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: Geopolymer concrete (GPC) allows a complete replacement of cement clinker with industrial wastes,
Available online 24 December 2021 contributing to a reduction in global carbon footprint. The recent evolution of GPC cured at ambient tem-
peratures, widens its suitability and applicability to concrete-based structures. To improve the mix
Keywords: design method of such a material, in this study, a predictive model for the compressive strength of fly
Nonlinear regression ash and slag blended GPC was developed using non-linear regression analysis, as a machine learning tool,
Sustainable on a database of 156 samples, aggregated from the literature. Statistical investigations were carried out
Machine learning
on the influence of parameters, identified from existing literature while developing the model. 75% of the
Blended slag
Blended fly ash
data were randomly chosen to train the model while the remaining 25% was used to test the efficacy of
GPC the proposed model. Root Mean Squared Error (RMSE) of the predictions on the training and testing data-
sets and R-squared value of the model based on training data were used as criteria to select the model.
The proposed model shows RMSE of 6.3 MPa, and 6.7 MPa, respectively, for the training and test datasets.
The statistical investigation reveals that the compressive strength is significantly affected by the com-
bined effect of the proportion of Na2SiO3-to-NaOH ratio, activator-to-binder ratio, Silica-to-Alumina
(Si-to-Al) ratio, Lime-to-Silica (Ca-to-Si) ratio, and the curing temperature. From the prediction of the
proposed model, it was found that the compressive strength increases with the increase in Ca-to-Si ratio
and curing temperature while it decreases with the increase in the activator-to-binder ratio. Further, the
model prediction indicates that the Fly ash and Slag blended GPC can reach a compressive strength of
about 70–80 MPa, for the following range of mix proportions and conditions: the ratio of Na2SiO3-to-
NaOH (1.5–2.5), activator-to-binder ratio (0.2–0.4), Si-to-Al ratio (1.4–1.8), Ca-to-Si ratio (0.4–0.6), and
curing temperature (30 °C to 60 °C).
Copyright Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the F-EIR Conference 2021
on Environment Concerns and its Remediation: Materials Science
1. Introduction Granulated Blast Furnace Slag (GGBS) with the help of alkaline
solutions such as NaOH and Na2SiO3 can produce three-
The production of cement is utilized in the construction of dimensional polymeric chains and ring structures consisting of
infrastructure which is essential to the development of society, Polysialates-siloxy geopolymers [6]. The amorphous materials pro-
however, it increases the CO2 emission into the atmosphere [1], duced during geo-polymerization contribute to the hardened
causing environmental problems such as the greenhouse effect geopolymer concrete (GPC) with a smooth surface and high rigid-
along with the reduction in fossil fuels [2–5]. Hence, there is a need ity. Fly-ash is a by-product of burned coal originated from thermal
for an environment-friendly alternate binding material that can power plants, is used as a binding material which under high tem-
replace cement. The natural pozzolanic materials or the industrial perature and alkaline environment gets polymerized, leading to
by-products which are rich in reactive silica and alumina can be the formation of amorphous and crystalline compounds, required,
utilized as cementitious materials to replace the cement. In this to attain desirable mechanical properties [7], however, for in-situ
line of view, aluminosilicate materials such as Fly ash and Ground application, providing a high amount of heat energy is not advis-
able [8]. One barrier to the widespread use of fly ash-based GPC
Abbreviations: GPC, Geopolymer Concrete; Ca-to-Si, Calcium oxide to Silica
is the necessity for high-temperature curing [9]. To mitigate the-
ratio; Si-to-Al, Silica-to-Alumina. heat demand, blending of slag which is rich in silica, alumina,
⇑ Corresponding author. and calcium is adopted. In geo-polymerization, blending of slag
E-mail address: mohdataullah.khan@mahindrauniversity.edu.in (M.A. Khan).
https://doi.org/10.1016/j.matpr.2021.12.026
2214-7853/Copyright Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the F-EIR Conference 2021 on Environment Concerns and its Remediation: Materials Science
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
and fly ash brings out the compounds such as calcium cial neural network, multilayer perceptron, radial basis function,
aluminosilicate hydrate (C-A-S-H) in addition to sodium alumi- and genetic programming.
nosilicate hydrate compound, which is produced by using Fly ash Most of the statistical analyses reported in the literature to
alone [10]. The compounds which are obtained by blending slag obtain optimal mix design were focused on low calcium fly ash-
and fly ash helps in early age strength gain and forms a dense based GPC cured at elevated temperature [27–29], and alkali-
structure which could lead to enhanced mechanical properties at activated slag GPC cured at ambient conditions [23]. In the pro-
ambient curing conditions [8]. Yazdi et al. [11] reported that the posed study, a non-linear regression analysis technique was used
compressive and flexural strengths increased up to 100 MPa and to develop a model to predict the compressive strength of fly ash
10 MPa, respectively, when the mixture contains 50% GGBS. In and slag blended GPC cured at elevated and ambient temperatures.
comparison to ordinary Portland Cement Concrete (OPCC), the In the process, an experimental database was developed and the
amorphous structure of GPC provides several remarkable features, influence of material mix proportions on the compressive strength
including resistance to acid/ sulfate attack [12,13], chloride pene- was presented. To the best of the authors’ knowledge, limited work
tration [14], and fire [15], and an ability to reach compressive has been carried out on the statistical analysis of fly ash and slag
strength greater than 100 MPa [16]. GPC is an attractive choice blended GPC to predict the compressive strength using material
for 3D printing applications owing to its viscous nature in the fresh mix parameters. The developed model can be very useful to opti-
state, contributing to enhanced bonding between the printed lay- mize the ratios of different mix proportions.
ers [17]. The utilization of fly ash and slag blended GPC in compar-
ison to conventional concrete, results in a reduction of carbon
2. Methodology
dioxide emissions in the atmosphere, consumption of energy, dis-
posal of waste, and cost of the construction.
