ACI MATERIALS JOURNAL TECHNICAL PAPER

Title No. 121-M48

Mixture Design for Lightweight Geopolymer Concrete

by Sathya Thukkaram and Arun Kumar Ammasi

Lightweight concrete (LWC) finds wide-ranging applications in typically have lesser density and higher porosity than
the construction industry due to its reduced dead load, good fire conventional aggregates, which reduces the overall density
resistance, and low thermal and acoustic conductivity. Light- of concrete. Lightweight pozzolanic materials, such as fly
weight geopolymer concrete (LWGC) is an emerging type of ash or slag cement, can also contribute to the lesser density
concrete that is garnering attention in the construction industry
and higher strength in LWC. The proportioning of ingredi-
for its sustainable and eco-friendly properties. LWGC is produced
ents must be carefully controlled to ensure that the resulting
using geopolymer binders instead of cement, thereby reducing the
carbon footprint associated with conventional concrete production. concrete has the desired properties, such as strength, dura-
However, the absence of standard codes for geopolymer concrete bility, and so on. Finally, the mixing process must be care-
restricts its widespread application. To address this limitation, fully controlled to ensure the concrete is thoroughly mixed,
an investigation focused on developing a new mixture design for workable, and free from voids or defects.
LWGC by modifying the existing ACI 211.2-98 provisions has been LWC typically exhibits lesser compressive strength than
carried out. In this study, crucial parameters of LWGC, such as conventional concrete, which can limit its use in high-rise
alkaline-binder ratio (A/B), molarity, silicate/hydroxide ratio, and buildings. The compressive strength of LWC can range
curing temperature, were established using machine learning tech- from 5 to 70 MPa, depending on the type of lightweight
niques. As a result, a simple and efficient method for determining aggregate and cementitious material used. The availability
the mixture proportions for LWGC has been proposed.
of lightweight aggregates and cementitious materials varies
Keywords: geopolymer; lightweight aggregates; lightweight concrete depending on the region. In some areas, these materials may
(LWC); mixture design. not be available at all, which can limit the use of LWC.7
Lightweight geopolymer concrete (LWGC) is a relatively
INTRODUCTION new type of concrete that is gaining attention in the construc-
Lightweight concrete (LWC) is a type of concrete with tion industry due to its sustainability and eco-friendliness.
a lesser density than conventional concrete. According The use of geopolymer binders in concrete can reduce
to ACI 213R-14,1 concrete with a density in the range of the carbon footprint of concrete production by up to 80%
1120 to 1920 kg/m3 is generally classified as LWC, and to compared to conventional portland cement-based concrete.8
be considered for structural application, LWC must have a In geopolymer concrete, the heat of hydration is generally
compressive strength of more than 17 MPa. LWC is typi- less compared to portland cement-based concrete due to the
cally made using lightweight aggregates, such as expanded different chemical reactions. Geopolymer concrete, which is
clay, shale, or slate, as well as lightweight pozzolanic mate- typically composed of aluminosilicate materials such as fly
rials, such as fly ash or slag cement. The lesser density of ash or slag, undergoes a geopolymerization reaction instead
LWC reduces the structure’s overall weight, which can be of the hydration reaction that occurs in portland cement.9
beneficial in seismic regions.2 LWC has excellent thermal Geopolymer concrete also exhibits excellent mechanical
and acoustic insulation properties, which make it an ideal properties, including high compressive strength, flexural
choice for exterior walls and roofs that also reduces the strength, and good durability. This makes it ideal for use in
energy consumption of buildings in addition to improving structural applications where high strength and durability
its comfort and privacy.3,4 LWC also has good fire resistance, are required.10 Geopolymer concrete can be produced using
which makes it an ideal choice for fireproofing walls and a variety of alumina- and silica-rich materials, including
floors.5 industrial wastes such as fly ash and slag.11
The current standard adopted for the mixture design of Several studies have been conducted to develop stan-
LWC is the American Concrete Institute’s ACI 211.2-98 dard mixture designs for geopolymer concrete. However,
code,6 which was formulated in 1998 and reapproved in the mixture design process for geopolymer concrete is
2004. The ACI code was adopted for this study due to its still an area of active research and development, and there
wide use in multiple countries over other national or regional is no universally accepted standard mixture design for
standards, such as British, Indian, European, or Australian geopolymer concrete. The absence of standard code provi-
standards, which hold significance only in their respective sions for geopolymer concrete can be attributed to several
regions.
ACI Materials Journal, V. 121, No. 5, September 2024.
The production of LWC involves several factors, including MS No. M-2023-311.R3, doi: 10.14359/51742040, received June 3, 2024, and
the selection of lightweight aggregate, the selection of light- reviewed under Institute publication policies. Copyright © 2024, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
weight cementitious materials, the proportioning of ingre- obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
dients, and the mixing process. Lightweight aggregates is received within four months of the paper’s print publication.

ACI Materials Journal/September 2024 91

Table 1—Chemical composition of fly ash, wt. %
SiO2 Al2O3 K2O MgO Fe2O3 P2O5 SO3 TiO2 MnO CaO Na2O
57.20 29.50 1.62 0.64 3.70 0.47 0.15 1.38 0.03 0.93 0.15

Table 2—Properties of fly ash

Specific gravity Loss on ignition, % Bulk density, kg/m3 Moisture content, % Specific surface area, m2/kg
2.24 0.29 1930 2.00 390

