Construction and Building Materials 166 (2018) 472–481

Contents lists available at ScienceDirect

Construction and Building Materials

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

Design of fly ash geopolymer concrete mix proportions using

Multivariate Adaptive Regression Spline model
Weena Lokuge a, Aaron Wilson a, Chamila Gunasekara b,⇑, David W. Law b, Sujeeva Setunge b
a
School of Civil Engineering and Surveying, University of Southern Queensland, Springfield, Queensland 4300, Australia
b
Civil and Infrastructure Engineering, School of Engineering, RMIT University, 124, La Trobe Street, Melbourne, Victoria 3000, Australia

h i g h l i g h t s

Developed a mix design method for fly ash geopolymer concrete using MARS model.

Key variables are water/solid, activator/fly ash, Na2SiO3/NaOH and NaOH molarity.

Contour plots can be used to design fly ash geopolymer concrete mix proportions.

Targeted 28-day strength from 30 MPa to 55 MPa achieved in laboratory experiments.

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

Article history: Many research studies have been conducted during recent years on the topic of geopolymer materials
Received 11 December 2017 based on the engineering performance of the concrete. What has been missing is the combination of this
Received in revised form 25 January 2018 research in a way that would provide a simple to use design tool for geopolymer concrete as a replace-
Accepted 29 January 2018
ment to concrete based on Portland Cement. This research paper addresses this requirement for develop-
Available online 22 February 2018
ing a standard mix design method for Class F, low calcium fly ash based geopolymer concrete using
Multivariate Adaptive Regression Spline (MARS) model. Published geopolymer concrete research data
Keywords:
was combined into a database and analysed to give the ratios of water/solid, alkaline activator/fly ash,
Fly ash
Geopolymer concrete
Na2SiO3/NaOH, and NaOH molarity. Targeted compressive strengths ranging from 30 MPa to 55 MPa at
Multivariate Adaptive Regression Spline 28 days were achieved with laboratory experiments, using the proposed MARS mix design methodology.
model Thus, this tool has the capability to provide a novel approach for the design of geopolymer concrete mixes
Mix design to achieve the desired compressive strength appropriate for the construction requirement.
Compressive strength Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction This increased demand for housing, not only in China but world-
wide, is feeding the global demand for building materials, in partic-
1.1. Background ular Portland Cement (PC) as the main binder in concrete.
Globally, we currently use approximately 2.8 billion tonnes of
As the population of the world increases so too does the PC per annum and this is expected to increase to at least 4 billion
requirement for housing and development of infrastructure. tonnes per annum. The PC manufacturing process generates
Berkelmans and Wang [1] estimated that 1.9 billion square metres between 0.6 and 0.8 kg of CO2 per every kilogram of cement pro-
of residential floor space was built in China alone in 2011. To put duction [2,3], and this is responsible for 5 to 7% of CO2 emissions
this into perspective in one year China built as much floor space worldwide [4,5]. The damage that this level of pollution is doing
as there is in all of Spain. It is also estimated that this growth will to the atmosphere is unsustainable, and this has inspired research
not peak until 2017, although others believe that this decline in into environmentally friendly geopolymer concrete. This sustain-
construction will be short lived due to the underlying demand able geopolymer concrete can reduce CO2 emission between 26
which is driven by higher salaries and increased urban population. and 45% with the 100% replacement of PC in concrete [6,7]. Low
calcium, Class F fly ash has been identified as a possible source
material for geopolymer concrete [8–12]. Other concerns have also
⇑ Corresponding author.
highlighted the use of coal as a primary energy source and the
E-mail addresses: Weena.Lokuge@usq.edu.au (W. Lokuge), chamila.
gunasekara@rmit.edu.au (C. Gunasekara), david.law@rmit.edu.au (D.W. Law), release of fly ash as a by-product, some of which becomes environ-
sujeeva.setunge@rmit.edu.au (S. Setunge). mental waste [13]. In addition to the CO2 reduction attributable to

https://doi.org/10.1016/j.conbuildmat.2018.01.175
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481 473

the use of geopolymer concrete, fly ash has an important advan- 2. Proposed mix design procedure
tage as it is a by-product which would be a waste product requir-
ing disposal with associated costs [5]. 2.1. Mix design database
In the geopolymeric reaction, silica and alumina oxides con-
tained in fly ash react under highly alkaline conditions, typically A comprehensive literature review was conducted to establish a
provided by sodium based hydroxide and silicate solution, and pro- database for mix designs of low calcium, Class F 100% fly ash
duce an amorphous three-dimensional network of silicon and alu- geopolymer concrete, published between 2000 and 2016 based
minium atoms linked by oxygen atoms in a fourfold coordination. on 28-day compressive strength. Much of the initial research on
The presence of sodium cations is important due to charge balanc- 100% fly ash geopolymer was based on geopolymer paste and mor-
ing and catalytic properties. The main outcome of the geopoly- tar. The database was used to identify those publications that pro-
meric reaction is thus sodium-aluminosilicate gel, which governs vided the experimental details required for the MARS model.
the performance of geopolymer concrete.
To date many of the mix design parameters both chemical and 2.2. Multivariate Adaptive Regression Splines (MARS) model
physical have been identified as affecting the compressive strength
of 100% low calcium fly ash based geopolymer concrete, such as The MARS model was originally proposed by Friedman [51]. It is
particle size distribution and specific surface area of the fly ash, a form of a stepwise linear regression and suitable for higher
SiO2/Al2O3 and Na2O/Al2O3 ratio of fly ash, amorphous and CaO dimensional inputs. In the reported literature [52], The MARS
content of fly ash and workability [14–18]. On the other hand, model appears to be more popular than its counterparts such as
Hardjito and Rangan [16] reported that the water to solid ratio of Artificial Neural Networks and Extreme Learning Machine [52]
geopolymer mix has a positive influence to the final compressive because of its adaptively synthesized model structure compared
strength. Moreover, Rattanasak and Chindaprasirt [19] showed to the fixed model structures of its counterparts. MARS predictive
that a higher molarity of sodium hydroxide solution increases modelling has been widely used in hydro-meteorological analysis,
the leaching behaviour of aluminates and silicates from raw fly most recently in predicting the evaporation loss [53,54], and in
ash, which resulted in a higher geopolymerization reaction at an predicting the behaviour of fiber reinforced polymer confined con-
early stage. Bakiri et al. [20] and Songpiriyakij et al. [21] reported crete [55].
that the sodium silicate to sodium hydroxide ratio is a critical In MARS algorithm, training data sets are divided into separate
parameter in strength development, while literature [16,22–24] piecewise linear segments (splines) of different gradients (slopes).
reported that the curing temperature and curing time also have a These splines are connected together smoothly and the piecewise
positive effect on material properties of fly ash geopolymer curves are known as basis functions (BFs) producing a model able
concrete. to handle linear as well as nonlinear behaviour. The connection
Despite the in-depth research carried out in the field of points are called knots. Between any two knots, MARS charac-
geopolymer, the development of a suitable mix design procedure terises data either globally or using linear regression. BF(x) is the
for this novel concrete is still in the experimental stage. Ferdous basis function for the x intersects at the knot.
et al. [25], Lloyd and Rangan [26] and Anuradha et al. [27] proposed Let Y be the target dependent variable and X = (x1, x2. . .. . .xp) be
different methods to calculate mix proportions, however, all these the input independent variables. For a continuous response, the
are largely dependent on the trial and error approach. Moreover, relationship between Y and X can be expressed using the MARS
Montes et al. [28] clearly evidenced that unlike PC concrete, calcu- model:
lating the mix proportions of geopolymer concrete is more compli-
cated owing to range of variables being involved in the mix design. Y ¼ f ðx1 ; x2 . . . . . . . . . . . . xp Þ þ e ¼ f ðXÞ þ e ð1Þ
In this study, four key factors have been identified from the
where e is the fitting error and f(X) is the MARS model with the BFs.
review of the literature, water/solid ratio, alkaline activator/fly
For simplicity of this research, only piecewise linear functions are
ash ratio and Na2SiO3/NaOH ratio, and NaOH molarity. Based on
considered. In the MARS environment, the following expression
the parameters, a new standard mix design procedure for 100%
can be used to predict compressive strength of fly ash based
Class F, low calcium fly ash geopolymer concrete has been devel-
geopolymer concrete (Y).
oped using multivariate adaptive regression spline (MARS) model.
This method has then been used to develop mix designs for a range X
N
of compressive strengths. The model developed and the results of Y ¼ c0 þ cm BFðxÞ ð2Þ
the mix designs are presented in the paper to demonstrate the per- n¼1