To develop a model, in order to predict the performance of fly
The compressive strength is a parameter that indicates the stan-
ash and slag blended GPC, the following three steps were carried
dard of concrete that could be used as a structural element. Some
out viz. 1. Development of database from Scopus and web of
statistical methods, algorithms, and codes will give the empirical
science, 2. Identification of parameters affecting the performance
equations to evaluate the relationship between mechanical proper-
of GPC from literature and 3. Statistical analysis to develop a
ties, but the models for the relationship between the compressive
non-linear regression model along with validation criteria. These
strength and the material mix parameters of GPC are scarce [18].
steps are discussed in detail, in the following sub-sections.
The proposed study utilizes a non-linear regression analysis as a
statistical technique using MATLAB R2017a [19] to obtain a predic-
tive model to estimate the compressive strength. The developed 2.1. Geopolymer concrete database
model could help in deriving the standard mix design to facilitate
wide usage of GPC in construction. A systematic literature review was carried out on slag and fly
Previous studies identified that the chemical compositions and ash blended GPC from the Scopus and Web of Science (WoS) and
the physical properties of the mix, were the most influential factors a database was developed using the published literature [8,33–
affecting the compressive strength of GPC. These factors are alka- 42] and is presented in Appendix. The output parameter is the
line liquid to binder ratio, Na2SiO3-to-NaOH ratio, additional compressive strength of GPC (y) at 28 days in MPa. The motiva-
water-to-binder ratio, NaOH to fly ash ratio, Na2SiO3 to fly ash tion to do so is that the compressive strength value of the con-
ratio, sand to fly ash ratio, and workability [20-22]. On the other crete also provides the overall measure of the performance of
hand, Ukritnukun et al. [23] reported that the curing temperature, concrete in terms of durability. From the previous studies, as dis-
blast furnace slag content, and silicate modulus also have a positive cussed before, a total of eight input parameters were identified to
impact. Moreover, Asghar et al. [24] have given the optimal values influence the compressive strength of GPC. These input parame-
for molar ratios Na + K/Si ((sodium + potassium)/silica), Ca/Si (cal- ters are listed as
cium oxide/silica), and volume ratio (H2O/solid) vol to increase the
compressive strength of GPC. Songpiriyakij et al. [25] reported that, x1 ¼ Na2SiO3-to-NaOH ratio :
at a Si-to-Al ratio of 15.9, GPC with a comparatively high compres- x2 ¼ Activator-to-binder ratio :
sive strength value of 73 MPa was obtained. Puertas et al. [26] x3 ¼ Si-to-Al ratio.
investigated the variation in strength and development of fly x4 ¼ Aggregate-to-binder ratio :
ash/slag paste hydration products. They reported that the com- x5 ¼ Ca-to-Si ratio :
pressive strength of the mix with a high fly ash/slag ratio of 1.0, x6 ¼ Superplasticizer dosage :
activated with 10 M NaOH solution, and cured at 25 °C reached x7 ¼ Molarity of NaOH :
more than 50 MPa after 28 days. A statistical analysis by Rai x8 ¼ Curing temperature.
et al. [27] reported that the combined effect of curing temperature,
the molarity of NaOH, and activator-to-binder ratio notably affects where, Activator-to-binder ratio (x2 ) is the ratio of the sum of
early compressive strength, while the NaOH/Na2SiO3 ratio is not Na2SiO3 and NaOH solutions in kg/m3 to the sum of fly ash and slag
statistically significant, and the target strength can be achieved fas- content in kg/m3. The Si-to-Al ratio (x3 ) is the ratio of total silica in
ter at elevated temperatures than at ambient temperatures. Lokuge kg/m3 to the total alumina in kg/m3, in the mixture of slag and fly
et al. [28] proposed the mix design for low calcium fly ash-based ash i.e., the binder. Ca-to-Si ratio (x5 ) is the ratio of the total lime in
GPC to obtain an optimal percentage of mix proportion to get the kg/m3 to the total silica in kg/m3, in the mixture of slag and fly ash
target compressive strength using the Multi-variate Adaptive i.e., the binder. The ranges of input parameters such as Na2SiO3-to-
Regression Splines (MARS) model. Ahmed et al. [18] proposed the NaOH ratio (0.5–6.3), activator-to-binder ratio (0.16–6.39), Si-to-Al
mix design using linear, nonlinear, and multi-logistic regression ratio (1.36–2.61), Ca-to-Si ratio (0.035–1.14), curing temperature
analysis to develop predictive models for the compressive strength (20–60 °C) are presented in Table 1. The overall influence of the
of the fly ash based-GPC and reported that the nonlinear regression type of the fly ash and slag on the performance is considered in this
model has better performance than the other two models. Previous study by including the respective oxide composition as an input
studies [30–32] have proposed models for the prediction of uncon- parameter. However, the influence of the particle size distribution
fined compressive strength of soft grounds, stabilized clayey soil, and other physical properties of the binder constituents which
and granulated blast furnace slag using techniques such as artifi- were not reported in the published literature were not accounted
467
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
Table 1
Statistics for input and predictive parameters.