factors. Geopolymer concrete is a relatively new material RESEARCH SIGNIFICANCE

compared to traditional portland cement-based concrete. The primary objective of this study is to combine two
Geopolymer concrete is a highly customizable material that essential aspects: simplifying the mixture design for LWGC
can be formulated using various combinations of alumino- and using modern optimization methods to predict the
silicate materials, alkali activators, and curing conditions, factors affecting its strength. The main significance of this
which makes it challenging to establish a single set of research is to propose a new mixture design for LWGC by
standardized provisions that can accommodate all possible modifying the existing code provision ACI 211.2-98. To
geopolymer concrete formulations and applications. improve the efficiency and accuracy in identifying and opti-
Different researchers and engineers may have variations in mizing the performance indicators of geopolymer concrete,
terminology, a combination of design methodologies, and modern machine learning tools are used to reduce the waste
performance evaluation criteria. This lack of consensus and cost associated with traditional practices. In recent years,
makes developing unified code provisions for geopolymer nearly 65 to 70% of all the geopolymer-based research was
concrete difficult.12-15 conducted using Class F fly ash activated using a combi-
Therefore, the mixture design process for geopolymer nation of sodium silicate and sodium hydroxide.24 There-
concrete may vary depending on the type of aluminosili- fore, the proposed mixture design is applicable to LWGC
cate materials and alkaline activator solution used, as well containing Class F fly ash, sodium hydroxide, and sodium
as the specific application and performance requirements silicate.
of the concrete. Recently, mixture design procedures to fix
the mixture proportions for ordinary geopolymer concrete EXPERIMENTAL PROCEDURE
have been developed by adopting and modifying the avail- Materials
able conventional mixture design standards to accommodate Fly ash is a common industrial by-product composed of
the special properties of geopolymer concrete.16 Concrete finely divided particles that contain predominantly amor-
compressive strength is one of the most critical performance phous aluminosilicates and other minerals. In geopolymer
indicators, and optimizing it is a common objective in concrete, fly ash can be used as the primary source of alumi-
concrete mixture design in all standards and codes.17 There nosilicate material to form a geopolymer binder. Several
is a general proportionality between the strength, freshness, studies have shown that fly ash-based geopolymer concrete
and durability properties of concrete, although it is important exhibits similar or even superior properties compared
to note that the relationship may not be linear.18 Simi- to portland cement concrete, such as high compressive
larly, several factors can affect the strength of geopolymer strength, low permeability, and excellent resistance to acid
concrete, such as composition, alkali activator concentration, and sulfate attack.25 This study used fly ash obtained from
binder-activator ratio, curing conditions, aggregate proper- a locally available thermal power station. An X-ray fluores-
ties, additives and admixtures, and so on.19 These factors can cence (XRF) testing was conducted to identify the type of fly
interact with each other, and their influence on geopolymer ash. The XRF results revealed that the silica, alumina, and
concrete strength can vary based on specific materials, ferric oxide content exceeded 70%, with silica having the
mixture proportions, and curing conditions. Proper optimi- highest percentage (57.20%) as shown in Table 1. Conse-
zation and consideration of these factors are necessary to quently, the fly ash was classified as Class F fly ash based on
achieve the desired strength in geopolymer concrete. Due to the criteria specified in ASTM C618-19. The physical prop-
the presence of multiple factors, it is important to incorpo- erties of the fly ash are shown in Table 2.
rate the relation between them when identifying the factors The alkaline activator is a critical component in
that affect the formation of geopolymer concrete. Machine geopolymer concrete, as it initiates the reaction between the
learning techniques have been used in this study to identify aluminosilicate materials and forms a geopolymer binder.
and optimize these factors instead of traditional trial-and- The alkaline activator solution typically consists of an alkali
error methods.20 By employing machine learning algorithms metal silicate, such as sodium or potassium silicate, and an
such as random forest regressor (RFR), gradient boosting alkaline earth metal oxide or hydroxide, such as sodium or
regressor (GBR), support vector regression (SVR), and arti- potassium hydroxide. An investigation carried out on the
ficial neural networks (ANNs), hidden insights and intricate effect of different types of alkaline activators on the prop-
patterns within the data can be uncovered.21,22 This leads erties of fly ash-based geopolymer concrete found that the
to improved efficiency and accuracy in predicting concrete type and concentration of the activator solution significantly
strength, enabling more targeted material testing and experi- influenced the compressive strength and microstructure of
mentation, thereby reducing both cost and waste.23 concrete.26 The effect of different ratios of sodium/potassium

92 ACI Materials Journal/September 2024

Fig. 1—Steps involved in production of lightweight aggregate.
silicate to sodium/potassium hydroxide in the activator solu- aggregate typically involves mixing fly ash with water and
tion on the properties of fly ash-based geopolymer concrete then shaping the mixture into pellets using a pelletizer.
was studied recently, and it was concluded that increasing The pellets are then heated to a high temperature of 800 to
the silicate-to-hydroxide ratio resulted in high compressive 1200°C in a sintering bed, which causes them to expand and
strength, and low porosity, indicating improved durability.27 harden into lightweight aggregates. The resulting aggregates
In this experimental study, the alkaline activator used was have a low bulk density and high porosity, making them
a combination of NaOH and Na2SiO3, and the ratio of the suitable for use in LWC applications. Several studies have
solutions was determined by using machine learning tech- investigated the properties and applications of fly ash-based
niques, as discussed in the following sections. An alkaline- lightweight aggregate in concrete production, demonstrating
grade Na2SiO3 solution with a silicate-to-sodium ratio of their effectiveness in producing LWC with good mechanical
2.2:1 was used, while NaOH was prepared by dissolving properties, thermal insulation, and permeability properties.31
sodium hydroxide pellets in distilled water with a density Lightweight aggregates were developed specifically for this
of 2120 kg/m3, molar mass of 40 g/mol, and pH of 13. investigation using Class F fly ash by sintering process. The
The molarity of sodium hydroxide (molarity), silicate- steps involved in the production of lightweight aggregates
hydroxide ratio (silicate/hydroxide ratio), and alkaline- are shown in Fig. 1. Pelletization was done in a 3.7 m diam-
binder ratio (A/B) were determined using a machine learning eter pelletizer by feeding fly ash into the pan with water
model with the help of previously published data, and the sprayed periodically until the formation of pellets. The
final values were also experimentally verified as discussed pellets were then transported to a sintering bed, where the
in the following sections. gradual heating of fly ash aggregates began. The sintering
Manufactured sand (M-sand) passing through a 4.75 mm process of the aggregates is shown in Fig. 2.
sieve was used as fine aggregate in this study. M-sand has The particle size distribution of the fly ash-based light-
become an increasingly popular alternative to natural river weight aggregate developed by the incineration method
sand in construction applications due to its consistent is shown in Fig. 3. Aggregate passing through a 4.75 mm
particle size, low silt and clay content, and higher compres- sieve and retained on a 20 mm sieve was used for this
sive strength.28 investigation. It is clear from Fig. 3 that 70% of the light-
Lightweight aggregates are used in the production of weight aggregate developed is feasible to be used in LWC,
LWC to reduce the weight of the concrete while maintaining indicating a successful pelletization process. The phys-
its strength and durability. Lightweight aggregates can be ical properties of the lightweight aggregate developed are
natural or artificial, and they are typically characterized shown in Table 3 with reference to the most commonly
by their low bulk density and high porosity.29 In this study, used commercially available fly ash-based lightweight
Class F fly ash-based lightweight aggregates were used in the aggregate in the construction industry. The physical prop-
production of LWC.30 The production of fly ash lightweight erties of lightweight aggregate produced were similar to the

ACI Materials Journal/September 2024 93

Table 3—Properties of fly ash-based lightweight aggregate
Property Bulk density, kg/m3 Specific gravity Water absorption, % Loss on ignition, %
Fly ash-based lightweight
740 1.47 14.80 2.93
aggregate
Reference 800 1.83 16.70 2.90

Note: Reference is commercially available fly ash-based lightweight aggregate.