formance of the model.

x is the input variable, c0 is a constant and cm is the coefficient of BF
(x).
During the construction of MARS model, the basis functions are
1.2. Significance of research
selected based on the generalized cross validation (GCV).
0 1
Published research to date on fly ash based geopolymer con- PN
 y^i Þ A
2
1 m¼1 ðyi
crete has primarily focused on the effect of physical and chemical GCVðNÞ ¼ @ 2
ð3Þ
n ð1  CðMÞ Þ
properties of precursor fly ash and ratio of mix proportions on the n
compressive strength evolution. However, only very limited
research has focused on developing a standard mix design proce- n is the number of data points, yi is the actual value of data point i, y^i
dure for geopolymer concrete in order to obtain a desired strength is the predicted value for data point i and C(M) is the penalty factor
grade. Thus, this research, addresses the development of a standard defined as follows:
mix design method for 100% Class F, low calcium fly ash based CðMÞ ¼ M þ dM ð4Þ
geopolymer concrete using multivariate adaptive regression spline
(MARS) model. Hence, the statistical model presented here will be d is the cost penalty factor of each basis function optimisation.
extremely useful to calculate mix proportions for target compres- When several basis functions are selected in the forward phase,
sive strength of 100% Class F, low calcium fly ash geopolymer over-fitting can occur. Therefore, deleting some basis functions in
concrete. the backward phase is important to select the optimised model.
474 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481

2.3. MARS model development A scatter plot of the actual and predicted compressive strength
in the training and testing regimes are shown in Fig. 1. Straight line
The algorithm for the predictive model for geopolymer concrete shows when actual and predicted compressive strengths are equal.
mix design was developed using MATLAB programming environ- Fig. 1 illustrates that the training data set gives a better correla-
ment. In order to develop the MARS model, the database shown tion (0.768) than the testing data set. This is also evident from the
in Table 1 was analysed to establish the key mix design parame- other two performance indices RMSE and MAE (Table 3). Small val-
ters. These were identified as the water/solid ratio, alkaline activa- ues for RMSE and MAE and large value for R indicate a better rela-
tor/fly ash ratio, Na2SiO3/NaOH ratio and NaOH molarity as tionship between the predicted and the actual compressive
predictive variables. Table 2 shows the statistics for both input strengths. The adopted model can be further refined using a larger
variables and the output considered in the model development. database in the future to gain better model performance. Fig. 2
The low values of skewness and kurtosis for water/solid, NaOH shows the normalized predicted compressive strengths with
molarity and compressive strength are an indication of the asym- respect to the data points related to the training and testing phase.
metry about the mean values and they are light tailed too. There is a slight deviation of the predicted compressive strengths
When developing the MARS predictive model, it is important to for both training and testing data sets, which is a function of the
select a training data subset and a testing subset to evaluate the limitation of data available in the literature.
model performance. The portioning of the available database Having trained and tested the model developed in this research,
between the training and testing subsets was decided based on mix designs for four target compressive strengths are determined
each application and there is no fixed approach for this division. and an experimental program is used to validate the model.
In the literature, it is reported that 63–80% of the available data
has been used for the training [52,56–58]. In a more recent study,
3. Design of geopolymer concrete mixes using MARS model
60% of the data was selected for training, 20% selected for testing
while the remaining was selected for validation [52]. It was
3.1. Contour plots
decided in this study to select 68 (70%) of the available data for
the training and use the remainder as the testing subset. A random
The six different contour plots obtained from MARS model are
sampling process was used in partitioning the database into train-
illustrated in Fig. 3(a–f). As seen from the previous section, the
ing and testing to achieve optimum results. The randomly sampled
compressive strength of geopolymer concrete is dependent on all
data set was used in the development and training stage of the
the four variables that have been identified. Thus, it is necessary
model while the testing dataset was used in the model verification
to use all of these contour maps to produce a mix design to the
stage. As the initial stage of the model development, all the input
required compressive strengths at 28-days.
and output data sets were normalized to get a range between 0
and 1 using the following equation:
(a) Effect of water/solid ratio
ðx  xminimum Þ
xnormalized ¼ ð5Þ
ðxmaximum  xminimum Þ Fig. 3(a) and (d) show that when the water/solid ratio is smaller,
the compressive strength is larger and vice versa. However, Fig. 3
where x is any data point (input or output), xnormalised is the normal- (e) indicates that even if the water/solid ratio is lower, a higher
ized value of the data set, xminimum is the minimum value of the set compressive strength can still be achieved if the Na2SiO3/NaOH
of data and xmaximum is the highest value of the same data set. In the ratio is higher.
MARS model construction, 19 basis functions were used in the for-
ward phase and 6 of them were deleted in the backward phase leav- (b) Effect of activator/fly ash ratio
ing 13 basis functions in the final optimum MARS model.
Fig. 3(a) shows that a combination of a lower activator/fly ash
2.4. MARS model evaluation ratio and a lower water/solids ratio will yield a higher compressive
strength. On the contrary, Fig. 3(b) shows that lower activator/fly
Once the predictive model is developed, it is important to test ash ratio combined with a higher Na2SiO3/NaOH ratio can result
the model by using the actual and predicted compressive strengths in higher compressive strengths. Fig. 3(c) indicates that for a cer-
of geopolymer concrete. The performance of the MARS model tain NaOH molarity, if the activator/fly ash ratio changes then dif-
developed was evaluated using coefficient of correlation (R), mean ferent compressive strengths can be achieved.
square error (RMSE) and mean absolute error (MAE) [52].
Pn (c) Effect of Na2SiO3/NaOH ratio
 
