Construction

and Building

Construction and Building Materials 21 (2007) 1928–1939

MATERIALS
www.elsevier.com/locate/conbuildmat

Optimization on proportion for recycled aggregate in concrete

using two-stage mixing approach
a,*
Vivian W.Y. Tam , C.M. Tam b, Y. Wang c

a
Griffith School of Engineering, Gold Coast Campus, Griffith University PMB50 Gold Coast Mail Centre, Qld 9726, Australia
b
Department of Building and Construction, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong
c
Department of Computer Science, Tsinghua University, China

Received 27 September 2005; received in revised form 27 February 2006; accepted 31 May 2006
Available online 25 September 2006

Abstract

Recycled aggregate (RA) is well acknowledged having a poorer quality due to its higher porosity resulted from cement mortar remains
attaching to its surface that hampers the recycling rate of concrete waste. Many previous researches recorded reduction in strength for
concrete made with RA. As a result, the use of RA is mainly confined to low-grade applications. Tam et al. [Tam WYV, Gao XF, Tam
CM. Micro-structural analysis of recycled aggregate concrete produced from two-stage mixing approach. Cem Concr Res
2005;35(6):1195–203] proposed a two-stage mixing approach (TSMA) for improving the strength of recycled aggregate concrete
(RAC), by testing mixes with up to 30% RA replacement. This paper extends Tam et al.’s work (2005) by exploring RA substitutions
ranging from 0% to 100% and compares their performance with the traditional mixing procedure. Based upon the experimental works,
improvements on strength and rigidity of RAC using TSMA were compared with those of traditional mixing procedure based on dif-
ferent percentages of RA replacements. The results were then optimized using general regression neural networks (GRNN) and RA
replacements of 25–40% and 50–70% were found to be optimal when TSMA was adopted. It confirms the conservative recommendation
of 20% RA substitution by many previous researchers and public users.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Optimization; Strength; Recycled aggregate; Concrete; Construction

1. Introduction In the Netherlands, about 14 million tons of building and

demolishing wastes per annum are produced, in which about
Sustainable development has been advocated all over 8 million tons are recycled, mainly for unbound road base
the world [1]. ERL [2] indicated that the amount of demo- courses [4]. National waste strategies in UK require increase
lition debris dumped at landfill sites in the United King- in household waste recycling and recovery rates and these
dom is in excess of 20 million tons per annum. The bulk measures will require the development of the infrastructure
of this material is concrete (50–55%) and masonry (30– to support national high-intensity recycling and composting
40%) with only small percentages of other materials such schemes and the construction of at least 35 new municipal
as metals, glass and timber [3]. Recovery of these waste wastes incinerators. Although 40% of demolition waste in
materials can reduce the consumption rate of landfill areas the United Kingdom is recycled, they are mainly used for
and natural gravel resources. low grade application such as fills and hardcore. Higher
grade utilization such as concrete has been discouraged by
a lack of suitable specifications [5]. The demand from low
grade applications is almost saturated and a further increase
*
Corresponding author. in reuse must be realized through other sectors of the build-
E-mail address: v.tam@griffith.edu.au (V.W.Y. Tam). ing industry, such as aggregates for concrete.

0950-0618/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2006.05.040
 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1929