Fig. 2—Sintering process of fly ash aggregates.

process, the fly ash particles are heated to high temperatures,
causing the carbon residue to oxidize and leave a black color
on the interior of the aggregate. This does not necessarily
impact the performance or properties of the lightweight
aggregate.
Class F fly ash-based geopolymer concrete is usually
cured at elevated temperatures in a controlled environment.
Fly ash, an aluminosilicate material, requires heat curing to
enhance its reactivity and promote the geopolymerization
process. Elevated temperatures accelerate the dissolution
of fly ash particles and the activation of the alkali activator,
resulting in the formation of a more robust geopolymer
Fig. 3—Size distribution of fly ash-based aggregates. network. This improved reactivity leads to higher compres-
sive strength and enhanced mechanical properties. Heat
commercially available fly ash-based aggregate.32 The fully
curing also helps in drying out the fly ash geopolymer
developed lightweight aggregate by pelletization is shown
concrete more rapidly, which reduces the duration of the
in Fig. 4 along with internal and external microstructure.
drying phase and mitigates the potential for drying shrinkage
The pellets exhibited a reddish-brown color after incinera-
cracks. By applying controlled heat, moisture can be effec-
tion and a black interior. During the sintering process, the fly
tively removed from the concrete, allowing it to reach the
ash particles are heated to a high temperature in a reducing
desired moisture content and dimensional stability.33 The
atmosphere, which contains a limited amount of oxygen. In
carbon dioxide (CO2) emissions associated with the heat
this environment, iron oxide present in the fly ash can react
curing of fly ash-based geopolymer concrete are relatively
with the carbon present in the atmosphere to form metallic
lesser compared to the CO2 emissions associated with
iron. The metallic iron then deposits onto the surface of the
traditional portland cement-based concrete. However, heat
aggregate, as shown in Fig. 4(c), giving it a reddish-brown
curing still contributes to some level of CO2 emissions due
color (full-color PDF can be accessed at www.concrete.org).
to energy consumption.34 Using high-calcium materials such
This phenomenon is more common in fly ash-based light-
as ground-granulated blast-furnace slag or Class C fly ash
weight aggregates that have higher iron content. The interior
can negate the need for curing at elevated temperatures for
of a fly ash-based lightweight aggregate is often black due to
geopolymer concrete, but the heat of hydration is signifi-
the presence of unburned coal and fly ash residue from the
cantly high, which can lead to thermal cracking, reduced
sintering process as shown in Fig. 4(b). During the sintering
workability, and impact long-term performance.35

94 ACI Materials Journal/September 2024

Fig. 4—Lightweight aggregate image and microstructure: (a) fully developed lightweight aggregate; (b) external microstruc-
ture showing ferrous deposit; and (c) internal microstructure showing unreacted fly ash.
ANALYTICAL INVESTIGATION between different features. Both algorithms are ensemble
Geopolymer concrete, with its remarkable strength, methods that build multiple decision trees and combine their
exhibits impressive early-age performance, comparable to or predictions to arrive at a final result. Ensemble methods tend
even surpassing traditional concrete. As geopolymer binders to yield more accurate predictions compared to individual
replace environmentally detrimental portland cement, they decision trees, which helps in achieving better performance
significantly reduce carbon emissions and promise a greener in prediction.51
future. Geopolymer mortar and concrete lack a standard- In the pursuit of understanding the factors influencing the
ized mixture design, as they are relatively novel materials compressive strength of geopolymer mortar, an extensive
in widespread use and are considered a unique type of review of past literature was conducted. This investigation
construction material. Currently, the approach is to use a revealed six key parameters that significantly impact the
trial-and-error method by varying crucial parameters that strength of geopolymer concrete. These include the molarity
impact the geopolymerization process and strength gain, of sodium hydroxide, curing temperature of the concrete,
which is an exhausting and uneconomical process. Studies ratio of alkaline solution to binder, ratio of sodium silicate
on geopolymer mortars have already been conducted over to sodium hydroxide, the stand time after curing, and fine-
the past few decades, and there is ample data available to ness of fly ash.52 Each factor’s importance was quantified
determine the critical parameters using statistical or machine using a metric known as mean decrease impurity (MDI) as
learning methods without the need for experimental investi- shown in Fig. 5. MDI is a metric used in machine learning to
gation, thereby saving time, material, and cost. quantify the importance of a feature in a model, particularly
Machine learning can provide valuable insights and make in decision tree-based models such as random forests and
accurate predictions without relying on explicit statistical gradient boosting. The factors with higher MDI values have
assumptions with its ability to automatically learn patterns more impact on the compressive strength of geopolymer
and relationships from data. One of the key advantages of concrete.53 Upon careful consideration, it was decided to
machine learning is its flexibility and adaptability to complex focus on the first four factors for further analysis due to their
and nonlinear data. Unlike traditional statistical methods higher feature importance.
that often assume specific data distributions or relationships, Previous studies show that the maximum compressive
machine learning algorithms can uncover intricate patterns strength is achieved within the molarity range of 10 to 12,
and make predictions based on the data itself, allowing for and when the silicate/hydroxide ratio is within the range
more accurate and robust modeling.36 of 2.5 to 3. However, beyond these ranges, a reduction in
This study employs two machine learning techniques, compressive strength occurs due to precipitation. Molarity
namely GBR37 and RFR,38 to predict the compressive is a concentration unit commonly used in chemistry that
strength of geopolymer mortar. Previously published represents the number of moles of solute dissolved in 1 L
literature on 1:1 geopolymer mortar mixtures containing of solution. When the molarity of NaOH is too high, it can
Class F fly ash as the binder material was used to develop lead to over-dissolution of the aluminosilicate particles, and
the machine learning model.39-50 RFR and GBR are popular excess Al3+ and Si4+ ions are released into the solution. At
machine learning algorithms used for the prediction of certain concentrations, these ions can precipitate back into
compressive strength of concrete due to their ability to handle solid forms, reducing their availability for geopolymeriza-
complex relationships in the data, handle high-dimensional tion. The precipitation occurs as aluminosilicate hydrates,
features, and reduce overfitting. Compressive strength is which are not part of the desired geopolymer structure.
influenced by various factors, and the relationship between Additionally, higher curing temperatures significantly
these factors and the strength is often nonlinear. Both RFR contribute to increased compressive strength. It is evident
and GBR are capable of capturing nonlinear relationships, from these studies that increasing the A/B beyond 0.35
making them suitable for modeling the complex interactions leads to a decrease in compressive strength. These findings