i¼1 ðY ai  Y a ÞðY pi  Y p Þ
R ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6Þ Fig. 3(b) shows a strong relationship between Na2SiO3/NaOH
Pn  2 Pn ðY pi  Yp Þ2
i¼1 ðY ai  Y a Þ i¼1 ratio and activator/fly ash ratio. For a range of Na2SiO3/NaOH of
1.5 to 5.5, a range of compressive strengths can be achieved
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn 2
(30–55 MPa) if the activator to fly ash ratio is changed between
i¼1 ðY ai  Y pi Þ 0.65 and 0.75. Fig. 3(e) depicts that for a range of Na2SiO3/NaOH
RMSE ¼ ð7Þ
n of 1.5 to 4.5, a range of compressive strengths can be achieved
Pn (30–55 MPa) if the water/solid ratio is changed between 0.3 and
i¼1 jY ai  Y pi j 0.35.
MAE ¼ ð8Þ
n
(d) Effect of NaOH molarity
Yai and Ypi are actual and predicted compressive strengths, Ya and Yp
are the mean of the actual and predicted values while n is the num- Fig. 3(c) and (d) show that for a certain NaOH molarity, different
ber of data samples. The performance indicators of the MARS model compressive strength can be achieved if the activator/fly ash and
for the training and testing data sets are shown in Table 3. The water/solid ratios are changed. Fig. 3(f) shows that even if the
training data set is having a better correlation with the actual NaOH molarity reduces, a range of compressive strengths (30 to
values. 60 MPa) can be achieved by increasing the Na2SiO3/NaOH ratio.
 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481 475

Table 1
Reported mix designs of low calcium fly ash based geopolymer concrete (kg/m3).

Reference study Fly ash Aggregates Alkaline activator Added NaOH Solid% in Heat curing Strength
water molarity (M) Na2SiO3 (MPa)
Coarse Fine NaOH Na2SiO3 SiO2 Na2O t (hrs) T(°C)
Ahmed et al. [29] 400 950 850 57 143 80 12 29.43 14.26 24 70 22.6
400 950 850 57 143 60 12 24 70 37.3
400 950 850 57 143 48 12 48 60 44.8
400 950 850 57 143 40 12 24 70 53.5
Deb et al. [30] 400 1222 658 40 100 0 14 30.7 11.4 72 20–23 25.0
400 1222 658 56 84 0 14 72 20–23 27.0
Hardjito and Rangan [31] 408 1294 554 41 103 21.3 14 29.4 14.7 24 30 32.0
408 1294 554 41 103 10.7 14 24 30 35.0
408 1294 554 51.5 103 16.5 14 24 60 36.0
408 1201 647 41 103 20.7 14 4 90 37.0
408 1201 647 41 103 26.5 16 24 60 40.0
408 1294 554 41 103 16.5 14 24 60 40.0
408 1294 554 51.5 103 16.5 14 24 60 41.0
408 1294 554 51.5 103 16.5 14 24 60 41.0
408 1294 554 51.5 103 16.5 14 24 60 42.0
408 1201 647 41 103 14.4 12 24 60 42.0
408 1201 647 41 103 17.6 14 24 60 43.0
408 1294 554 41 103 0 14 24 60 44.0
408 1201 647 41 103 7.5 10 24 60 45.0
408 1201 647 55.4 103 0 8 24 75 44.0
476 1294 554 48 120 0 8 24 60 57.0
408 1201 647 41 103 0 8 24 60 63.0
476 1294 554 48 120 0 14 24 60 68.0
408 1201 647 41 103 0 14 24 90 89.0
Galvin and Llyod [32] 408 1246 554 41 103 20 8 31.5 12.6 18 60 29.0
408 1080 554 41 103 20 8 18 60 29.0
408 1243 554 41 103 20 8 18 60 25.0
Hardjito et al. [33] 408 1232 616 48 103 0 14 29.4 14.7 24 60 28.0
408 1232 616 41 103 21.3 14 24 60 32.0
408 1232 616 55.4 103 0 8 24 60 33.0
408 1232 616 41 103 10.6 14 24 60 35.0
408 1232 616 41 103 20.7 14 29.4 14.7 24 60 45.0
408 1232 616 41 103 26.5 16 24 60 48.0
408 1232 616 41 103 14.4 12 24 60 51.0
408 1232 616 41 103 7.5 10 24 60 52.0
408 1232 616 41 103 0 8 24 60 55.0
408 1232 616 41 103 0 14 24 60 66.8
420.6 1032.0 555.7 37.6 80.1 113.6 10 24 60 33.8
Kong and Sanjayan [34] 350 1200 645 41 103 35 8 29.4 14.7 24 80 20.0
Joseph and Matthew [35] 365.2 1118.0 602.0 34.3 73.0 103.5 10 34.64 16.27 24 100 35.3
254.5 1290.0 694.7 22.8 48.5 68.7 10 24 100 36.8
309.9 1204.0 648.4 27.7 59.0 83.7 10 24 100 42.0
400.0 1265.0 540.0 42.3 105.7 24.3 16 24 100 44.0
405.0 1235.0 545.0 52.9 132.4 28.0 16 24 100 46.0
380.0 1233.0 540.0 56.5 141.3 14.6 16 24 100 49.0
400.0 1356.0 535.0 51.5 128.6 12.7 16 24 100 52.0
420.0 1125.0 750.0 40.0 100.0 0.0 16 24 100 70.5
Kubiantoro et al. [36] 350 1200 645 41 103 35 8 29.4 14.7 24 60 48.0
408 1294 554 41 103 22.5 14 24 60 36.0
Lloyd and Rangan [26] 408 1294 554 41 103 0 8 29.4 14.7 24–96 60 56.0
400 950 850 57 144 48 12 24–96 60 48.5
Nuruddin et al. [37] 400 1209 651 45.7 114.3 0 12 29.4 14.7 24–96 60–90 26.0
Nath and Sarker [38] 428.6 1177.0 623.0 68.6 102.9 28.5 14 30.7 11.4 72 20–23 28.6
Olivia and Nikraz [39] 394.3 1201.0 647.0 52.6 105.1 21.5 14 29.4 14.7 24 60 29.7
408.9 1177.0 623.0 57.2 85.9 24.5 14 24 75 35.7
480.0 1153.0 599.0 56.0 112.0 23.7 14 12 70 37.1
444.4 1177.0 623.0 44.4 111.1 18.6 14 24 60 38.7
498.5 1153.0 599.0 59.8 89.7 26.5 14 24 60 39.9
461.5 1177.0 623.0 46.2 92.3 18.6 14 24 75 42.5
462.9 1153.0 599.0 52.9 132.2 21.2 14 24 75 49.6
424.6 1177.0 623.0 36.4 91.0 16.0 14 12 70 54.9
406.0 1194.0 643.0 41.0 102.0 26.8 14 12 70 37.0
Rahman and Sarker [40] 404 1195 640 41 102 20 16 29.4 14.7 24 60 50.0
408 1294 554 41 103 22.5 16 24 60 45.0
Rangan [41] 408 1294 554 41 103 0 8 29.4 14.7 4–96 60 58.0
408 1201 647 62 93 4 14 4–96 60 32.0