This paper aims at the following objectives: water/cement ratio of 0.45 and grading requirements
according to BS 882 [10].
(i) Study the extent of construction waste; The RA used for the experiments were collected from
(ii) Examination of the current obstacles on recycling the Tuen Mun Area 38 recycling plant with average
concrete wastes; absorption values of 1.65% for 20 mm and 2.63% for
(iii) Comparison and analysis of recycled aggregate con- 10 mm aggregate. The quality of RA is poorer than virgin
crete (RAC) made from the two-stage mixing aggregate (water absorption of virgin aggregate at 0.77%
approach (TSMA) developed by Tam et al. [6] with and 0.57% for 20 mm and 10 mm respectively), indicating
recycled aggregate (RA) replacements ranging from that the major weakness of RA is its high porosity.
0% to 100% and that of the traditional mixing proce- The compressive and flexural strengths of the mixes
dure; and were then compared, which are the most important
(iv) Optimization of percentages of RA replacements mechanical properties of concrete in concrete mix design.
using TSMA by general regression neural networks Cubes of 100 mm and 100 mm · 500 mm beams were used
(GRNN). for testing the compressive and flexural strength develop-
ment respectively under the standard curing conditions
for 7, 14, 28 and 56 days. Three cubes and beams were
2. Obstacles in using recycled aggregate tested at 7, 14, 28 and 56 days and then the average
strength was taken, according to BS 1881: Part 116 [11]
Although it is environmentally beneficial to use RA, the and BS 1881: Part 118 [12] for compressive and flexural
current legislation and experience, however, are not suffi- strength respectively. Density of hardened concrete was
cient to support and encourage recycling of demolished also measured. Furthermore, the static modulus of elastic-
concrete waste. These technical problems, include weak ity using cylinders of 100 mm diameter under 28-day curing
interfacial transition zones between cement paste and conditions were examined based on BS 1881: Part 121 [13].
aggregate, porosity and traverse cracks within demolished Three preloading cycles, using the same loading and
concrete, high level of sulphate and chloride contents, unloading rate, were used. One hundred and sixty-eight
impurity, cement remains, poor grading, high variations specimens of 500 mm · 100 mm beams, cubes of 100 mm
in quality, which render the use of RA for structural appli- and cylinders of 100 mm diameter had been used for mea-
cations difficult [7,8]. suring the density, compressive strength, flexural strength
and modulus of elasticity respectively.
3. Two-stage mixing approach
5. Experimental results
In order to improve the quality of recycled aggregate
All the mix proportions of recycled aggregate concrete
concrete, a mixing method: two-stage mixing approach
mixed using TSMA and NMA were collected with a slump
(TSMA) was developed by Tam et al. [6], which divides
of 75 mm as required by BD [9]. The results of density,
the mixing process into two parts and proportionally splits
compressive strengths, flexural strength and static modulus
the required water into two parts which are added after
of elasticity to mixing with different proportions of RA
mixing one part with fine and coarse aggregate and cement;
using NMA and TSMA are shown in Figs. 1–13 respec-
while the normal mixing approach only puts all the ingre-
tively. A clear strength enhancement in using TSMA can
dients of concrete and mix them. Improvement of strength
be found when compared with that of NMA.
can be achieved up to 21.19% for 20% of RA under 28-day
In TSMA, during the first stage of mixing, the use of half
curing conditions using TSMA as recorded by Tam et al.
of the required water for mixing leads to the formation of a
[6].
thin layer of cement slurry on the surface of RA which per-
meates into the porous old cement mortar, filling up the old
4. Experimental method cracks and voids. In the second stage of mixing, the remain-
ing water is added to complete the cement hydration pro-
The experimental works by Tam et al. [6] was limited to cess. A stronger interfacial zone (ITZ) is thus developed.
30% RA substitutions. This study further explores RA The quality of ITZ depends on the surface characteris-
replacement ratios of 40%, 50%, 60%, 70%, 80%, 90% tics of the aggregate particles, the degree of bleeding, chem-
and 100% experimented under the normal mixing approach ical bonding and the specimen preparation technique
(NMA) and TSMA in addition to the proportions of 0%, which, however, are difficult to measure. Although these
5%, 10%, 15%, 20%, 25% and 30% previously recorded effects have been reported by some investigators, the results
by Tam et al. [6]. The designated mix proportions follow are difficult to reconcile. Nonetheless, it is generally agreed
the specifications of Buildings Department (BD) of the that as the paste-aggregate bond strength increases, the
Hong Kong Special Administrative Region [9] with concrete strength also increases [14]. Fig. 14 illustrated
1:1.8:1.8:0.9 (ordinary Portland cement/fine aggregate/ the interfacial zone of concrete using TSMA, which is
20 mm coarse aggregate/10 mm coarse aggregate) with a much better than that of NMA shown in Fig. 15.
1930 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939

2550

2500

NMA
Density (kg/m3)

2450
TSMA

2400

2350

2300
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 1. Density of mixes with different proportions of RA using NMA and TSMA at 7 days.