ACI Materials Journal/September 2024 95

Fig. 5—Feature importance of geopolymer mortar parameters (MDI).

Fig. 6—Data summary.

highlight that the four main factors—namely molarity, silicate- helps to prevent overfitting, where the algorithm becomes
hydroxide ratio, A/B, and curing temperature—have a signif- too specialized to the training data and is unable to gener-
icant impact on the compressive strength of geopolymer alize to new data.54 GBR is also a type of machine learning
mortar. The prediction model is developed based on these algorithm that helps predict a numerical value. It works by
four significant factors which are responsible for the majority creating multiple decision trees, where each tree tries to
of strength development in geopolymer mortar. predict the value of the target variable. The algorithm then
A total of 405 data were collected with the molarity combines all the trees’ predictions and iteratively adjusts
ranging from 2 to 18, the silicate to hydroxide ratio ranging each tree’s prediction to improve the final prediction. The
from 1.00 to 3.00, the A/B ranging from 0.30 to 0.40, and algorithm learns from the mistakes of the previous trees
the curing temperature ranging from 30 to 90°C. These and tries to reduce the errors by giving more weight to the
data covered the essential ranges for all the factors to study instances that were poorly predicted.55 The result is a highly
their impact on the compressive strength of the geopolymer accurate prediction that improves with each iteration.
mortar. A summary of the data collected is shown in Fig. 6. The final output can be represented by the following
Sufficient amount of available data for a prediction model model:
improves performance and reduces error in the prediction
model. The collected data was curated to remove outliers final prediction = base value +
and handle missing values as noisy or irrelevant data can (learning rate × 1st residual prediction by model 1) +
harm model performance and therefore, the collected data (learning rate × 2nd residual prediction by model 2) …… +
was curated to remove outliers and handle missing values. (learning rate × n-th residual prediction by model n) (1)
RFR is a type of machine learning algorithm that helps
to predict a numerical value, such as compressive strength, The accuracy of a prediction model can be evaluated using
based on historical data and relevant attributes. RFR works multiple metrics, including R-squared (R2) and mean abso-
by creating multiple decision trees that each make a predic- lute error (MAE). Both metrics provide valuable insights
tion. The algorithm then takes the average of all the predic- into the performance of the model, but they capture different
tions to provide a more accurate final prediction. Each tree aspects of accuracy. R2 measures the proportion of variance
in the random forest is created by randomly selecting a in the dependent variable that is explained by independent
subset of the available data, and each tree is also created variables in the model. It ranges from 0 to 1, where a value
by using different features or characteristics of data. This of 1 indicates a perfect fit. A higher R2 value suggests that

96 ACI Materials Journal/September 2024

Fig. 7—Ensemble learning procedure to fix A/B, molar ratio, molarity, and curing temperature.

Fig. 8—Example of models predicting compressive strength for geopolymer binder.

a larger percentage of the variability in the dependent vari- comparison between the experimental values, the predicted
able is accounted for by the independent variables. However, values of GBR and RFR as well as the value adopted from
R2 does not indicate the magnitude of prediction errors, or the literature for the training data for the mixtures E1 to
how close the predicted values are to the actual values.56 On E13 are shown in Table 4. The 13 mixtures were selected
the other hand, MAE measures the average absolute differ- based on Fig. 7, where the procedure to fix the parameters
ence between the predicted values and the actual values. It of geopolymer mortar is shown in a simplified manner. The
provides a measure of the average magnitude of errors in experimental results were found to be 7 to 10% different
the predictions. A lower MAE value indicates better predic- from the predicted results, as clearly shown in Fig. 7. This
tion accuracy, as it signifies smaller differences between variation in results can be attributed to the use of M-sand
the predicted and actual values.57 The lower the MAE, the instead of river sand which was used in the literature data.
more accurate the predictions are. The MAE and R2 values The use of manufactured sand reduced the water-cement
of the RFR and GBR methods are as follows: MAE of GBR: ratio (W/C), which contributed to the higher strength of
0.72; MAE of RFR: 1.63; R2 value of GBR: 0.99; R2 value the geopolymer mortar. The higher angularity and surface
of RFR: 0.97 texture of the M-sand particles also contributed to the higher
Figure 7 shows the step-by-step procedure followed in compressive strength, which allowed for better bonding
the prediction of the mixture design parameters, and Fig. 8 with the cement paste.58 A maximum compressive strength
shows an example of a machine learning model predicting the of almost 60 MPa was achieved by using a molarity of 12,
compressive strength from the given inputs. When working Na2SiO3 to NaOH ratio of 2.5, an A/B of 0.35, and curing at
with machine learning models, it is essential to verify their a temperature of 90°C for one day followed by 28 days at
predictions experimentally to assess their accuracy and room temperature. These mixture parameters were chosen
reliability. To verify the model, experimental casting and for preparing the mixture design for LWGC.
testing were conducted in the laboratory to see if the model’s
predictions were accurate. This process is known as cross Mixture design procedure
validation and it involved the casting of 13 mixtures as per The mixture proportioning was done by modifying the
ASTM C109-20. The results of the prediction model and existing ACI 211.2-98 for selecting proportions for LWC.
experimental work are shown in Table 4. Figure 9 shows the The step-by-step mixture design procedure based on

