Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: In this study, an artificial neural networks study was carried out to predict the core compressive strength
Received 24 August 2010 of self-compacting concrete (SCC) mixtures with mineral additives. This study is based on the determi-
Received in revised form 19 October 2010 nation of the variation of core compressive strength, water absorption and unit weight in curtain wall
Accepted 13 November 2010
elements. One conventional concrete (vibrated concrete) and six different self-compacting concrete
Available online 2 June 2011
(SCC) mixtures with mineral additives were prepared. SCC mixtures were produced as control concrete
(without mineral additives), moreover fly ash and limestone powder were used with two different
Keywords:
replacement ratios (15% and 30%) of cement and marble powder was used with 15% replacement ratio
Self-compacting concrete (SCC)
The core compressive strength
of cement. SCC mixtures were compared to conventional concrete according to the variation of compres-
Mineral admixture sive strength, water absorption and unit weight. It can be seen from this study, self-compacting concretes
Artificial neural network consolidated by its own weight homogeneously in the narrow reinforcement construction elements.
Experimental results were also obtained by building models according to artificial neural network
(ANN) to predict the core compressive strength. ANN model is constructed, trained and tested using these
data. The results showed that ANN can be an alternative approach for the predicting the core compressive
strength of self-compacting concrete (SCC) mixtures with mineral additives.
Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.
0950-0618/$ - see front matter Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2010.11.108
4106 M. Uysal, H. Tanyildizi / Construction and Building Materials 25 (2011) 4105–4111
Table 1
Properties of portland cement and mineral additives.
Table 2
Mix proportions of SCC and vibrated concrete for 1 m3.
Materials (kg/m3) Vibrated concrete Control FA15 SCC FA30 SCC LP15 SCC LP30 SCC MP15 SCC
Cement 550 550 467 385 467 385 467
Limestone powder – – – – 83 165 –
Marble powder – – – – – – 83
Fly ash – – 83 165 – – –
Water 182 182 182 182 182 182 182
w/c ratio 0.33 0.33 0.39 0.51 0.37 0.47 0.39
w/p ratio 0.33 0.33 0.33 0.33 0.33 0.33 0.33
Sand 865 869 865 878 866 860 863
CSI 466 467 457 445 464 461 463
CSII 320 311 305 297 311 307 312
pressive strength. For this purpose, a computer program was devel- fly ash, marble powder and limestone powder were used as mineral and filler addi-
tives in SCC to utilize it. Specific surface area by Blaine and 28th day compressive
oped in MATLAB. Furthermore, the results obtained from the ANN
strength of cement were 399.6 m2/kg, and 48.3 MPa, respectively.
model were compared with the average results of the experiments. Marble powder (MP) was provided from a marble managing plant in Bilecik di-
rectly used in SCC without any processes. The specific surface area by Blaine of MP
is 889 m2/kg. Limestone powder was a by-product of quarry crushers and collected
2. Experimental procedure from the filtration system of a quarry crushers. The characteristic properties, min-
eralogical composition and particle size distribution of filler materials are given
2.1. Materials Table 1.
Polycarboxylate based and high range water reducing superplasticizer was also
The Portland cement used in this study was produced according to the European used in the mixtures at the ratio of 1.6% of binder materials by weight for reducing
Standards EN-197/1 and labeled as CEM I/42.5 R. The physical and chemical proper- the water/binder ratio of SCC. The solid content and pH of superplasticizer were 21%
ties of the Portland cement is listed in Table 1. The maximum size of coarse aggre- and 8%, respectively. Tap water used was obtained from the city waterworks of Sak-
gate was selected as 16 mm in order to avoid the blocking effect of SCC. Besides, arya for the production of concrete mixtures during the experimental procedure.
3
2 4
1 5
6
9 10
8 11
7 12
15 16
14 17
13 18
22
21
20 23
19 24
Table 3
Test results of CC, LP15, LP30 and MP15 mixtures in curtain wall element.