2440

2420

2400

NMA
Density (kg/m3)

2380
TSMA

2360

2340

2320

2300
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 2. Density of mixes with different proportions of RA using NMA and TSMA at 14 days.

6. Optimization on RA proportions in using TSMA sary for further analysis, like comparison, accumulation,
and optimization. The crystallization of concrete is, how-
Because the proportions of RA replacement are taken in ever, a complex physical–chemical process affected by many
discrete sampling space, regression of discrete experimental ingredients that are hard to control, so the experimental
data into a continuous real function is important and neces- data is usually inevitably noisy. In the some cases, large
 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1931

2460

2440

2420

2400 NMA
Density (kg/m3)

TSMA
2380

2360

2340

2320

2300
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 3. Density of mixes with different proportions of RA using NMA and TSMA at 28 days.

2420

2400

2380

NMA
Density (kg/m3)

2360
TSMA

2340

2320

2300

2280
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 4. Density of mixes with different proportions of RA using NMA and TSMA at 56 days.

noises even fake data with high frequency can be found. In After attempting various mathematical models including
comparing the two mixing methods, TSMA and NMA, polynomial curve fitting, multi-layer perceptron (MLP) and
convictive continuous regression results were assumed so radial basis network (RBN) (all of which bear some prob-
that the representative trends amongst these noisy data lems in analyzing the experimental results [15–17]), general
could be modeled with robust mathematical methods. regression neural network (GRNN) was found to work
1932 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939

60

50
Compressive Strength (MPa)

40
NMA

TSMA
30

20

10

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 5. Compressive strengths of mixes with different proportions of RA using NMA and TSMA at 7 days.

70

60
Compressive Strength (MPa)

50

NMA
40
TSMA

30

20

10

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 6. Compressive strengths of mixes with different proportions of RA using NMA and TSMA at 14 days.

better in harmonizing the noises by introducing a normal- taken with Matlab version 7. Improvements of compressive
ized dot product weight function to substitute the weighted strength, flexural strength, and static modulus of elasticity
sum in RBN [18]. It presents the improvement resulted from of concrete samples mixed by TSMA and NMA with
different proportions of RA substitutions recorded in the respect of various percentages of RA substitutions and cur-
various experimental results. The modeling work is under- ing times are shown in Figs. 16–24 respectively. It can be
 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1933

70

60
Compressive Strength (MPa)

50

NMA
40
TSMA

30

20

10

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 7. Compressive strengths of mixes with different proportions of RA using NMA and TSMA at 28 days.

90

80

70
Compressive Strength (MPa)

60
NMA
50 TSMA

40

30

20

10

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 8. Compressive strengths of mixes with different proportions of RA using NMA and TSMA at 56 days.

seen that the values obtained by TSMA are higher than local maxima with different RA substitutions are also high-
those of NMA for any percentage of RA substitution. lighted in each figure. The local maxima and their improve-
The noise-tolerant regression, it can be used to compare ment percentages are tabulated in Table 1.
TSMA with NMA quantitatively. The percentage improve- From the results of the compressive strength shown in
ments’ of TSMA over NMA are shown in Figs. 16–24. The Figs. 16–19, it is clearly indicated that around 25–40% of
1934 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939

6
Compressive Strength (MPa)

NMA
4
TSMA

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 9. Flexural strengths of mixes with different proportions of RA using NMA and TSMA at 7 days.

6
Compressive Strength (MPa)

5 NMA

TSMA
4

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 10. Flexural strengths of mixes with different proportions of RA using NMA and TSMA at 14 days.