ACI Materials Journal/September 2024 97

Table 4—Experimental verification for ensemble model to predict compressive strength of geopolymer
mortar
Actual 28 days
Silicate Curing GBR predicted RFR predicted compressive Difference (actual-predicted
Mixture Molarity hydroxide ratio A/B temperature, °C results, MPa results, MPa strength, MPa GBR average in %)
E1 10 1 0.30 90 29.26 30.31 32.93 11.14
E2 10 1 0.35 90 41.27 40.34 43.89 7.03
E3 10 1 0.40 90 30.20 27.36 32.27 6.72
E4 10 1 0.35 90 34.53 32.23 38.42 10.11
E5 10 1.5 0.35 90 40.12 42.86 46.72 14.13
E6 10 2 0.35 90 51.16 49.33 55.16 7.25
E7 10 2.5 0.35 90 57.42 55.56 61.41 7.99
E8 10 3 0.35 90 48.91 47.09 51.99 5.93
E9 8 2.5 0.35 90 40.43 40.31 38.89 3.96
E10 12 2.5 0.35 90 59.93 56.32 64.93 10.48
E11 14 2.5 0.35 90 51.49 49.11 54.78 6.00
E12 *
12 2.5 0.35 30 19.65 18.34 — —
E13 12 2.5 0.35 60 60.96 58.64 62.66 2.79
*
E12 sample did not form well enough to be considered for testing.
Note: E1 is experimental mixture 1; E2 is experimental mixture 2; E3 is experimental mixture 3; and so on.

Step-by-step procedure
The step-by-step mixture design procedure is given as
follows:
Step 1. Conversion of units—The conversion standards
are adopted in accordance with the metric system, such as
1 N/mm2 = 145.04 psi; 1 kg/m3 = 1.69 lb/yd3.
Step 2. Selecting water-cement ratio—The W/C for
normal LWC is selected by assuming the required compres-
sive strength and interpolating the corresponding W/C from
Fig. 11. LWGC uses an alkaline solution instead of water.
The alkaline solution consists of a combination of NaOH
and Na2SiO3 solution. Therefore, the corresponding amount
of alkaline solution content is taken instead of water for
LWGC. The quantity of water—that is, the alkaline solu-
tion content (Wasc), in kg/m3—is determined from Fig. 12 by
using the required slump value and exposure of the concrete
and the corresponding entrapped air is deducted.
Fig. 9—Comparison between machine learning, actual data, Step 3. Calculation of binder content—The binder content
and experimental result. is calculated by dividing the solution content by the W/C.
ACI 211.2-98 is shown in Fig. 10. ACI 211.2-98 uses pounds Fly ash quantity (Wb), in kg/m3 =
(lb) and inches (in.) as units for calculations. For ease of alkaline solution content ( ​W​asc​)
__________________________
understanding, pounds and inch units are converted to ​​         ​ ​ (2)
water-cement ratio (​ W/C)​
Newtons and meters as per standard conversion rules. Data,
wherever required, are assumed as per standard conven- The binder content is obtained in kg/m3 and the corre-
tions. The notations and abbreviations used are as follows: sponding volume is calculated by Eq. (3) given as follows
W/C is water-cement ratio; Ww is water content; A/B is
alkaline-binder ratio; Wasc is alkaline solution content (ASC); Volume of fly ash (Vb), in m3 =
Wb is fly ash quantity; Vb is fly ash volume; WLWA is quan- ​w​b​
__________________________
​  
    ​​ (3)
tity of lightweight aggregate; VLWA is volume of lightweight specific gravity of fly ash × 1000
aggregate; and WFA is quantity of fine aggregate.
Step 4. Calculation of lightweight aggregate content—
Figure 13 specifies the volume of coarse aggregate for the
unit volume of concrete based on the nominal maximum
size of aggregate and fineness modulus of fine aggregate.

98 ACI Materials Journal/September 2024

Fig. 10—Procedure to fix mixture proportions using ACI 211.2-98.

Fig. 11—Relationship between W/C and compressive strength of concrete for non-air-entrained concrete as per ACI 211.2-98,
Table 3.2.2.3.
It shows the relationship between fineness modulus of Quantity of lightweight aggregate (WLWA), in kg/m3 =
fine aggregate, volume of coarse aggregate, and maximum Volume from Table 3.2.2.4 (VLWA) / density of LWA (4)
aggregate. The section also outlines that the corresponding
quantity of lightweight aggregate for a unit of concrete can The lightweight aggregate is known to have higher water
be determined by multiplying the value from Fig. 13 by the absorption than normal aggregates and therefore it is to be
density of coarse aggregate. This calculation is used to deter- adjusted for surface-saturated conditions.
mine both the volume and weight of coarse aggregate.

ACI Materials Journal/September 2024 99

 Fig. 13—Relationship between fineness modulus of fine
aggregate and volume of coarse aggregate in kg/m3 as per
ACI 211.2-98, Table 3.2.2.4.
Fig. 12—Relationship between water content in kg/m3
and slump as per ACI 211.2-98, Table 3.2.2.2 for non-air-
entrained concrete.
WLWA (surface-saturated condition), in kg/m3 =
WLWA × water absorption factor (5)