(continued on next page)

476 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481

Table 1 (continued)

Reference study Fly ash Aggregates Alkaline activator Added NaOH Solid% in Heat curing Strength
water molarity (M) Na2SiO3 (MPa)
Coarse Fine NaOH Na2SiO3 SiO2 Na2O t (hrs) T(°C)

Sarker et al. [42] 408 1201 647 62 93 0 14 28.9 9.1 24 60 36.0

408 1201 647 68 103 0 14 24 60 48.0
368 1294 554 53 131 0 8 24 60 54.7
Shi et al. [43] 532.8 0 1600.8 41 102.5 0 8 29.4 14.7 24 80 56.5
Sofi et al. [44] 378 1294 554 50 124 0 12 29.4 14.7 24 30–35 34.6
Sujatha et al. [45] 408 1201 647 63 138 0 12 29.4 14.7 24 70 54.2
404 1190 640 41 102 25.5 14 24 70 37.0
Sumajouw and Rangan [46] 404 1190 640 41 102 17 14 29.4 14.7 24 60 46.0
404 1190 640 41 102 13.5 14 24 60 76.0
408 1202 647 41 103 26 16 24 60 42.0
Sumajouw et al. [23] 404 1190 640 41 102 16.5 14 29.4 14.7 24 60 66.0
428 1170 630 57 114 86 14 24 60 20.0
Vora and Dave [47] 428 1170 630 57 114 64 14 35.0 16.84 24 75 24.0
428 1170 630 49 122 43 14 24 75 28.0
428 1170 630 49 122 43 14 48 75 29.0
444 1170 630 44 111 43 14 24 75 30.0
428 1170 630 49 122 43 14 48 75 30.0
428 1170 630 57 114 43 14 24 75 30.0
428 1170 630 57 114 43 8 24 75 32.0
428 1170 630 57 114 43 10 24 75 35.0
428 1170 630 57 114 43 14 24 75 40.0
428 1170 630 57 114 43 12 24 75 43.0
408 1294 554 41 103 22.5 14 24 75 36.0
Wallah and Rangan [48] 408 1294 554 41 103 0 8 29.4 13.7 24 60 56.0
Diaz-Loya et al. [16] 494 858 691 99 99 0 14 29.4 14.7 72 60 42.1
Gunasekara et al. [49] 416 927 699 65 292 8 15 29.4 14.7 24 80 82.5
416 927 699 65 292 8 15 24 80 36.9
420 936 706 92 241 15 15 24 80 24.9
412 918 693 39 342 0 15 24 80 29.6
Wardhono, A. [50] 409 909 686 129 204 10 15 29.4 14.7 24 80 22.4

Table 2
Statistics of the inputs and the output for the MARS model.
a
Variable Minimum Maximum Average SD Skewness Kurtosis
Water/solids 0.022 0.393 0.228 0.073 0.132 1.285
Alkaline activator/fly ash 0.269 0.925 0.401 0.117 3.003 9.557
Na2SiO3/NaOH 1 8.769 2.405 0.801 5.488 42.003
Molarity of NaOH 8 16 12.589 2.444 0.936 0.514
Compressive strength (MPa) 20 89 41.936 13.338 1.008 1.411
a
Standard deviation.

4. Validation of mix designs using experimental data

Table 3
Model performance metrics. In order to assess the MARS model developed it was used to
develop mix designs for four low calcium fly ash geopolymer
R (MPa) RMSE (MPa) MAE (MPa)
concretes with nominal strengths of 30 MPa, 40 MPa, 50 MPa and
Training 0.768 8.764 6.418
55 MPa.
Testing 0.123 16.37 10.96

4.1. Example calculation – 40 MPa-Geopolymer concrete

On the other hand, while the trend shown in Fig. 3(d) is that The contour plots obtained from MARS model, shown in Fig. 3,
decreasing NaOH molarity will give increasing compressive were used to identify the values for four mix design variables in
strength, Fig. 3(f) shows that for one particular NaOH molarity, a order to design the 40 MPa-geopolymer concrete mix. The input
range of compressive strengths can be achieved by selecting a values. Derived from the model, for the water/solid ratio, activa-
Na2SiO3/NaOH ratio. This shows that effect of NaOH molarity is tor/fly ash ratio, Na2SiO3/NaOH ratio and NaOH molarity were
closely related to the other three parameters discussed. 0.25, 0.50, 3.0 and 10.0 respectively. The binder content was
It is noted that Fig. 3(a)–(d) can be used to design a geopolymer selected based on a review of the database. Most of the previously
concrete with a desired target compressive strength at 28 days. reported geopolymer concrete mixes in Table 1 used a fly ash con-
Once it is decided, Fig. 3(e) and (f) can either be used as a check tent of 400–420 kg. Taking the median value, the total fly ash con-
or to further refine the mix design. tent in the mix designs used for this study is taken as 410 kg.
 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481 477