RA substitutions yield optimal properties using TSMA (see Table 1). Another local maximum happens at around
with improvements of 17.39%, 10.83%, 17.68% and 55–70% of RA substitution with improvements of
12.70% in 7, 14, 28 and 56 days of curing [the optimal 11.93%, 6.03%, 7.82% and 9.13% of 58.59%, 68.69%,
RA percentages obtained were 37.37%, 36.36%, 24.24% 58.59% and 56.57% of RA substitution after 7, 14, 28
and 39.39% in 7, 14, 28 and 56 days of curing respectively] and 56 days of curing respectively.
 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1935

6
Compressive Strength (MPa)

5 NMA

TSMA
4

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 11. Flexural strengths of mixes with different proportions of RA using NMA and TSMA at 28 days.

6
Compressive Strength (MPa)

5 NMA

TSMA
4

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 12. Flexural strengths of mixes with different proportions of RA using NMA and TSMA at 56 days.

Similar situations occurred in flexural strength and static tively (see Figs. 20–23). At 31.31% of RA substitutions, an
modulus of elasticity. The optimal points occurred at RA improvement of 14.35% on static modulus of elasticity was
substitutions of 39.39%, 28.28%, 31.31% and 29.29% with recorded (see Fig. 24). Furthermore, improvements in flex-
improvements in flexural strength of 21.79%, 21.54%, ural strength of 3.56%, 9.38%, 2.25% and 2.40% with RA
7.59% and 12.86% at 7, 14, 28 and 56 days of curing respec- substitutions of 60.61%, 60.61%, 69.70% and 67.68% at 7,
1936 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939

40000

35000

30000
Modulus of Elasticity (N(mm2)-1)

25000
NMA

20000 TSMA

15000

10000

5000

0
0% 20% 40% 60% 80% 100% 120%
% of RA
Fig. 13. Modulus of elasticity of mixes with different proportions of RA using NMA and TSMA.

Fig. 14. Interfacial zone for TSMA.

Fig. 15. Poorer new interfacial zone for NMA.

Fig. 16. Improvement on compressive strength with 7-day curing.

V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1937

Fig. 17. Improvement on compressive strength with 14-day curing.

Fig. 18. Improvement on compressive strength with 28-day curing.

Fig. 19. Improvement on compressive strength with 56-day curing.

Fig. 20. Improvement on flexural strength with 7-day curing.

Fig. 21. Improvement on flexural strength with 14-day curing.

Fig. 22. Improvement on flexural strength with 28-day curing.

1938 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939

Fig. 23. Improvement on flexural strength with 56-day curing.

Fig. 24. Improvement on static modulus of elasticity with 28-day curing.

Table 1
Local maxima of percentages of RA and respective improvement percentages
Strength Curing conditions (days) Local maxima (% of RA) Improvement when compared with
NMA (%)
Compressive strength 7 6.06 37.37 58.59 10.54 17.39 11.93
14 10.10 36.36 68.69 10.23 10.83 6.03
28 24.24 58.59 81.82 17.68 7.82 3.03
56 39.39 56.57 12.70 9.13
Flexural strength 7 2.02 22.22 39.39 60.61 12.79 14.00 21.79 3.56
14 28.28 60.61 90.91 21.54 9.38 8.44
28 31.31 69.70 94.95 7.59 2.25 3.29
56 10.10 29.29 67.68 10.44 12.86 2.40
Static modulus of elasticity 28 31.31 46.46 89.90 14.35 12.07 4.40

14, 28 and 56 days of curing are recorded respectively. An aggregate concrete. A series of full scale experiments testing
improvement in static modulus of elasticity of 12.07% with RA substitution between 0% and 100% is reported in this
RA substitution of 46.46% was also recorded. paper. Optimization of RA substitution is modeled.
With the inherent weakness of RA including high poros- Around 25–40% of RA substitution is found to be most
ity, high amounts of cracks, high level of impurity and high favourable in using TSMA. Further, around 50–70% of
cement mortar remains, the mechanical properties of RAC RA replacement can also give some improvement although
are affected, which is thus limited to lower-grade applica- the enhancements are less significant when compared with
tions. The pre-mix process can provide a better interfacial that of 25–40%.
behaviour between the aggregate and cement paste. This
study has identified that the optimal proportions of RA Acknowledgement
when using TSMA is around 25–40% and 50–70% of RA
substitution. Therefore, the current regulation in Hong The work described in this paper, was fully supported by
Kong and other countries limiting the adoption of 20% a grant from the Housing Authority Research Fund of the
of RA substitution is on the conservative side. The use of Hong Kong Special Administrative Region, China (Project
TSMA can allow higher percentages of RA substitution. Ref. No. 9460004).