for, therefore, the coefficient of determination was expected to be individual parameters. A detailed explanation of correlations
less than unity in the modelling stage. The output parameter i.e., between the input parameters and the compressive strength is
the compressive strength ranges between 10.5 and 89.6 MPa provided in Appendix-A.
which corresponds to low to high strength concrete. In Table 1,
the mean value for all variables, median, which shows the variation 2.3. Non-linear regression modelling
of a specific variable, and standard deviation, used to determine
the spread of datasets compiled from the literature [8,33–42], are The non-linear regression model developed with the help of a
likewise provided. machine learning technique wherein the database was grouped
into training and testing datasets by randomly selecting the train-
2.2. Statistical assessment ing subset while the remaining subset was allocated for testing.
75% of datasets were used to train the non-linear regression model
As discussed before, the dataset comprises eight input parame- and determine the best weights and biases. The remaining 25% of
ters with output parameter compressive strength while the param- datasets were used for testing to confirm the predictability of the
eters include a Na2SiO3-to-NaOH ratio, activator-to-binder, Si-to-Al proposed model. The training dataset was employed to build the
ratio, aggregate-to-binder ratio, Ca-to-Si ratio, superplasticizer model in two stages using MATLAB R2017a [19]. In the first stage,
dosage, the molarity of NaOH, curing temperature in oC. Statistical a linear regression model was developed with the help of the fitlm
analysis was performed to investigate whether any strong correla- function in MATLAB. The function fitlm develops a linear regression
tions exist between the input parameters and the compressive model of the dependent variable, fit to the input predictors, of the
strength of fly ash and slag blended GPC with the help of the cor- form given by
rplot function in MATLAB R2017a [19]. The function corrplot gives
y ¼ a0 þ a1 Z 1 þ a2 Z 2 þ an Z n þ ð1Þ
a measure of covariance between the individual variables such that
it ranges between 1 and 1. The corrplot presented in Fig. 1, graph- where, y is the dependent variable, ai are the model coefficients,
ically showcases the frequency distribution of each of the datasets Z 1 ; Z 2 ; Z n are the predictors, and is the residual error, not
and the Pearson linear coefficient between all the parameters explained by the model. In the second stage of the modelling, a
which will help in visualizing the subsets of data and the relation polynomial regression model of order two was developed. A generic
between them. The diagonal entries show the histogram of each second-order polynomial regression model with multiple predictor
dataset while the off-diagonal entries depict the fit between variables will have the following form
468
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
y ¼ a0 þ a1 Z 1 þ a2 Z 2 þ an Z n þ a11 Z 21 þ a22 Z 22 þ ann Z 2n generalized. Note that the training dataset was not used for the
development of the model. The R-squared value obtain based on
þ a12 Z 1 Z 2 þ aij Z i Z j þ ð2Þ
the predictions of the training dataset is 0.682. From the total eight
where, Z i ; Z 2i and Z i Z j are the linear, the quadratic, and the interac- parameters considered based on the preliminary investigations,
the significant parameters influencing the compressive strength
tion terms, respectively, ai andaij are the model coefficients. In this
found after the regression are the Na2SiO3-to-NaOH ratio (x1 ), the
stage, stepwise non-linear regression is carried out using the step
activator-to-binder ratio (x2 ), Si-to-Al ratio (x3 ), Ca-to-Si ratio
function in MATLAB on the linear model of the form presented in
(x5 ), and curing temperature (x8 ). In most of the predictive mod-
Eq. (1). The step function will modify the model by removing the
elling studies [16,20,23] the molarity of NaOH as an input param-
parameters from Eq. (1) and/or by adding quadratic or interaction
eter has a significant effect on the compressive strength. In this
terms from Eq. (2), in a stepwise manner. The basis for selecting
study, the dataset of NaOH molarity has not affected the compres-
or rejecting a term in a model is its significance in predicting the
sive strength due to the usage of an identified optimal value of
response. The function step utilizes the p-value of an F-statistic to
molarity of NaOH (about 10 M) by most of the recent studies, used
test models with and without a potential term at each step. The null
to develop database, as observed from the histogram of respective
hypothesis employed is that the term to be removed/added has a
data, presented in Fig. 1. It is a convention to present a validation
zero coefficient. In case, sufficient evidence is gathered against the
plot akin to the present study to understand the performance of
null hypothesis, the term is added to the model. At the end of the
the different models based on training and test datasets (see for
last step in regression analyses of the step function, a non-linear
instance [43–45]. The relationship between the predicted com-
regression model is obtained along with the statistical significance
pressive strength and experimentally measured compressive
of model coefficients in terms of standard error, t-statistics, and
strength of fly ash and slag blended GPC for training and testing
p-value. To ensure statistical significance, in case the p-value of
datasets are displayed in Fig. 2a and 2b, respectively. In these fig-
the model coefficients is found to be greater than 0.05, the respec-
ures, the values being closer to the diagonal line depicts the accu-
tive term is removed from the model.