80 80
(a) Training (b) Tesng

Predicted Compressive Strength (MPa)

Predicted compressive strength (MPa)

70 70

60 60

50 50

40 40

30 30

20 20
20 30 40 50 60 70 80 20 30 40 50 60 70 80
Actual Compressive Strength (MPa) Actual compressive strength (MPa)

Fig. 1. Performance of MARS model based on (a) training and (b) testing.

1.2 1.2
Normalised Compressive Strength

Normalised Compressive Strength

(a) Training (b) Tesng
1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 10 20 30 40 50 60 70 0 5 10 15 20 25 30
Actual Compressive Strength (MPa) Actual Compressive Strength (MPa)

Fig. 2. Performance of the model using normalized predicted compressive strength.

(a) Calculate alkaline activator content: V Flyash þ V Na2 SiO3 þ V NaOH þ V Added water þ V Sand þ V Aggregate ¼ 1

Activ ator Na2 SiO3 þ NaOH MFly ash

þ
M Na2 SiO3
þ
M NaOH
þ
M Added water
þ
M Sand
þ
M Aggregate
¼1
¼ ¼ 0:50 ð9Þ qFly ash qNa2 SiO3 qNaOH qAdded water qSand qAggregate
Fly ash Fly ash
ð12Þ
Na2 SiO3
¼ 3:0 ð10Þ V Aggregate
NaOH ¼ 0:65 ð13Þ
V Sand þ V Aggregate
After solving Eqs. (9) and (10);
Na2SiO3 = 153.8 kg and NaOH = 51.2 kg. where q is specific material density (kg/m3).
After solving Eqs. (12) and (13);
(b) Calculate required added water content: M Sand ¼ 588 kg and M Aggregate ¼ 1157.5 kg.

Water Na2 SiO3water þ NaOHwater þ Added Water 4.2. Materials used

¼ ¼ 0:25 ð11Þ
Solid Fly ash þ Na2 SiO3solid þ NaOHsolid
The geopolymer concrete was produced using 100% Class F low
Na2SiO3 NaOH Added water Total calcium fly ash, conforming to AS 3582.1 standard [60], which is
obtained from Gladstone power plant in Australia. X-ray fluores-
Solid 67.8 14.8 0 82.6 cence analysis, X-ray diffraction analysis and Malvern particle size
Water 86.0 36.4 w 122.4 + w (Mastersizer X) analysis were conducted in order to investigate the
chemical composition (Table 4), mineralogical composition
After solving Eq. (11); Added water (w) = 0.75 kg. (Table 5) and particle size distribution of fly ash (Table 5), respec-
tively. The surface area of fly ash was determined using Brunauer
(c) Calculate fine and coarse aggregate content: Emmett Teller (BET) method by N2 absorption.
Commercially available sodium hydroxide solution (10 M) and
The fine and coarse aggregate content in geopolymer concrete sodium silicate solution (Na2O = 14.7% and SiO2 = 29.4% by mass,
mix is calculated based on the absolute volume (V) method [59]. specific gravity = 1.53) were used as alkaline activator in the
That is; geopolymer production. The fine aggregate and coarse aggregate
478 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481

0.45 0.75
(a) (b) 50
0.7
0.4 30
0.65

Acvator/Fly ash rao

35

30
0.6
Water/Solid rao

0.35

35
35

40

45
0.55

55
50
0.3 40 0.5

60
0.45
40
0.25 45
0.4

0.35
0.2

35
50

45
40
45 0.3

65
55

70
60
50
0.15 0.25
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 1.5 2 2.5 3 3.5 4 4.5 5 5.5
Acvator/Fly ash rao Na2SiO3/NaOH rao
0.45
30
0.8 (c) (d)
30
30 35
0.4
0.7
35 35
Acvator/Fly ash rao

40

Water/Solid rao
35
0.6 0.35
40
40 40
0.5 40 0.3
45
40
0.4 45
0.25
45
50
0.3
50 45 0.2
55 45
50
0.2
0.15
6 7 8 9 10 11 12 13 14 15 16 6 7 8 9 10 11 12 13 14 15 16
NaOH Molarity NaOH Molarity
0.35
(e) (f)
45
50
35
30

0.3
NaOH Molarity
Water/Solid rao

40

0.25 55
45

50

60
35

0.2
40

55

65
50

60
45

70

0.15
1.5 2 2.5 3 3.5 4 4.5
Na2SiO3/NaOH rao Na2SiO3/NaOH rao

Fig. 3. Effect of input parameters on the compressive strength.

were prepared with respect to the Australian Standards, AS 1141.5 4.4. Mixing, casting and curing
[61]. River sand in uncrushed form (specific gravity = 2.5 and fine-
ness modulus = 2.8) was used as fine aggregate, and 10 mm grain The Gladstone fly ash, river sand and coarse aggregate were
size crushed granite aggregate (specific gravity = 2.65 and water mixed using a 60 L concrete mixer for 4 min. Next sodium silicate,
absorption = 0.74%) was used as coarse aggregate in concrete. sodium hydroxide and water were added and mixed continuously
Demineralized water was used throughout in the mixing. for another 8 min in order to obtain a glossy and well combined
concrete mix. The geopolymer concrete mix was poured into 100
4.3. Mix design  100  100 mm cubic Teflon moulds, and then vibrated using a
vibration table for 30 s to remove air bubbles. Finally, the concrete
Table 6 shows the mix design variables used to design four dif- moulds were kept at laboratory conditions (24 °C temperature and
ferent fly ash geopolymer concrete mixes using MARS model. Each 70% relative humidity) for 24 h and then heat cured in an oven for
geopolymer concrete mix targeted to achieve specific compressive 1 day at 80 °C temperature). After heat curing, concrete moulds
strength at 28 days. The calculated mix proportions of four differ- were removed from the oven and left until cooling to laboratory
ent concrete mixes are tabulated in Table 7. conditions before demoulding. The geopolymer concrete specimens
 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481 479

Table 4
Chemical composition of fly ash.