2.2. Mixture proportions imen was determined using DIN 4226 [33] and unit weight of each core specimen
was determined using TS EN 12390-7 [34]. Curtain wall specimen is presented in
One conventional concrete (CC, vibrated concrete) and six different SCC mix- Fig. 1.
tures with mineral additives were prepared. Table 2 presents the composition
and labeling of the SCCs. As seen in that table, the mixtures were labeled such that
the ingredients were identifiable from their IDs. One control concrete (without min- 3. Experimental results
eral additives) was produced and fly ash and limestone powder were used with two
different replacement ratios (15%, and 30%) of cement (FA15, FA30, LP15, LP30).
Moreover, marble powder was used with 15% replacement ratio of cement In order to determine the variation of core compressive
(MP15). After the preliminary investigations, the water–powder ratio (w/p) was se- strength, water absorption and unit weight to find out homogene-
lected as 0.33 and the total powder content was fixed to 550 kg/m3. ity, there were tests applied on the wall elements and the results
are summarized in Tables 3 and 4. Fig. 1 also shows core specimen
2.3. Preparation and testing of specimens numbers to compare the variation among core specimens. Mean-
while, number 1 is accepted concrete pouring point. When SCC
After self-compactibility was determined by fresh concrete experiments con-
cretes were poured from a point at the top of the mold and placed in molds without mixtures are compared to conventional concrete as compressive
vibration. Conventional concrete was vibrated by two sides as L shape. All of the strength, there is no significant variance among the results of the
curtain wall specimens were demoulded after a day, then, samples were cured at same mix at different depths of the wall elements in SCC. The obvi-
22 ± 2 °C with 100% RH during 28 days period. At the end of 28 days, by taking into ous trend is that, the bottom specimens have higher strength val-
account reinforcement points, the different height levels of the curtain wall core
samples were removed and these samples were made on the compressive strength,
ues than the top specimens, probably due to higher degree of
water absorption and unit weight tests. Compressive strength of each core speci- compaction due to the weight of the concrete above. This also
men was determined using TS EN 12504-1 [32], water absorption of each core spec- shows that the segregation has not occurred while concrete is
4108 M. Uysal, H. Tanyildizi / Construction and Building Materials 25 (2011) 4105–4111
Table 4
Test results of control, FA15 and FA30 mixtures in curtain wall element.
being poured down a 1.5 m high formwork. This indicates that the than the others. Although limestone powder was assumed to have a
mixes had adequate stability. But, conventional concrete has sig- filling effect, this mixture incorporating high volumes of LP (30%
nificant variance among the results of the bottom specimens and replacement rate with cement), exhibits the highest water absorp-
the top specimens. The variation between the top compressive tion values than other mixtures. This is likely to be due to the high
strength values and the bottom compressive strength is higher in volume of limestone powder hindering the overall cementitious
conventional concrete than all of the SCC. According to compres- reactions, hence creating a more porous microstructure. In general,
sive strength test results, the concrete consists of fly ash 15% any factor tending to improve the compressive strength of the con-
replacement ratio with cement has the highest average compres- crete will have a beneficial effect upon the water-tightness. There-
sive strength values. fore, a similar trend in the compressive strength is observed for
Water absorption determination was done to cylindrical core water absorption results for different compositions.
specimens before compressive strength tests were done. The obvi- Unit weight determination was also done to cylindrical core
ous trend is that, the top specimens have higher water absorption specimens before compressive strength tests were done. The trend
values than the bottom specimens. The bottom specimens consoli- is that, the bottom specimens have higher unit weight values than
date better than the top specimens because of hydrostatic effect and the top specimens. The bottom specimens consolidate better than
bottom specimens have less permeable depending on this situation. the top specimens and they exhibit denser and less porous micro-
Control concrete wall element exhibits the lowest water absorption structure which means higher unit weight values. According to
values comparing to the other compositions. Such results suggest unit weight test results, control concrete exhibits the highest aver-
that control concrete was denser and more resistant to fluid ingress age unit weight values.