Step 5. Calculation of fine aggregate content—After

determining the quantity of other ingredients, fine aggregate
content and volume are calculated by difference. Specifically,
fine aggregate content is calculated by assuming the weight
of the first batch of concrete from previous experience for
non-air-entrained concrete. Theoretical weight calculation
is also done to find the expected density of concrete before
casting to make sure it conforms to LWC specifications as
per the recommendations of ACI. The relationship between
the specific gravity of lightweight aggregate and the theo-
retical weight of air-entrained concrete is shown in Fig. 14.
For non-air-entrained concrete, the theoretical weight is Fig. 14—Relationship between specific gravity of light-
calculated from previous experience. However, in this weight aggregate and theoretical weight of air-entrained
study, theoretical weight is calculated as per Fig. 14 for both concrete in kg/m3 as per ACI 211.2-98, Table 3.2.2.5.
air-entrained and non-air-entrained concrete for the ease of
understanding. casting and testing. The requirement for LWGC is assumed
as given in Table 5.
Theoretical weight, in kg/m3 = Wb + Ww + WLWA + WFA
(6) Calculation of alkaline-binder ratio
For the specified compressive strength of 20 MPa, the
Quantity of fine aggregate (WFA) in kg/m3 = W/C as per Fig. 11 is 0.68. The W/C is specified in Fig. 11
weight of concrete as per Fig. 14 (–) (Wb + Ww + WLWA) (7) for a specified slump value of 75 to 100 mm and moderate
exposure. For non-air-entrained concrete, the quantity of
The obtained quantity for fine aggregate as per Eq. (7) alkaline solution to be taken is 201.74 kg/m3 as per Fig. 12.
is the saturated surface-dry weight of fine aggregate and The A/B = 0.68; quantity of alkaline solution (Ww) =
therefore no adjustments are made for water absorption. The 201.74 kg/m3; and deduct 2% for air entrapment from Ww, as
theoretical weight as per Eq. (6) should be in the range of per Table 3.2.2.2 in ACI 211.2 = 197.71 kg/m3.
1440 to 1840 kg/m3 to conform to LWC specifications. If
not, the process is repeated with modifications until the theo- Calculation of fly ash content
retical weight is obtained within standards. ​Ww​ ​
________________
Fly ash content (Wb) = ​ ​= 291 kg/m3
A/B
EXPERIMENTAL RESULTS AND DISCUSSIONS
The implementation of the proposed mixture design Calculation of lightweight aggregate content
is shown herein and is also verified through experimental Volume of lightweight aggregate as per Fig. 13 for 20 mm
maximum size of aggregate and 2.80 fineness modulus of

100 ACI Materials Journal/September 2024

fine aggregate is 0.70 m3; volume of lightweight aggregate Experimental verification
(VLWA) = 0.70 m3; quantity of lightweight aggregate (WLWA) = Casting was done with 150 x 150 x 150 mm cubes to
0.70 × density of lightweight aggregate = 0.70 × 740 = find the density as well as the compressive strength of the
518 kg/m3; water absorption of lightweight aggregate = LWC and LWGC as per ASTM C109-20. LWC was cured
14.80%; and adjustment for water absorption = WLWA × for 28 days with normal water curing, and LWGC was
1.148 = 518 × 1.148 = 595 kg/m3. oven-cured at 900°C for the first 24 hours immediately after
casting, and the ambient curing method was followed for the
Calculation of fine aggregate content next 28 days at room temperature. The results are tabulated
The quantity of fine aggregate is calculated by assuming in Table 6. The difference in the curing regimes is due to
the weight of concrete. The weight of concrete is assumed the different chemical reactions occurring in the LWC and
from previous experience or calculated from Fig. 14 LWGC mixtures. Normal concrete undergoes hydration
based on the air entrainment. However, for ease of under- (C-S-H) in water curing and geopolymer concrete undergoes
standing, theoretical weight is assumed from Fig. 14 for this geopolymerization to form a solid matrix. Geopolymeriza-
experiment. tion is a different process that does not involve traditional
Quantity of fine aggregate (WFA) = weight of concrete – water curing. The alkaline activator reacts with the alumi-
W(B + LWA + W); weight of concrete as per Fig. 14 for nosilicate components in the source materials and forms a
corresponding specific gravity of lightweight aggregate = three-dimensional geopolymer network. The primary prod-
1803 kg/m3; and quantity of fine aggregate = 1803 – ucts are aluminosilicate and calcium-silicate-hydrate-like
(197.71 + 291 + 595) = 719.29 kg/m3. phases. In fly ash-based geopolymers, geopolymerization
occurs at high temperatures and does not rely on the tradi-
Final mixture proportions tional hydration process (water curing).59
Weight of binder Wb = 291 kg/m3; weight of alkaline The experimental results showed that the compressive
solution Ww = 197.71 kg/m3; weight of lightweight aggre- strength of LWGC was 32% higher than that of LWC concrete
gate WLWA = 595 kg/m3; and weight of concrete WFA = and 55% higher than the target compressive strength of
719.29 kg/m3. 20 MPa. This was expected because the geopolymerization
The expected density of LWGC = 1803 kg/m3, which is process results in the formation of a more compact and dense
within the range of 1440 to 1840 kg/m3 as mentioned by structure than conventional cement-based concrete, resulting
ACI. in higher strength and stiffness. However, the density was
less than the expected density of 1803 kg/m3, which may be
attributed to the high-temperature curing of the geopolymer
concrete, which was not considered in the mixture design.
The W/C curve adopted from ACI 211.2-98 for this mixture
Table 5—Requirements for LWGC concrete design is not suitable for LWGC, as shown in the compres-
Required compressive strength 20 MPa sive strength results where the LWGC mixture outperformed
the target compressive strength. Therefore, to facilitate
Slump 75 to 100 mm
future mixture design and better understand the behavior
Exposure Moderate of LWGC, a new curve showing the relationship between
Air entrainment Non-air-entrained concrete the A/B and the compressive strength has to be proposed
Alkaline solution ratio 2.50 for designing LWGC. The relationship between W/C and
compressive strength is shown in Fig. 11. Experimental
Molarity of NaOH 12
casting was performed by adopting A/B of 0.41, 0.48, 0.57,
Solid content in NaOH 48% 0.68, and 0.82, as mentioned in Fig. 11. The compressive
Solid content in Na2SiO3 35.50% strength results of LWGC are shown in Fig. 15, along with
Binder Fly ash
the W/C to compressive strength ratio of LWC. It is clear
from the results that the compressive strength performance
Specific gravity of fly ash 2.24
of LWGC is higher than LWC. Therefore, the A/B cannot be
Fineness modulus of fine aggregate 2.80 adopted from the existing W/C to the compressive strength
Specific gravity of fine aggregate 2.72 curve shown in Fig. 11 as it would result in the production of
Density of lightweight aggregate 740 kg/m3
LWGC, which would exceed the target compressive strength
by a considerable margin. An empirical equation has been
Specific gravity of lightweight aggregate 1.47
derived to demonstrate the expected difference between the
Maximum size of aggregate 20 mm target compressive strength and actual compressive strength
Water absorption of lightweight aggregate 14.80%