Fly ash Component (wt%)

SiO2 Al2O3 Fe2O3 CaO P2O5 TiO2 MgO K2O SO3 MnO Na2O LOIa
Gladstone 47.9 28.0 14.1 3.8 1.8 2.0 0.9 0.6 0.3 0.2 0.4 0.4
a
Loss on ignition (unburnt carbon content).

Table 5 60
Physical and mineralogical properties of fly ash. 7-day 28-day
Properties investigated Gladstone 50
BET Surface Area, (m2/kg) 2363

Compressive Strength (MPa)

Fineness (%) at 10 micron 43.1
at 20 micron 61.9 40
at 45 micron 82.7
Amorphous content (%) 71.8
Crystalline (%) Quartz 6.8 30

55.5
50.6
Mullite 17.9

49.0
45.4
42.1
39.5
20

28.3
were stored in laboratory conditions until being tested at 7 and 28

26.6
days. 10

4.5. Testing 0
M30 M40 M50 M55
The 7-day and 28-day compressive strengths of geopolymer Geopolymer Concrete Mix Design
concrete were tested using a MTS machine with a loading rate of
20 MPa/min in accordance with AS 1012.9 standard [62]. Scanning Fig. 4. Compressive strength development.

Electron Microscopy (SEM) connected using a backscatter electron

detector of 15 eV of energy to examine the microstructure. Con- The microstructural changes of M30, M40, M50 and M55 con-
crete specimens were cut using a diamond saw to a size of 10 crete mixes at 28 days are shown in Fig. 5. A non-uniform, hetero-
mm (diameter)  5 mm (height), and then carbon coated and geneous aluminosilicate gel matrix with a number of
mounted on the SEM sample stage with conductive, double-sided unreacted/partially reacted fly ash spheres was noted in the M30
carbon tape. microstructure at 28 days. Some unreacted/partially reacted fly
ash spheres were slightly separated from the aluminosilicate gel
paste, and were not completely embedded into the geopolymer
5. Experimental results and discussion
matrix. This indicated a weak adhesion between the fly ash spheres
and geopolymeric binder. Moreover, these unreacted fly ash
The compressive strength development of four different low
spheres perform as composites. Steveson and Sagoe-Crentsil [63]
calcium fly ash based geopolymer concrete mixes, i.e. M30, M40,
reported that the interface between these composites and geopoly-
M50 and M55, between 7 and 28 days are displayed in Fig. 4. It
mer matrix is an area of weakness, which significantly affects the
was noted that all four geopolymer mixes obtained their targeted
overall strength of geopolymer concrete.
compressive strength or close to the targeted strength at 28 days,
In addition, M30 concrete contains more micro cracks than the
the M30 and M50 mixes showed a small reduction compared to
other geopolymer mixes. These micro cracks were distributed
the desired strength. Moreover, all geopolymer concretes achieved
throughout the gel matrix. The M30 geopolymer concrete mix con-
94% of their ultimate strength in the first 7 days.

Table 6
Mix design variables.

Mix notation Target strength Mix design variables

Water/Solid Activator/Fly ash Na2SiO3/NaOH NaOH molarity
M30 30 MPa 0.40 0.45 1.75 10
M40 40 MPa 0.25 0.50 3.00 10
M50 50 MPa 0.25 0.45 3.50 10
M55 55 MPa 0.25 0.50 4.00 10

Table 7
Mix proportions (kg/m3).

Mix notation Target strength Mix proportions (kg/m3)

Fly ash Sand Aggregates Na2SiO3 NaOH Added water
M30 30 MPa 410 530.6 1044.4 117.4 67.1 79.2
M40 40 MPa 410 588.0 1157.5 153.8 51.2 0.75
M50 50 MPa 410 590.2 1162.0 143.5 41.0 11.9
M55 55 MPa 410 586.8 1154.9 164.0 41.0 2.6
480 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481

M30 M40

Uneven geopolymeric
Embedded
gel distribuon
unreacted fly ash

Micro-cracks Parally
reacted fly ash
..Micro-cracks

M50 M55

Uniformly distributed
geopolymeric gel
Well compacted
geopolymer matrix

Unreacted fly ash

Fig. 5. Microstructure deviation of four different geopolymer mixes at 28 days.

tains more water (i.e. higher water/solid ratio, including 79 kg/m3 6. Summary and conclusions
of added water) compared to the other mixes. Thus, the occurrence
of these micro cracks is attributed to the evaporation of this excess A comprehensive database was developed using the available
water during the heat curing stage. It is hypothesized that a com- published research for Class F low calcium fly ash based geopoly-
bination of the uneven gel distribution with unreacted fly ash par- mer concrete. A new machine learning approach (Multivariate
ticles coupled with the presence of the micro cracks resulted in the Adaptive Regression Spline-MARS) has been utilised to design a
lower compressive strength obtained by this mix. geopolymer concrete with a target compressive strength at 28
Compared to M30 concrete mix, all other geopolymer days. The main contribution from this study is the use of MARS
microstructures showed a decrease in the number of unreacted/ model to develop contour plots to present the relationship
partially reacted fly ash particles at 28 days, indicating an between the four key parameters that influence the compressive
additional geopolymerization and concurrent gel formation with strength of geopolymer concrete. As a demonstration of the use
time. Moreover, both M30 and M40 geopolymer concretes had of these contour plots, a detailed calculation for a 40 MPa geopoly-
micro-cracks on their surface, but with M30 concrete displaying mer concrete mix design is presented. Similarly, four mix designs
more micro-cracks and a greater crack width compared to M40 were calculated, and an experimental program was conducted.
concrete. The SEM indicated the gel had diffused through the sur- The experimental results, ranging from 30 MPa to 55 MPa, are in
face covering and coalescing the remaining partially reacted fly ash good agreement with the predicted compressive strengths from
spheres together. The gel was also observed to fill the interior the contour plots hence validating the model. As such, the pro-
voids, resulting in the formation of a semi-homogeneous, but posed contour plots together with the methodology can be used
highly compacted dense microstructure, especially at M50 and to develop mix designs for low calcium Class F fly ash based
M55 geopolymer concretes. geopolymer concrete.
The overall results showed a good correlation between the tar-
geted and achieved compressive strength of low calcium fly ash
based geopolymer concrete by following the proposed MARS mix Acknowledgements
design method. Compressive strengths ranging from 30 MPa to
55 MPa were obtained with the laboratory experiments by using The authors wish to acknowledge Cement Australia Pty Ltd. for
the proposed mix design methodology. The current model has the supply of low calcium fly ash, and the X-ray facility and Micro-
identified four key mix design parameters. This could be poten- scopy & Microanalysis facility provided by RMIT University with
tially improved by considering other parameters, such as the the scientific and technical assistance. This research was conducted
chemical, physical and mineralogical properties of source fly ash. by the Australian Research Council Industrial Transformation
 W. Lokuge et al. / Construction and Building Materials 166 (2018) 472–481 481