racy of the model. The high accuracy reported in terms of RMSE
The following assessment criteria were used to choose the
values could be graphically observed in Fig. 2a and 2b. Note that
model: the prediction of the model should have an error of smaller
a higher error in the predictions of the testing dataset indicates
percentage between the true and predicted values, and have a
the overfitting characteristic of a model. However, the present
higher R-squared value closer to one. The Root Mean Squared Error
model shows efficacy in the predictions of both training and test-
(RMSE) of the model predictions are minimized for both the train-
ing datasets (see Fig. 2a and 2b).
ing and testing dataset while choosing the model. The RMSE is
computed as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 X 2 4. Results and discussions
RMSE ¼ pffiffiffiffi i
ytrue
i ypred i ð3Þ
N
To understand the influence of mix proportions on the compres-
where, ytrue and ypred are the true and predicted values, respectively, sive strength of fly ash and slag blended GPC, the proposed model,
i i
of the compressive strength for sample data i and N is the number given by Eqn. (3) is employed and contour maps are presented. The
of samples observed. The coefficient of determination or R-squared proposed model contains five input parameters as predictors. The
value is the measure of the amount of variability in the output contour plots depicted in Figs. 3 to 5 illustrate the dependency of
parameter explained by the input parameters which is computed the compressive strength on the two predictors simultaneously
using MATLAB R2017a [19]. Note that the adjusted R-squared value for chosen values of the remaining predictors. The constant values
accounts for the number of terms in the regression model. The R- of the predictors chosen here are the respective median values of
squared value for the model predictions of training dataset could the dataset obtained as {x1 , x2 , x3 , x5 , x8 } = {2.48, 0.45, 2.03, 0.36,
be obtained from MATLAB, however, the R-squared value for the 35.9 °C}. The contour plots should have to be interpreted carefully,
model predictions of training and/or test dataset could be computed as some material mix proportions are dependent on each other,
as below however, the influences of the material mix proportions on con-
crete strength are apparent. The influence of each parameter on
P true 2
the compressive strength is discussed below.
i yi ypred
i
R2 ¼ 1 P 2 ð4Þ In the first series, the influence of oxide constituents such as Ca-
true
i yi mean ytrue
i to-Si ratio, Si-to-Al ratio with remaining variables such as Na2SiO3-
to-NaOH ratio, activator-to-binder ratio, and curing temperature
on compressive strength was discussed through contour plots.
3. Compressive strength model for fly ash and slag blended GPC The contour plot, depicted in Fig. 3a, illustrates the variation in
compressive strength with the activator-to-binder ratio and the
Minimizing the RMSE of the model predictions for the training Ca-to-Si ratio in the binder. It was observed that the compressive
and testing datasets, and maximizing the R-squared value close to strength increases from about 20 MPa to 100 MPa as the Ca-to-Si
unity, the proposed nonlinear model for the compressive strength ratio increases from 0.15 to 1. This could be attributed to the
prediction of fly ash and slag blended GPC is obtained as: increase in the formation of C-A-S-H gel associated with a higher
Ca-to-Si ratio that helps in developing the compressive strength
f c ¼ b0 þ b1 x1 þ b2 x2 þ b3 x3 þ b4 x5 þ b6 x8 þ b7 x1 x3 ð5Þ
[10]. Fig. 3a also showcases that for a constant Ca-to-Si ratio, the
The coefficients bi ’s and their statistical significance values are compressive strength marginally decreases with an increase in
given in Table 2. The standard error, t-statistics, and the p-value the activator-to-binder ratio. Fig. 3b showcases a contour plot,
for the respective model co-efficient are also presented. Note that illustrating the variation in compressive strength in relation to
the p-value for the estimated model coefficients of all the predictor the Na2SiO3-to-NaOH ratio and the Ca-to-Si ratio. It can be seen
variables is less than 0.05. The adjusted R-squared value of the that the compressive strength decreases as the Na2SiO3-to-NaOH
model, accounting for the number of predictors is 0.815. The RMSE ratio increases. For instance, at a Ca-to-Si ratio of 0.8, the compres-
of the model predictions for the training and testing datasets are sive strength decreases from about 100 MPa to 80 MPa as the Na2-
6.3 MPa and 6.7 MPa, respectively, indicating that the model is SiO3-to-NaOH ratio increases from 0.5 to 6. This could be due to the
469
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
Table 2
Results of nonlinear regression model for fly ash and slag blended GPC using MATLAB: coefficient values, standard error, t-statistics, and p-value.
Estimated Coefficients:
Estimate Standard error t-statistics p-value
(Intercept) 85.8 12.672 6.7708 1.63E-09
x1 76.472 9.9704 7.6699 2.77E-11
x2 47.858 9.7495 4.9088 4.45E-06
x3 64.296 5.8663 10.96 7.15E-18
x5 95.046 6.4415 14.755 4.83E-25
x8 0.273 0.052594 5.1908 1.43E-06
x1 x3 40.219 4.6784 8.5968 3.85E-13
Fig. 2. Validation of the proposed model for compressive strength prediction of fly ash and slag blended GPC for (a) training dataset and (b) testing dataset.