Table 6—Experimental results of LWC and LWGC concrete

Property 28-day compressive strength, MPa Bulk density, kg/m3 Water absorption, %
LWC 23.54 1790 12.47
LWGC 31.07 1730 9.18

ACI Materials Journal/September 2024 101

Fig. 15—Comparison between ACI W/C to compressive strength relationship for LWC and proposed A/B to compressive
strength for LWGC.
obtained by comparing the five different mixtures of normal of all geopolymer research across the world. In the future,
LWC and LWGC, as shown in the equation as follows further studies can be conducted to develop an integrated
mixture design for geopolymer concrete, considering all
Difference in target compressive strength for LWGC = pozzolanic materials used in its development.
16.373 (A/B)–1.612 (8)
AUTHOR BIOS
The A/B notation used for LWGC as per Fig. 11 can be modi- Sathya Thukkaram is a PhD Scholar at the School of Civil Engineering,
Vellore Institute of Technology, Chennai, India. He received his bachelor’s
fied as W/C for LWC as per ACI 211.2-98. degree in civil engineering from Vel Tech High Tech Engineering College,
Chennai, India, and his master’s degree in structural engineering from
CONCLUSIONS Vellore Institute of Technology. His research interests include sustainable
construction and building materials.
A new mixture design procedure for lightweight
geopolymer concrete (LWGC) has been proposed and opti- Arun Kumar Ammasi is an Associate Professor at the School of Civil
mized for experimental use. Machine learning models were Engineering, Vellore Institute of Technology. He received his bachelor’s
degree in civil engineering from Sona College of Technology, Salem, India,
used to effectively determine the alkaline-binder ratio (A/B), and his master’s degree and PhD from Vellore Institute of Technology. His
molarity, molar ratio, and curing temperature. Machine research interests include sustainable construction and building materials.
learning offers a powerful and versatile approach to data
analysis and prediction, providing an alternative to tradi- ACKNOWLEDGMENTS
tional statistical methods. Machine learning has the ability to The authors would like to thank Vellore Institute of Technology, Chennai,
India.
learn from data and handle complex patterns, which makes
it an attractive choice for various applications, but careful
REFERENCES
validation and interpretation are necessary to ensure accu- 1. ACI Committee 213, “Guide for Structural Lightweight-Aggregate
rate and reliable results. Concrete (ACI PRC-213-14) (Reapproved 2023),” American Concrete
The ACI 211.2-98 standard for selecting proportions Institute, Farmington Hills, MI, 2014, 53 pp.
2. Nawy, E. G., Fundamentals of High-Performance Concrete, second
for lightweight concrete (LWC) was modified to suit the edition, John Wiley & Sons, Inc., Hoboken, NJ, 2000, 464 pp.
special needs of LWGC. Experimental results demon- 3. Pongsopha, P.; Sukontasukkul, P.; Zhang, H.; and Limkatanyu, S.,
strated that the compressive strength of LWGC exceeded the “Thermal and Acoustic Properties of Sustainable Structural Lightweight
Aggregate Rubberized Concrete,” Results in Engineering, V. 13, No. 3,
target compressive strength. The water-cement ratio (W/C) Mar. 2022.
versus compressive strength curve used for LWC cannot be 4. Carrillo, J.; Lizarazo, J. M.; and Bonett, R., “Effect of Lightweight
adopted for LWGC as there was a considerable difference in and Low-Strength Concrete on Seismic Performance of Thin Lightly-
Reinforced Shear Walls,” Engineering Structures, V. 93, 2015, pp. 61-69.
the compressive strength of LWC and LWGC produced by doi: 10.1016/j.engstruct.2015.03.022
adopting the same procedure. Therefore, a new A/B versus 5. Go, C. G.; Tang, J. R.; Chi, J. H.; Chen, C.-T.; and Huang, Y.-L.,
compressive strength curve is proposed to rationalize the “Fire-Resistance Property of Reinforced Lightweight Aggregate Concrete
Wall,” Construction and Building Materials, V. 30, 2012, pp. 725-733. doi:
higher compressive strength of LWGC. An empirical equa- 10.1016/j.conbuildmat.2011.12.081
tion has also been proposed to calculate the expected devi- 6. ACI Committee 211, “Standard Practice for Selecting Proportions
ation in compressive strength before initiating the mixture for Structural Lightweight Concrete (ACI 211.2-98) (Reapproved 2004),”
American Concrete Institute, Farmington Hills, MI, 1998, 20 pp.
design, which will help the user or the concrete technolo- 7. Tale Masoule, M. S.; Bahrami, N.; Karimzadeh, M.; Mohasanati, B.;
gist to analyze and achieve the targeted result. The proposed Shoaei, P.; Ameri, F.; and Ozbakkaloglu, T., “Lightweight Geopolymer
mixture design is only adaptable for fly ash-based LWGC Concrete: A Critical Review on the Feasibility, Mixture Design, Durability
Properties, and Microstructure,” Ceramics International, V. 48, No. 8,
activated using a combination of sodium silicate and sodium 2022, pp. 10347-10371. doi: 10.1016/j.ceramint.2022.01.298
hydroxide, which comprise of approximately 65 to 70%