Research Hub for nanoscience based construction material manu- [31] D. Hardjito, B.V. Rangan, Development and properties of low-calcium fly ash-
based geopolymer concrete, in: Research Report-GC1, Curtin University of
facturing (IH150100006) and funded by the Australian
Technology, Perth, Australia, 2005, pp. 1–103.
Government. [32] B. Galvin, N. Lloyd, Fly Ash Based Geopolymer Concrete with Recycled
Concrete Aggregate, in: Proceedings of the CONCRETE 2011 Conference,
2011. The Concrete Institute of Australia.
[33] D. Hardjito et al., Fly ash-based geopolymer concrete, Aust. J. Struct. Eng. 6 (1)
References (2005) 77–86.
[34] D.L.Y. Kong, J.G. Sanjayan, Effect of elevated temperatures on geopolymer
paste, mortar and concrete, Cem. Concr. Res. 40 (2) (2010) 334–339.
[1] L. Berkelmans, H. Wang, Chinese urban residential construction to 2040. 2012: [35] B. Joseph, G. Mathew, Influence of aggregate content on the behavior of fly ash
Economic Research Department, Reserve Bank of Australia. based geopolymer concrete, Scientia Iranica 19 (5) (2012) 1188–1194.
[2] J. Peng et al., Modeling of carbon dioxide measurement on cement plants, Adv. [36] A. Kusbiantoro et al., The effect of microwave incinerated rice husk ash on the
Mater. Res. 610–613 (1) (2013) 2120–2128. compressive and bond strength of fly ash based geopolymer concrete, Constr.
[3] C. Li et al., CO2 emissions due to cement manufacture, Mater. Sci. Forum 685 Build. Mater. 36 (2012) 695–703.
(1) (2011) 181–187. [37] M.F. Nuruddin et al., Utilisation of waste material in geopolymeric concrete,
[4] C. Chen et al., Environmental impact of cement production: detail of the Proc. Inst. Civ. Eng.-Constr. Mater. 164 (6) (2011) 315–327.
different processes and cement plant variability evaluation, J. Clean. Prod. 18 [38] P. Nath, P.K. Sarker, Use of OPC to improve setting and early strength
(5) (2010) 478–485. properties of low calcium fly ash geopolymer concrete cured at room
[5] C. Meyer, The greening of the concrete industry, Cem. Concr. Compos. 31 (8) temperature, Cem. Concr. Compos. 55 (2015) 205–214.
(2009) 601–605. [39] M. Olivia, H. Nikraz, Properties of fly ash geopolymer concrete designed by
[6] L.K. Turner, F.G. Collins, Carbon dioxide equivalent (CO2-e) emissions: a Taguchi method, Mater. Des. 36 (1) (2012) 191–198.
comparison between geopolymer and OPC cement concrete, Constr. Build. [40] M. Rahman, P. Sarker, Geopolymer concrete columns under combined axial
Mater. 43 (1) (2013) 125–130. load and biaxial bending, in: Proceedings of the CONCRETE 2011 Conference,
[7] G. Habert, J.B. d’Espinose de Lacaillerie, N. Roussel, An environmental 2011, The Concrete Institute of Australia.
evaluation of geopolymer based concrete production: reviewing current [41] B.V. Rangan, Low calcium fly ash based geopolymer concrete, in: Concrete
research trends, J. Clean. Prod. 19 (11) (2011) 1229–1238. Construction Engineering Handbook, 2008, pp. 1–19.
[8] F.U.A. Shaikh, Review of mechanical properties of short fibre reinforced [42] P.K. Sarker, R. Haque, K.V. Ramgolam, Fracture behaviour of heat cured fly ash
geopolymer composites, Constr. Build. Mater. 43 (2013) 37–49. based geopolymer concrete, Mater. Des. 44 (2013) 580–586.
[9] C. Gunasekara, D.W. Law, S. Setunge, Long term permeation properties of [43] X. Shi et al., Mechanical properties and microstructure analysis of fly ash
different fly ash geopolymer concretes, Constr. Build. Mater. 124 (2016) 352– geopolymeric recycled concrete, J. Hazard. Mater. 237 (2012) 20–29.
362. [44] M. Sofi et al., Engineering properties of inorganic polymer concretes (IPCs),
[10] A. Cwirzen, P. Sztermen, K. Habermehl-Cwirzen, Effect of Baltic seawater and Cem. Concr. Res. 37 (2) (2007) 251–257.
binder type on frost durability of concrete, J. Mater. Civ. Eng. 26 (2) (2013) [45] T. Sujatha, K. Kannapiran, S. Nagan, Strength assessment of heat cured
283–287. geopolymer concrete slender column, Asian J. Civ. Eng. 13 (5) (2012)
[11] D. Law et al., Durability properties of geopolymer mortars, Eleventh 635–646.
International Conference on Durability of Building Materials and [46] M. Sumajouw, B.V. Rangan, Low-calcium fly ash-based geopolymer concrete:
Components, 11dbmc, Istanbul Technical University, 2008. Reinforced beams and columns, 2006, Curtin University of Technology, Perth,
[12] D.L.Y. Kong, J.G. Sanjayan, K. Sagoe-Crentsil, Factors affecting the performance Australia, pp. 1–121.
of metakaolin geopolymers exposed to elevated temperatures, J. Mater. Sci. 43 [47] P.R. Vora, U.V. Dave, Parametric studies on compressive strength of
(3) (2008) 824–831. geopolymer concrete, Proc. Eng. 51 (2013) 210–219.
[13] T.R. Naik, S.S. Singh, Fly ash generation and utilization-an overview. Published [48] S. Wallah, B.V. Rangan, Low-calcium fly ash-based geopolymer concrete: Long-
in the book titled ‘‘Recent Trend in Fly Ash Utilization,” Ministry of term properties, in: Res. Report-GC2, 2006, Curtin University of Technology,
Environment and Forests Management, Government of India, 1993, pp. 1–25. Perth, Australia, pp. 1–107.
[14] C. Gunasekara et al., Zeta potential, gel formation and compressive strength of [49] C. Gunasekara, S. Setunge, D.W. Law, Long-term mechanical properties of