Fig. 3. Influence of input parameters on the compressive strength, Input parameters: (a) activator-to-binder ratio and Ca-to-Si (b) Na2 SiO3-to-NaOH ratio and Ca-to-Si ratio;
(c) Ca-to-Si ratio and curing temperature. (d) activator-to-binder ratio and Si-to-Al ratio; (e) Na2SiO3-to-NaOH ratio and Si-to-Al ratio; (f) Si-to-Al ratio and curing
temperature.
prevention of water evaporation delaying setting of GPC structure Fig. 3d also shows that compressive strength decreases from about
in presence of excess Na2SiO3 leading to diminishing of compres- 85 to 40 MPa as the Si-to-Al ratio increases from 1.4 to 2.6 at an
sive strength, although, Na2SiO3 works as a charge balancing ion activator-to-binder ratio of about 0.2. This could be attributed to
and is essential for the development of GPC [43,46]. The contour the formation of monomers in the GPC which majorly depends
plot in Fig. 3c shows the variation in compressive strength, as a upon the ratio of Si-to-Al present in the source material. Previous
function of the Ca-to-Si ratio and the curing temperature. It was studies have indicated that the total Si-to-Al ratio should be in
observed that, irrespective of the curing temperature, the compres- the range of 1.6–2.3, for the formation of stable geopolymer gel
sive strength increases from about 30 MPa to 100 MPa as the Ca- [47,48]. The contour plot in Fig. 3e depicts the variation in com-
to-Si ratio increases from about 0.15 to 1.0. The contour plot in pressive strength in relation to the Na2SiO3-to-NaOH ratio and
Fig. 3d demonstrates the variation in compressive strength in the Si-to-Al ratio. It was observed that at a lower Si-to-Al ratio of
response to the change in the activator-to-binder ratio and the about 1.5, the compressive strength increases with an increase in
Si-to-Al ratio. It can be observed that the compressive strength the Na2SiO3-to-NaOH ratio. Conversely, it was observed that at a
marginally increases as the activator-to-binder ratio decreases. relatively higher Si-to-Al ratio of about 2.5, the compressive
470
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
Fig. 5. Influence of input parameters on the compressive strength, Input parameters: (a) Na2SiO3-to-NaOH ratio and activator-to-binder ratio, (b) activator-to-binder ratio and
curing temperature, (c) Na2SiO3-to-NaOH ratio and curing temperature.
471
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
The chemical composition of the binder in terms of the Lime-to- (b). Activator-to-binder:
Silica (Ca-to-Si) ratio and the Silica-to-Alumina (Si-to-Al) ratio
has a significant influence on the performance of fly ash and The Activator-to-binder ratio of the fly ash and slag blended
slag blended GPC. GPC was in the range of 0.16–6.39 with a mean value of 0.45, the
The compressive strength was found to increase with the median of 0.45, the standard deviation of 0.10 as presented in
increase in Ca-to-Si ratio and curing temperature and it Table 1. The variation between compressive strength and
decreases with the increase in activator-to-binder ratio while Activator-to-binder, as well as the histogram analysis, is shown
a non-linear dependence was seen with the Si-to-Al ratio and in Fig. 1. As can be seen from the figure a very poor relationship
Na2SiO3-to-NaOH ratio. existed between compressive strength and the Activator-to-
A significant improvement in compressive strength greater than binder. However, confounding effects of other variables are
about 200% was noted with an increase in the Ca-to-Si ratio expected to show the significance of the activator-to-binder ratio
from about 0.2 to 1, irrespective of the other parameters. in the prediction of compressive strength. These confounding
The compressive strength in the range of about 80 MPa could be effects could be observed from the significance of the activator-
obtained by providing oxide compositions such as Ca-to-Si ratio to-binder ratio in the model.
of 0.5 and Si-to-Al ratio of 1.4 while maintaining the median
values of other parameters, i.e., the activator-to-binder ratio of (c). Si-to-Al ratio:
0.45, the Na2SiO3-to-NaOH ratio of 2.5, and the curing temper-
ature of 30 °C. The Si-to-Al ratio of the fly ash and slag blended GPC was in the
range of 1.36–2.61 with a mean value of 2.03, the median of 2.02,
The proposed model could be used to develop a mix design for the standard deviation of 0.23, as presented in Table 1. The rela-
fly ash and slag blended GPC which can be cured at ambient tionship between compressive strength and Si-to-Al ratio with His-
temperatures. togram of fly ash and slag blended GPC mixtures is presented in
Fig. 1. As it can be seen from the figure that the compressive
strength is increasing with an increase in the Si-to-Al ratio.
CRediT authorship contribution statement
(d). Aggregate-to-binder ratio:
Kailash Kumar Singaram: Validation, Formal analysis, Investi-
gation, Resources, Data curation, Writing – original draft. Mohd
The aggregate-to-binder ratio was in the range of 3.85–5.18.
Ataullah Khan: Conceptualization, Methodology, Software, Visual-
The statistical analysis of the dataset shows that the mean value
ization, Writing – review & editing. V. Talakokula: Project admin-
of the aggregate-to-binder ratio content was 4.54, the median
istration, Funding acquisition.
value of 4.63, and the standard deviation was 0.4095 were shown
in Table 1. From Fig. 1, a poor relationship is observed between
Declaration of Competing Interest compressive strength and the Aggregate-to-binder ratio.