M. Uysal, H. Tanyildizi / Construction and Building Materials 25 (2011) 4105–4111 4109
Σ Σ 0
10
Training-Blue Goal-Black
Inputs Σ
-1
10
Σ Σ Training
-2 Goal
10
Σ
Σ -3
10
bias bias
10-4
Fig. 2. Multilayer feed forward neural network structure.
10-5
4. Artificial neural network model for prediction of 0 50 100 150 200 250 300 350 400 450
experimental results 464 Epochs
Cement
Fly ash
Limestone
powders
Marble
powders
Natural
aggregate Compressive
Strength
I number
aggregate
II number
aggregate
Super
plasticizer
Unit
weight
Water
absorption
1.2 1.2
2
R = 0.95 1
2
R = 0.92
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0.5 0.6 0.7 0.8 0.9 1 1.1
Experimental Results Experimental Results
Fig. 5. Linear relationship between measured and predicted compressive strengths Fig. 7. Linear relationship between measured and predicted compressive strengths
(Fletcher–Powell conjugate gradient backpropagation algorithm). (Levenberg–Marquardt backpropagation algorithm).
When the test results were evaluated, it was shown that self- [15] Zong WJ, Guang NH, Yun HJ. The application of automatic acquisition of
knowledge to mix design of concrete. Cem Concr Res 1999;29(12):1857–80.
compacting concretes consolidated by its own weight homoge-
[16] Jepsen MT. Predicting concrete durability by using artificial neural network.
neously in the narrow reinforcement construction elements, Published in a special NCR-publication; ID. 5268; 2002.
when it was compared to conventional concrete and the use [17] Basma AA, Barakat S, Oraimi SA. Prediction of cement degree of hydration
of self-compacting concrete was optimum solution especially using artificial neural networks. Mater J 1999;96(2):166–72.
[18] Graham LD, Forbes DR, Smith SD. Modeling the ready mixed concrete delivery
in the column-beam integrated systems. system with neural network. Automat Constr 2006;15(5):656–63.
Between the algorithms of ANN tested for this study, Fletcher– [19] Yeh C. Exploring concrete slump model using artificial neural networks. J
Powell conjugate algorithm is found to be the best learning Comput Civil Eng ASCE 2006;20(3):217–21.
[20] Al-Metairie N, Terroand M, Al-Khaleefi A. Effect of recycling hospital ash on the
algorithm. The ANN model predicts the core compressive compressive properties of concrete. Build Environ 2004;39(5):557–66.
strength of SCC with the high correlation coefficient. Therefore, [21] Bai J, Wild S, Ware JA, Sabir BB. Using neural networks to predict workability of
the results show that the ANN can be used to predict the core concrete. Adv Eng Softw 2003;34(11–12):663–9.
[22] Oztas A, Pala M, Ozbay E, Kanca E, Caglar N, Bhatti MA. Predicting the
compressive strength of SCC using the ingredients of concrete. compressive strength and slump of high strength concrete using neural
network. Constr Build Mater 2006;20(9):769–75.
[23] Hadi MN. Neural networks applications in concrete structures. Comput Struct
2003;81(6):373–81.
References [24] Waszczyszyn Z, Ziemianski L. Neural networks in mechanics of structures and
materials. Comput Struct 2001;79(22–25):2261–76.
[1] Ozawa K, Maekawa K, Kunishima H, Okamura H. Performance of concrete [25] Ashour A, Alqedra M. Concrete breakout strength of single anchors in tension
based on the durability design of concrete structures. Proc Sec East Asia-Pacific using neural networks. Adv Eng Softw 2005;36(2):87–97.
Conf Struct Eng Const 1989;1:445–56. [26] Mukherjee A, Biswas SN. Artificial neural networks in prediction of mechanical
[2] Rols S, Ambroise J, Ṕera J. Effects of different viscosity agents on the properties behavior of concrete at high temperature. Nucl Eng Des 1997;178(1):1–11.
of self-levelling concrete. Cem Concr Res 1999;29(2):261–6. [27] FanChou S, Pellinen TK. Assessment of construction smoothness specification
[3] Okamura H et al. Self-compacting high performance concrete. In: proceedings pay. J Transport Eng 2005;131(7):563–70.
of the fifth EA SEC, vol. 3; 1995. p. 2381–8. [28] Tam CM, Fang CF. Comparative cost analysis of using high performance
[4] Campion MJ, Jost P. Self-compacting concrete. Concr Int: Des Constr concrete in tall building construction by artificial neural networks. Struct J
1999;22(4):31–4. 1999;96(6):927–36.