low calcium fly ash geopolymers, Constr. Build. Mater. 95 (1) (2015) 592–599. different fly ash geopolymers, ACI Struct. J. 114 (3) (2017) 743.
[15] C. Tennakoon et al., Distribution of oxides in fly ash controls strength evolution [50] A. Wardhono, The Durability of Fly Ash Geopolymer and Alkali-Activated Slag
of geopolymers, Constr. Build. Mater. 71 (2014) 72–82. Concretes, in School of Civil, Environmental and Chemical Engineering, 2015,
[16] E.I. Diaz-Loya, E.N. Allouche, S. Vaidya, Mechanical properties of fly-ash-based RMIT University: Melbourne, Australia, pp. 1–326.
geopolymer concrete, ACI Mater. J. 108 (3) (2011) 300–306. [51] J.H. Friedman, Multivariate adaptive regression splines, Ann. Statist. (1991) 1–
[17] A. Sathonsaowaphak, P. Chindaprasirt, K. Pimraksa, Workability and strength 67.
of lignite bottom ash geopolymer mortar, J. Hazard. Mater. 168 (1) (2009) 44– [52] R.C. Deo, P. Samui, D. Kim, Estimation of monthly evaporative loss using
50. relevance vector machine, extreme learning machine and multivariate
[18] K. Sagoe-Crentsil, L. Weng, Dissolution processes, hydrolysis and condensation adaptive regression spline models, Stochastic Environ. Res. Risk Assessment
reactions during geopolymer synthesis: Part II. High Si/Al ratio systems, J. 30 (6) (2016) 1769–1784.
Mater. Sci. 42 (9) (2007) 3007–3014. [53] P. Samui, Slope stability analysis using multivariate adaptive regression spline,
[19] U. Rattanasak, P. Chindaprasirt, Influence of NaOH solution on the synthesis of Metaheuristics Water Geotech. Transp. Eng. 14 (2012) 327.
fly ash geopolymer, Miner. Eng. 22 (12) (2009) 1073–1078. [54] J. Adamowski et al., Comparison of multiple linear and nonlinear regression,
[20] A.M. Al Bakri et al., The processing, characterization, and properties of fly ash autoregressive integrated moving average, artificial neural network, and
based geopolymer concrete, Rev. Adv. Mater. Sci 30 (2012) 90–97. wavelet artificial neural network methods for urban water demand
[21] S. Songpiriyakij et al., Compressive strength and degree of reaction of biomass- forecasting in Montreal, Canada, Water Resour. Res. 48 (1) (2012).
and fly ash-based geopolymer, Constr. Build. Mater. 24 (3) (2010) 236–240. [55] I. Mansouri et al., Predicting behavior of FRP-confined concrete using neuro
[22] A.A. Adam, Strength and durability properties of alkali activated slag and fly fuzzy, neural network, multivariate adaptive regression splines and M5 model
ash-based geopolymer concrete, School of Civil, Environmental and Chemical tree techniques, Mater. Struct. 49 (10) (2016) 4319–4334.
Engineering, RMIT University, Melbourne, Australia, 2009. [56] F.K. Boadu, Rock properties and seismic attenuation: neural network analysis,
[23] D. Sumajouw et al., Fly ash-based geopolymer concrete: study of slender Pure Appl. Geophys. 149 (3) (1997) 507–524.
reinforced columns, J. Mater. Sci. 42 (9) (2007) 3124–3130. [57] P.U. Kurup, N.K. Dudani, Neural networks for profiling stress history of clays
[24] J.L. Provis, J.S.J. van Deventer, Geopolymerisation kinetics. 1. In situ energy- from PCPT data, J. Geotech. Geoenviron. Eng. 128 (7) (2002) 569–579.
dispersive X-ray diffractometry, Chem. Eng. Sci. 62 (9) (2007) 2309–2317. [58] P. Samui, B. Dixon, Application of support vector machine and relevance vector
[25] W. Ferdous et al., Geopolymer concrete-filled pultruded composite beams– machine to determine evaporative losses in reservoirs, Hydrol. Process. 26 (9)
Concrete mix design and application, Cem. Concr. Compos. 58 (2015) 1–13. (2012) 1361–1369.
[26] N. Lloyd, V. Rangan, Geopolymer concrete with fly ash, in: Proceedings of the [59] A.M. Neville, Properties of Concrete, Fourth and Final Edition ed., Standards
Second International Conference on Sustainable Construction Materials and updated to 2002, 1996, Harlow: Pearson Education Limited.
Technologies, 2010, UWM Center for By-Products Utilization. [60] AS, Supplementary cementitious materials for use with portland and blended
[27] R. Anuradha et al., Modified guidelines for geopolymer concrete mix design cement, Part 1: Fly ash, in AS (Australian Standards), 1998, Standards
using Indian standard, Asian J. Civ. Eng. (Building and Housing) 13 (3) (2012) Australia: Australia, pp. 1–13.
353–364. [61] AS, Methods for sampling and testing aggregates, Method 5: Particle density
[28] C. Montes, et al., Statistical software to improve the accuracy of geopolymer and water absorption of fine aggregate, in AS (Australian Standards), 2000,
concrete mix design and proportioning. 2013. Standards Australia: Australia, pp. 1–8.
[29] M.F. Ahmed, M.F. Nuruddin, N. Shafiq, Compressive strength and workability [62] AS, Method of testing concrete, Method 9: Determination of the compressive
characteristics of low-calcium fly ash-based self-compacting geopolymer strength of concrete specimens, in AS (Australian Standards), 1999, Standards
concrete, Int. J. Civ. Environ. Eng. 3 (2) (2011) 72–78. Australia: Australia, pp. 1–12.
[30] P.S. Deb, Durability of fly ash based Geopolymer concrete, in: School of Civil [63] M. Steveson, K. Sagoe-Crentsil, Relationships between composition, structure
and Mechanical Engineering, Department of Civil Engineering, 2013, Curtin and strength of inorganic polymers, J. Mater. Sci. 40 (16) (2005) 4247–4259.
University.