The authors declare that they have no known competing finan- (e). Ca-to-Si ratio:
cial interests or personal relationships that could have appeared
to influence the work reported in this paper. The ratio of Ca-to-Si ratio was in the range of 0.035–1.14, with a
mean value of 0.36. The median and standard deviation were
0.3107 and 0.21, respectively, as shown in Table 1. The relationship
Acknowledgments
between compressive strength and Ca-to-Si ratio along with
respective histograms for the fly ash and slag blended GPC mixture
We acknowledge the research grant received for the project
is shown in Fig. 1. The figure illustrates that the compressive
titled ‘‘High-Performance Multi-functional Geopolymer Compos-
strength increases with the increase in the Ca-to-Si ratio.
ites” from Mahindra University, Ecole Centrale School of
Engineering.
(f). Superplasticizer dosage:
Appendix A. Correlation of input parameters to strength The superplasticizer dosage was varying in the range of 0–
180 kg/m3, with a mean value of 87.05, the median value of 144,
As discussed before, the correlation between the input parame- and the standard deviation of 80.83 as depicted in Table 1. The
ters and strength is performed with the help of the corrplot func- relationship between compressive strength and superplasticizer
tion in MATLAB. The diagonal entries in Fig. 1, present the dosage along with their respective histograms is shown in Fig. 1.
histograms of the input and the output variables based on the col- It can be seen that the compressive strength increases with the
lated dataset. A histogram typically helps in visualizing the distri- increase in superplasticizer dosage.
bution of data in terms of spread, skewness, and symmetry. The
correlations between the input parameters and the compressive (g). Molarity of NaOH:
strength, as depicted in Fig. 1, are discussed below:
The Molarity of NaOH was varying in the range of 4–16 M, with
a mean value of 9.4 M, the median value of 8 M, and the standard
(a). Na2SiO3-to-NaOH ratio:
deviation of 2.94 M as depicted in Table 1. Variation between com-
pressive strength and molarity of NaOH along with their respective
The ratio of Na2SiO3-to-NaOH was in the range of 0.5–6.3, with
histograms is illustrated in Fig. 1. As it can be seen that a very poor
a mean of 2.48. The median and standard deviation were 2.5 and
relationship existed between compressive strength and the molar-
1.29 respectively as reported in Table 1. The relationship between
ity of NaOH. Although the molarity of NaOH is an important
compressive strength and Na2SiO3-to-NaOH with respective his-
parameter influencing the performance of GPC, however, most of
tograms for slag and fly ash blended GPC mixtures is shown in
the recent studies have used an optimal value of 10 M which was
Fig. 1. As can be seen from this figure that the compressive strength
depicted in the histogram of the dataset.
is increasing with an increase in the Na2SiO3-to-NaOH ratio.
472
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
The range of the curing temperature was varying from 20 to 60 The compressive strength of the fly ash and slag blended GPC
°C, with a mean value of 35.9 °C and the median value of 30 °C, and was in the range of 10.5–89.6 MPa, with a mean value of
the standard deviation of 16.05 °C as depicted in Table 1. Fig. 1 42.71 MPa and the median value of 42.78 MPa, and the standard
illustrates that the compressive strength increases with the deviation of 15.32 MPa as depicted in Table 1.
increase in curing temperature.
Appendix B. Mix design database for fly ash and slag blended GPC
Fly Ash Fly Ash GGBS GGBS Composition Aggregates Alkaline SP NaOH Temp. Comp. Ref.
(kg/m3) Composition (kg/m3) (% weight) (kg/m3) Activators dosage Molarity (°C) strength
(% weight) (kg/m3) (kg/m3) (MPa)
SiO2 Al2O3 CaO SiO2 Al2O3 CaO Fine Coarse NaOH Na2SiO3
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 30 33.83 [33]
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 30 36.19
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 30 31.11
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 30 25.71
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 54 135.0 168 8 30 36.69
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 60 150.0 168 8 30 38.16
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 66 165.0 168 8 30 26.17
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 72 180.0 168 8 30 25.11
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 57.9 144.6 180 8 30 33.81
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 64.3 160.7 180 8 30 38.96
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 70.7 176.8 180 8 30 25.71
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 77.1 192.9 180 8 30 20.76
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 30 42.32
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 30 47.92
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 30 46.87
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 30 43.38
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 54 135.0 168 8 30 46.08
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 60 150.0 168 8 30 51.22
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 66 165.0 168 8 30 44.67
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 72 180.0 168 8 30 46.57
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 57.9 144.6 180 8 30 45.88
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 64.3 160.7 180 8 30 50.69
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 70.7 176.8 180 8 30 41.85
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 77.1 192.9 180 8 30 45.68 [33]
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 30 55.37
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 30 59.79
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 30 51.61
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 30 46.68
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 54 135.0 168 8 30 56.86
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 60 150.0 168 8 30 60.38
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 66 165.0 168 8 30 46.48
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 72 180.0 168 8 30 47.57
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 57.9 144.6 180 8 30 48.91
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 64.3 160.7 180 8 30 58.53
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 70.7 176.8 180 8 30 48.45
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 77.1 192.9 180 8 30 47.64
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 60 41.53
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 60 42.56
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 60 33.94
252 60.11 26.53 4 108 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 60 28.58
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 54 135.0 168 8 60 40.41
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 60 150.0 168 8 60 43.5
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 66 165.0 168 8 60 31.24
294 60.11 26.53 4 126 34.06 20 32.6 810.6 966 72 180.0 168 8 60 26.89
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 57.9 144.6 180 8 60 38.63
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 64.3 160.7 180 8 60 44.49
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 70.7 176.8 180 8 60 35.2
473
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
Appendix B (continued)
Fly Ash Fly Ash GGBS GGBS Composition Aggregates Alkaline SP NaOH Temp. Comp. Ref.
