 7.331 0.96 7.3297 0.96
were used. In this study, 52 experimental results were used: 30 of
them were employed for training to assess fuzzy base rules,
whereas 22 were used for testing to validate model results. In this
paper, 33 triangular-type membership functions were used to run a
Sugeno-type fuzzy inference system. All of the proposed member-
ship functions in this study consisted of five inputs and two outputs. Table 5. Comparison of Experimental Results with Testing Results
The membership function plots of input variables used in the Obtained from Fuzzy Logic Model
training of Rf and Rc are shown in Figs. 10 and 11. Testing results obtained
Comparisons between experimental results and training and Experimental results from fuzzy logic model
testing results, respectively, are shown in Tables 4 and 5. Number Rc (MPa) Rf (MPa) Rc (MPa) Rf (MPa)
In addition, root-mean-squared error (RMSE) and mean abso-
lute percentage error (MAPE) are calculated (Topçu and Sandemir 1 23.23 4.4 22.0172 5.1
2 17.5 3.8 15.4209 4.5
2008). MAPE is the mean of all the ðAPEÞi defined by
3 8.94 3 14.6021 3.8987
APEi ¼ jti − oi j=ti (1) 4 9.58 2.7 17.7678 3.101
5 8.83 2.5 20.0787 4.4996
where t = target value; and o = output value. The root-mean- 6 20.94 3.8 22.0859 4.3
7 15.82 3.1 13.0383 3.002
squared error (RMSE) is the square root of the mean square error
8 10.05 2 12.077 3.349
where the square error ðSEÞi is defined by 9 7.5 1.9 11.1017 3.0001
10 9.07 2.15 9.7315 1.9
SEi ¼ ðti − oi Þ2 (2) 11 21.13 4 16.4821 3.5977
12 15.39 3 22.0735 3.3016
RMSE of testing and training sets are smaller than 0.5 and 13 19.14 3.9 18.0397 5.2993
5.5 MPa, respectively, for flexural and compressive strengths 14 11.16 3.4 13.3798 3.7
(Table 6). The fuzzy logic model is more precise for flexural 15 6.92 2.4 10.2609 3.1026
strength forecasts. Estimates of compressive strengths are not 16 19.43 3.4 22.541 3.8978
realistic with MAPE in the range of 20–33%. 17 11 2.4 16.6326 2.7006
Figs. 12 and 13 show a comparison between experimental re- 18 8.78 1.6 7.1095 2.3
sults and predicted results, in the case of training and testing results, 19 14.159 1.31 15.8821 2.9567
respectively, for flexural and compressive strengths. 20 7.12 0.96 4.5721 2.4237
Fig. 14 shows the effect of two factors at a time (i.e., with three 21 49.147 3.897 34.5517 3.3658
22 37.49 3.22 31.3602 4.3209
other input factors that are constant) on the surface plot of flexural

1356 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012

J. Mater. Civ. Eng. 2012.24:1350-1358.


Table 6. Statistical Parameters of Testing and Training Sets
Statistical parameters RMSE of training RMSE of testing MAPE of training MAPE of testing
of fuzzy logic model set (MPa) set (MPa) set (percentage) set (percentage)
Flexural strength Rf 0.48477 0.40029 4.05 1.95
Compressive strength Rc 4.55015 5.41807 20.09 32.34
Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 12. Comparison between predicted and experimental flexural strengths results (Rf ) with fuzzy logic method (training and testing results)

Fig. 13. Comparison between predicted and experimental compressive strengths results (Rc ) with fuzzy logic method (training and testing results)

strengths. This kind of result could help optimize the mass com- focused on the strong relation between porosity and mechanical
position of mortar–rubber aggregates mixtures to obtain the desired properties, such as flexural and compressive strengths. A fuzzy
mechanical properties. logic model is elaborated to predict the mechanical behavior of
the mortar–rubber aggregate fillers composites with different
percentages of rubber particles (0, 10, 30, and 50%) and with
Conclusions and Perspective 5% siliceous or limestone fillers. The comparison between exper-
imental data and simulated results shows good accordance.
The main objective of this paper is to reveal the addition effect both The next step of this work is to validate this model and compare
of the rubber aggregates and fillers in mortar and where this kind of the fuzzy logic method with other software, such as Abacus and
eco-material could be used in civil engineering. Also, attention is Sophistic.

JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012 / 1357

J. Mater. Civ. Eng. 2012.24:1350-1358.


Downloaded from ascelibrary.org by Mohamed Turki on 01/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Combined effects of two factors on flexural strength surface plot; unit of flexural strengths, MPa

Acknowledgments activation and silica fume addition.” J. Cleaner Prod., 19(6–7),


757–763.
The authors would like to thank the Research Group on Intelligent Ramesh, G., Sotelino, E. D., and Chen, W. F. (1996). “Effect of transition
Machines (REGIM) at the National School of Engineers of Sfax, zone on elastic moduli of concrete materials.” Cem. Concr. Res., 26(4),
Tunisia, for its technical support during the simulation work using 611–622.
fuzzy logic method. Segre, N., and Joekes, I. (2000). “Use of tire rubber particles as addition to
cement paste.” Cem. Concr. Res., 30(9), 1421–1425.
Sukontasukkul, P., and Chaikaew, C. (2006). “Properties of concrete pedes-
References trian block mixed with crumb rubber.” Constr. Build. Mater., 20(7),
450–457.
AFNOR. (1995). “Méthodes d’essai des ciments. Parti I: Détermination des Topçu, I. B. (1997). “Assessment of the brittleness index of rubberized
résistances mécaniques.” EN 196-1, France. concretes.” Cem. Concr. Res., 27(2), 177–183.
Albano, C., Camacho, N., Royes, J., Felin, J. L., and Hernandez, M. (2005). Topçu, I. B., and Sandemir, M. (2008). “Prediction of rubberized mortar
“Influence of scrap rubber addition to Portland I Concrete composites: De- properties using artificial neural network and fuzzy logic.” J. Mater.
structive and non-destructive testing.” Compos. Struct., 71(3–4), 439–446. Process. Technol., 199(1–3), 108–118.
Eldin, N. N., and Senouci, A. B. (1993). “Observations on rubberized Turatsinze, A., Bonnet, S., and Granju, J. L. (2005). “Mechanical character-
concrete behavior.” Cem. Concr. Aggregates, 15(1), 74–84. isation of cement-based mortar incorporating rubber aggregates from
Fedroff, D., Ahmad, S., and Savas, B. Z. (1996). “Mechanical properties of con- recycled worn tyres.” Build. Environ., 40(2), 221–226.
crete with ground waste tire rubber.” Transp. Res. Rec., 1532, 66–72. Turki, M., Bretagne, E., Rouis, M. J., and Quéneudec, M. (2009). “Micro-
Ganjian, E., Khorami, M., and Maghsoudi, A. A. (2009). “Scrap-tyre- structure, physical and mechanical properties of mortar–rubber aggre-
rubber replacement for aggregate and filler in concrete.” Const. Build. gates mixtures.” Constr. Build. Mater., 23(7), 2715–2722.
Mater., 23(5), 1828–1836. Turki, M., Molines, E., Dheilly, R. M., Rouis, M. J., and Quéneudec, M.
Garros, M. (2007). “Composites cimentaires incorporant des granulats (2007). “Valorization of rubber aggregate in cementitious composites:
de caoutchouc issus du broyage de pneus usagés: Optimization de la Influence of the interface between the aggregates and the cement paste.”
composition et de la caractérisation.” Ph.D. thesis, Paul Sabatier Univ– Proc., 8th Int. Conf. of Eco-Materials—ICEM8, Brunel Univ., UK.
Toulouse III, Toulouse, France. Turki, M., Zarrad, I., Bretagne, E., Rouis, M. J., and Quéneudec, M. (2010).
Inan, G., Göktepe, A. B., Ramyar, K., and Sezer, A. (2007). “Prediction of “Properties of rubberized mortar containing either siliceous or calcitic
sulfate expansion of PC mortar using adaptive neuro-fuzzy methodol- filler.” Int. Conf. on Structural Faults and Repairs, Abstract 95, Engi-
ogy.” Build. Environ., 42(3), 1264–1269. neering Technic Press, Edinburgh, UK.
Kaloush, K. E., Way, G. B., and Zhu, H. (2005). “Properties of crumb Xi, Y., Li, Y., Xie, Z., and Lee, J. S. (2004). “Utilization of solid wastes
rubber concrete.” Trans. Res. Rec., 1914, 8–14. (waste glass and rubber particles) as aggregates in concrete.” Proc.,
Khaloo, A. R., Dahestani, M., and Rahmatabadi, P. (2008). “Mechanical Int. Workshop on Sustainable Development and Concrete Technology,
properties of concrete containing a high volume of tire-rubber par- K. Wang, ed., Beijing, China, 45–54.
ticles.” Waste Manage., 28(12), 2472–2482. Yilmaz, A., and Degirmenci, N. (2009). “Possibility of using waste tire
Khatib, Z. K., and Bayony, F. M. (1999). “Rubberized Portland cement rubber and fly ash with Portland cement as construction materials.”
concrete.” J. Mater. Civ. Eng., 11(3), 206–213. Waste Manage., 29(5), 1541–1546.
Pelisser, F., Zavarise, N., Longo, T. A., and Bernardin, A. M. (2011). Zadeh, L. A. (1999). “Fuzzy sets as a basis for a theory of possibility.”
“Concrete made with recycled tire rubber: Effect of alkaline Fuzzy Sets Syst., 100(1), 9–34.

1358 / JOURNAL OF MATERIALS IN CIVIL ENGINEERING © ASCE / NOVEMBER 2012

J. Mater. Civ. Eng. 2012.24:1350-1358.

You might also like