[5] Khayat KH, Manai K, Trudel A. In situ mechanical properties of wall elements [29] Tarefder RA, White L, Zaman M. Neural network model for asphalt concrete
using self-consolidating concrete. ACI Mater J 1997;94(6):491–500. permeability. J Mater Eng 2005;17(1):468–73.
[6] Yurugi M. Application of self-compacting concrete in Japan. In: proceedings [30] Pala M, Ozbay E, Oztas A, Yuce MI. Appraisal of long-term effects of fly ash and
23rd OWICS conference, Singapore: CI-Premier Ptc. Ltd.; 1998. p. 29–42. silica fume on compressive strength of concrete by neural networks. Constr
[7] Petersson O. Application of self-compacting concrete for bridge castings. Swed Build Mater 2007;21(2):384–94.
Cem Concr Res Inst 1998;54(2):5. [31] Ji T, Lin T, Lin X. A concrete mix proportion design algorithm based on artificial
[8] Arima I, Takeguchi M, Sakurai S, Hayashi J. The quality of superworkable neural network. Cem Concr Res 2006;36(7):1399–408.
concrete used for the construction of Akashi-Kaikyo Bridge 4A anchorage. Proc [32] TS EN 12504-1. Testing concrete in structures – part 1: cored specimens –
Jpn Concr Inst 1994;16(1):25–30. taking, examining and testing in compression. TSE Ankara; 2002 [in Turkish].
[9] Nehdi M, Pardhanb M, Koshowskic S. Durability of self-consolidating concrete [33] DIN 4226. Test method for determination the specific gravity the absorbtion
incorporating high-volume replacement composite cements. Cem Concr Res water and the void ratio in hardened concrete; 1981.
2004;34(11):2103–12. [34] TS EN 12390-7. Testing hardened concrete – part 7: density of hardened
[10] Bartos PJM, Grauers M. Self-compacting concrete. Concrete 1999;3(4):9–13. concrete. TSE Ankara; 2010 [in Turkish].
[11] Skarendahl A. Market acceptance of self-compacting concrete. The Swedish [35] Hanbay D, Turkoglu I, Demir Y. An expert system based on wavelet
experience. In: Ozawa K, Ouchi M, editors. Proc. 2nd inter. symp. on SCC, Coms decomposition and neural network for modeling chua’s circuit. Exp Syst
Engineering Corporation Tokyo; 2001. p. 1–13. Appl 2008;34(4):2278–83.
[12] Okamura H, Ozawa K, Maekawa K, Tangtermsinikul S. High performance [36] Haykin S. Neural networks, a comprehensive foundation. College Publishing
concrete mechanism of super-fluidized concrete. EIT-JSCE-AIT joint seminar Comp. Inc.; 1994.
on solution to urban infrastructure problems through civil engineering [37] Hanbay D, Turkoglu I, Demir Y. Prediction of wastewater treatment plant
technology, Bangkok; 1992. p. 16. performance based on wavelet packet decomposition and neural networks.
[13] Ozawa K, Maekawa K, Kunishima M, Okamura H. High-performance concrete Exp Syst Appl 2008;34(2):1038–43.
based on the durability of concrete structures. In: proceedings of the second [38] Bilim C, Atisß CD, Tanyildizi H, Karahan O. Predicting the compressive strength
East Asia Pacific conference on structural engineering and construction; 1989. of ground granulated blast furnace slag concrete using artificial neural
[14] Yeh I. Modeling of strength of high-performance concrete using artificial network. Adv Eng Softw 2009;40(5):334–40.
neural networks. Cem Concr Res 1998;28(12):1797–808.