(kg/m3) Composition (kg/m3) (% weight) (kg/m3) Activators dosage Molarity (°C) strength
(% weight) (kg/m3) (kg/m3) (MPa)
SiO2 Al2O3 CaO SiO2 Al2O3 CaO Fine Coarse NaOH Na2SiO3
315 60.11 26.53 4 135 34.06 20 32.6 760.5 972 77.1 192.9 180 8 60 29.46
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 60 55.57 [33]
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 60 57.37
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 60 53.39
216 60.11 26.53 4 144 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 60 51.22
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 54 135.0 168 8 60 58.8
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 60 150.0 168 8 60 62.29
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 66 165.0 168 8 60 58.93
252 60.11 26.53 4 168 34.06 20 32.6 810.6 966 72 180.0 168 8 60 56.64
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 57.9 144.6 180 8 60 56.56
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 64.3 160.7 180 8 60 61.3
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 70.7 176.8 180 8 60 54.38
270 60.11 26.53 4 180 34.06 20 32.6 760.5 972 77.1 192.9 180 8 60 49.83
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 46.3 115.7 144 8 60 61.96
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 51.4 128.6 144 8 60 65.26
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 56.6 141.4 144 8 60 53.39
180 60.11 26.53 4 180 34.06 20 32.6 774 1090.8 61.7 154.3 144 8 60 52.4
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 54 135.0 168 8 60 57.54
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 60 150.0 168 8 60 64.28
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 66 165.0 168 8 60 50.64
210 60.11 26.53 4 210 34.06 20 32.6 810.6 966 72 180.0 168 8 60 49.44
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 57.9 144.6 180 8 60 54.97
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 64.3 160.7 180 8 60 61.9
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 70.7 176.8 180 8 60 54.38
225 60.11 26.53 4 225 34.06 20 32.6 760.5 972 77.1 192.9 180 8 60 53.79
293 46 33 2.6 88 21 17 56.1 760 1005 143.3 71.7 6 20 37.40 [34]
293 46 33 2.6 88 21 17 56.1 760 1005 107.5 107.5 6 20 26.60
293 46 33 2.6 88 21 17 56.1 760 1005 143.3 71.7 4 20 29.90
293 46 33 2.6 88 21 17 56.1 760 1005 107.5 107.5 4 20 30.70
344 46 33 2.6 34 21 17 56.1 760 1005 143.3 71.7 4 20 15.50
316 46 33 2.6 63 21 17 56.1 760 1005 143.3 71.7 4 20 23.00
293 46 33 2.6 88 21 17 56.1 760 1005 143.3 71.7 4 20 30.60
253 46 33 2.6 108 21 17 56.1 760 1005 143.3 71.7 4 20 39.01
272 46 33 2.6 108 21 17 56.1 760 1005 143.3 71.7 4 20 27.80
253 46 33 2.6 126 21 17 56.1 760 1005 143.3 71.7 4 20 28.00
237 63.53 27.4 1.26 158 34.26 11.32 38.34 547 1277 52 129 7.9 8 32 28.36 [35]
237 63.53 27.4 1.26 158 34.26 11.32 38.34 547 1277 52 129 7.9 4 32 34.84
237 63.53 27.4 1.26 158 34.26 11.32 38.34 547 1277 52 129 7.9 4 32 37.24
237 63.53 27.4 1.26 158 34.26 11.32 38.34 547 1277 52 129 7.9 4 32 33.2
400 53.71 27.2 1.9 0 29.96 12.25 45.45 651 1209 45.7 114.3 14 23 25.6
360 53.71 27.2 1.9 40 29.96 12.25 45.45 651 1209 45.7 114.3 14 23 38.3
340 53.71 27.2 1.9 60 29.96 12.25 45.45 651 1209 45.7 114.3 14 23 46.6
400 53.71 27.2 1.9 0 29.96 12.25 45.45 655.9 1218.1 40 100 6 14 23 32.5
360 53.71 27.2 1.9 40 29.96 12.25 45.45 655.9 1218.1 40 100 6 14 23 33.3
200 52.6 34.1 1.32 200 33.8 14.8 38.8 716 1074 9 56 31.7 [37]
200 52.6 34.1 1.32 200 33.8 14.8 38.8 712 1068 12 74 49.3
200 52.6 34.1 1.32 200 33.8 14.8 38.8 708 1062 15 93 58.2
200 52.6 34.1 1.32 200 33.8 14.8 38.8 712 1068 19 62 44.4
200 52.6 34.1 1.32 200 33.8 14.8 38.8 708 1062 15 93 58.2
200 52.6 34.1 1.32 200 33.8 14.8 38.8 712 1068 12 74 49.3 [37]
100 52.6 34.1 1.32 300 33.8 14.8 38.8 712 1068 12 74 52.1
0 52.6 34.1 1.32 400 33.8 14.8 38.8 712 1068 12 74 63.1
200 52.6 34.1 1.32 200 33.8 14.8 38.8 716 1074 15 93 65.7
200 52.6 34.1 1.32 200 33.8 14.8 38.8 708 1062 15 93 58.2
200 52.6 34.1 1.32 200 33.8 14.8 38.8 700 1050 15 93 42.3
360 53.71 27.2 1.9 40 29.96 12.25 45.45 651 1209 45.7 114.3 14 22 40 [38]
320 53.71 27.2 1.9 80 29.96 12.25 45.45 651 1209 45.7 114.3 14 22 47
360 53.71 27.2 1.9 40 29.96 12.25 45.45 651 1209 64 96 14 22 43
474
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
Appendix B (continued)
Fly Ash Fly Ash GGBS GGBS Composition Aggregates Alkaline SP NaOH Temp. Comp. Ref.