7. Conclusion References

Recycled aggregate is strongly advocated by the Hong [1] Kawano H. The state of reuse of demolished concrete in Japan. In:
Kong government such as the motives of setting up the Integrated design and environmental issues in concrete technology:
proceedings of the international workshop ‘rational design of
centralized recycling plant, issuance of practice guidelines,
concrete structures under severe conditions:’ Hakodate, Japan, 7–9
specification and technical circulars in guiding the use of August 1995. London: E&FN Spon; 1995. p. 243–9.
recycled aggregate. The porous nature of recycled aggre- [2] ERL (Environmental Resources Limited). Demolition waste. Lan-
gate reduces the mechanical properties of the recycled caster, (England): Construction Press Limited; 1980.
 V.W.Y. Tam et al. / Construction and Building Materials 21 (2007) 1928–1939 1939

[3] Mulheron M. The recycling of demolition debris: current practice, [9] BD (Buildings Department). Practice note for authorized persons and
products and standards in the United Kingdom. Demolition and registered structural engineers: use of recycled aggregates in concrete.
reuse of concrete and masonry: reuse of demolition waste. London: Hong Kong, 2003.
Chapman & Hall; 1988. p. 510–519. [10] BS 882. Specification for aggregates from natural sources for
[4] Hendriks CF. Certification system for aggregates produced from concrete. London: (United Kingdom): British Standards Institution;
building waste and demolished buildings. In: Environmental aspects 1992.
of construction with waste materials: proceeding[s] of the interna- [11] BS 1881: Part 116. Method for determination of compressive strength
tional conference on environmental implications of construction of concrete cubes. London: (United Kingdom): British Standards
materials and technology developments, Maastricht, The Nether- Institution; 1983.
lands; 1–3 June 1994. p. 821–34. [12] BS 1881: Part 118. Method for determination of flexural
[5] Collins RJ. Reuse of demolition materials in relation to specifications strength. London: (United Kingdom): British Standards Institu-
in the UK. In: Demolition and reuse of concrete and masonry: tion; 1983.
guidelines for demolition and reuse of concrete and masonry: [13] BS 1881: Part 121. Method for determination of static modulus of
proceedings of the third international RILEM symposium on elasticity in compression. United Kingdom: British Standard Institu-
demolition and reuse of concrete masonry, Odense, Denmark, 24– tion; 1983.
27 October 1993. London: E&FN Spon; 1993. p. 49–56. [14] Mindess JF, Young DD. Concrete. Upper Saddle River, (NJ):
[6] Tam WYV, Gao XF, Tam CM. Micro-structural analysis of recycled Prentice Hall; 2003.
aggregate concrete produced from two-stage mixing approach. Cem [15] Cybenko G. Approximation by superpositions of a sigmoidal
Concr Res 2005;35(6):1195–203. function. Mathe Control Signals Systems 1989;2:303–14.
[7] Chan CY, Fong FK. Development in recycling of construction and [16] Hartman EJ, Keller JD, Kowalski JM. Layered neura networks with
demolition materials. Civil Engineering Department, Hong Kong, Gaussian hidden units as universal approximations. Neural Computa
Homepage. Available from: http://www.ced.gov.hk/eng/services/ 1990;2(2):210–5.
recycling/development.pdf, 2002. [17] Mitchell TM. Machine learning. McGraw-Hill Companies, Inc;
[8] Kawano H. Barriers for sustainable use of concrete materials. 1997.
Concrete technology for a sustainable development in the 21st [18] Specht DF. A general regression neural network. IEEE Trans on
century. London: New York: E&FN Spon; 2000. p. 288–93. Neural Networks 1991;2(6):568–75.