(kg/m3) Composition (kg/m3) (% weight) (kg/m3) Activators dosage Molarity (°C) strength
(% weight) (kg/m3) (kg/m3) (MPa)
SiO2 Al2O3 CaO SiO2 Al2O3 CaO Fine Coarse NaOH Na2SiO3
320 53.71 27.2 1.9 80 29.96 12.25 45.45 651 1209 64 96 14 22 54
400 53.71 27.2 1.9 0 29.96 12.25 45.45 658 1222 40 100 6 14 22 25
360 53.71 27.2 1.9 40 29.96 12.25 45.45 655 1216 40 100 6 14 22 27
320 53.71 27.2 1.9 80 29.96 12.25 45.45 655 1216 40 100 6 14 22 35
400 53.71 27.2 1.9 0 29.96 12.25 45.45 658 1222 56 84 6 14 22 27
360 53.71 27.2 1.9 40 29.96 12.25 45.45 655 1216 56 84 6 14 22 27
320 53.71 27.2 1.9 80 29.96 12.25 45.45 655 1216 56 84 6 14 22 45
360 53.24 26.42 3.65 40 36.77 13.56 37.6 644 1197 53 107 4 10 22 21.9 [8]
340 53.24 26.42 3.65 60 36.77 13.56 37.6 646 1200 53 107 4 10 22 28
320 53.24 26.42 3.65 80 36.77 13.56 37.6 648 1203 53 107 4 10 22 41.5
300 53.24 26.42 3.65 100 36.77 13.56 37.6 650 1207 53 107 4 10 22 46
280 53.24 26.42 3.65 120 36.77 13.56 37.6 652 1210 53 107 4 10 22 56.5
340 53.24 26.42 3.65 60 36.77 13.56 37.6 646 1200 53 107 4 12 22 35
320 53.24 26.42 3.65 80 36.77 13.56 37.6 648 1203 53 107 4 12 22 45
300 53.24 26.42 3.65 100 36.77 13.56 37.6 658 1222 53 107 4 12 22 57 [8]
340 53.24 26.42 3.65 60 36.77 13.56 37.6 651 1209 47 93 4 10 22 30
320 53.24 26.42 3.65 80 36.77 13.56 37.6 661 1227 47 93 4 10 22 40
300 53.24 26.42 3.65 100 36.77 13.56 37.6 671 1246 47 93 4 10 22 48
340 53.24 26.42 3.65 60 36.77 13.56 37.6 637 1184 64 96 4 10 22 28.5
300 53.24 26.42 3.65 100 36.77 13.56 37.6 659 1223 46 114 4 10 22 45
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 8 10.5 [39]
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 10 13
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 12 19
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 14 22
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 16 24
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 108 54 4.05 14 22.4
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 64.8 97.2 4.05 14 27
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 54 108 4.05 14 33
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 46.28 115.72 4.05 14 30
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 40.5 121.5 4.05 14 18.2
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 8 22.2
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 10 23.2
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 12 24
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 14 24.8
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 81 81 4.05 16 21.8
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 108 54 4.05 14 23.8
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 64.8 97.2 4.05 14 33.8
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 54 108 4.05 14 34.5
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 46.28 115.72 4.05 14 35
303.75 45.8 21.4 13.7 101.25 34.52 20.66 32.43 683 1269 40.5 121.5 4.05 14 31.7
400 37.73 14.42 37.34 810 990 57.1 143 8 12 25 89.6 [40]
400 37.73 14.42 37.34 810 990 57.1 143 12 12 25 89.2
400 37.73 14.42 37.34 810 990 57.1 143 16 12 25 88.7
400 37.73 14.42 37.34 810 990 57.1 143 20 12 25 86.11
400 37.73 14.42 37.34 810 990 57.1 143 24 12 25 84.1
204.5 65.6 28 1 204.5 30.61 16.24 34.48 554 1293 41 102 10 53.5 [41]
102 65.6 28 1 307 30.61 16.24 34.48 554 1293 41 102 10 55.5
0 65.6 28 1 409 30.61 16.24 34.48 554 1293 41 102 10 58.6
400 31.63 13.42 36.35 740 1110 12.9 82.5 22 34.6 [42]
400 31.63 13.42 36.35 785 1085 12.9 82.5 22 53.6
400 31.63 13.42 36.35 790 1065 12.9 82.5 22 66.7
475
K.K. Singaram, M.A. Khan and V. Talakokula Materials Today: Proceedings 61 (2022) 466–476
476