102 ACI Materials Journal/September 2024

 8. Elmesalami, N., and Celik, K., “A Critical Review of Engineered Self-Compacted Geopolymer Concrete,” Materials Today: Proceedings,
Geopolymer Composite: A Low-Carbon Ultra-High-Performance V. 62, 2022, pp. 4168-4176. doi: 10.1016/j.matpr.2022.04.688
Concrete,” Construction and Building Materials, V. 346, Sept. 2022. doi: 28. Praveen Kumar, K.; Radhakrishna; Ramesh, P. S.; and Aravinda,
10.1016/j.conbuildmat.2022.128491 P. T., “Effect of Fines on Strength and Durability of Concrete with Manu-
9. Mohamed, R.; Abd Razak, R.; Abdullah, M. M. A. B.; Rahim, factured Sand,” Materials Today: Proceedings, V. 66, 2022, pp. 2381-2386.
S. Z. A. A; Yuan-Li, L.; Subaer; Sandu, A. V.; and Wystocki, J. J., “Heat 29. Bejan, G.; Barbua, M.; Vizitiu, R. S.; and Burlacu, A., “Lightweight
Evolution of Alkali-Activated Materials: A Review on Influence Factors,” Concrete with Waste – Review,” Procedia Manufacturing, V. 46, 2020,
Construction and Building Materials, V. 314, 2022. doi: 10.1016/j. pp. 136-143.
conbuildmat.2021.125651 30. Thukkaram, S., and Kumar, A. A., “Behaviour of Sewage Sludge-
10. Lahoti, M.; Tan, K. H.; and Yang, E. H., “A Critical Review of Based Lightweight Aggregate in Geopolymer Concrete,” Materials
Geopolymer Properties for Structural Fire-Resistance Applications,” Research Express, V. 11, No. 5, 2024, p. 055501. doi: 10.1088/2053-1591/
Construction and Building Materials, V. 221, 2019, pp. 514-526. doi: ad4198
10.1016/j.conbuildmat.2019.06.076 31. Yu, L.; Wang, S.; Qiao, Z.; Xu, L.; Wu, K.; Li, P.; and Yang, Z.,
11. Ozcelikci, E.; Kul, A.; Gunal, M. F.; Ozel, B. F.; Yildirim, G.; “Effect of Curing Time and Temperature on the Mechanical Properties of
Ashour, A.; and Sahmaran, M., “A Comprehensive Study on the Compres- Green and Ultra-High-Strength Non-Sintered Aggregate Via Autoclave
sive Strength, Durability-Related Parameters and Microstructure of Technology,” Construction and Building Materials, V. 374, 2023. doi:
Geopolymer Mortars Based on Mixed Construction and Demolition 10.1016/j.conbuildmat.2023.130874
Waste,” Journal of Cleaner Production, V. 396, Apr. 2023. 32. Zaid, O.; Alsharari, F.; Althoey, F.; Elhag, A. B.; Hadidi, H. M.; and
12. Ansari, M. A.; Shariq, M.; and Mahdi, F., “Geopolymer Concrete Abuhussain, M. A., “Assessing the Performance of Palm Oil Fuel Ash and
for Clean and Sustainable Construction – A State-of-the-Art Review on Lytag on the Development of Ultra-High-Performance Self-Compacting
the Mix Design Approaches,” Structures, V. 55, 2023, pp. 1045-1070. doi: Lightweight Concrete with Waste Tire Steel Fibers,” Journal of Building
10.1016/j.istruc.2023.06.089 Engineering, V. 76, Oct. 2023. doi: 10.1016/j.jobe.2023.107112
13. Onoue, K.; Sagawa, Y.; Atarashi, D.; and Takayama, Y., “Optimiza- 33. Gultekin, A., and Ramyar, K., “Effect of Curing Type on Micro-
tion of Mix Proportions and Manufacturing Conditions of Fly Ash-Based structure and Compressive Strength of Geopolymer Mortars,” Ceramics
Geopolymer Mortar by Parameters Design with Dynamic Characteristics,” International, V. 48, No. 11, 2022, pp. 16156-16172. doi: 10.1016/j.
Cement and Concrete Composites, V. 133, 2022. ceramint.2022.02.163
14. Gopalakrishna, B., and Dinakar, P., “The Study on Various Tempera- 34. Dong, W.; Huang, Y.; Cui, A.; and Ma, G., “Mix Design Optimi-
ture Condition of Fly Ash Based Geopolymer Mortar,” Materials Today: zation for Fly Ash-Based Geopolymer with Mechanical, Environmental,
Proceedings, V. 93, 2023, pp. 234-238. doi: 10.1016/j.matpr.2023.07.176 and Economic Objectives Using Soft Computing Technology,” Journal of
15. Ghafoor, M. T.; Fujiyama, C.; and Maekawa, K., “Mix Design Building Engineering, V. 72, 2023. doi: 10.1016/j.jobe.2023.106577
Processing for Self Compacting Geopolymer Mortar,” Journal of Advanced 35. Kamath, M.; Prashant, S.; and Kumar, M., “Micro-Characterisation
Concrete Technology, V. 19, No. 11, 2021, pp. 1133-1147. doi: 10.3151/ of Alkali Activated Paste with Fly Ash-GGBS-Metakaolin Binder System
jact.19.1133 with Ambient Setting Characteristics,” Construction and Building Mate-
16. Pavithra, P.; Srinivasula Reddy, M.; Dinakar, P.; Hanumantha rials, V. 277, 2021. doi: 10.1016/j.conbuildmat.2021.122323
Rao, B.; Satpathy, B. K.; and Mohanty, A. N., “A Mix Design Procedure 36. Xu, C.; Fu, L.; Lin, T.; Li, W.; and Ma, S., “Machine Learning in
for Geopolymer Concrete with Fly Ash,” Journal of Cleaner Production, Petrophysics: Advantages and Limitations,” Artificial Intelligence in
V. 133, 2016, pp. 117-125. doi: 10.1016/j.jclepro.2016.05.041 Geosciences, V. 3, 2022, pp. 157-161. doi: 10.1016/j.aiig.2022.11.004
17. Nunez, I.; Marani, A.; Flah, M.; and Nehdi, M. L., “Estimating 37. Khan, M. I., and Abbas, Y. M., “Robust Extreme Gradient Boosting
Compressive Strength of Modern Concrete Mixtures Using Computational Regression Model for Compressive Strength Prediction of Blast Furnace
Intelligence: A Systematic Review,” Construction and Building Materials, Slag and Fly Ash Concrete,” Materials Today. Communications, V. 35,
V. 310, 2021. doi: 10.1016/j.conbuildmat.2021.125279 2023, p. 105793 doi: 10.1016/j.mtcomm.2023.105793
18. Adessina, A.; Fraj, A. B.; and Barthélémy, J. F., “Improvement of 38. Zhang, J.; Ma, G.; Huang, Y.; sun, J.; Aslani, F.; and Nener,
the Compressive Strength of Recycled Aggregate Concretes and Relative B., “Modelling Uniaxial Compressive Strength of Lightweight Self-
Effects on Durability Properties,” Construction and Building Materials, Compacting Concrete Using Random Forest Regression,” Construc-
V. 384, 2023. doi: 10.1016/j.conbuildmat.2023.131447 tion and Building Materials, V. 210, 2019, pp. 713-719. doi: 10.1016/j.
19. Luan, C., Shi, X., Zhang, K., Utashev, N.; Yang, F.; Dai, J.; and conbuildmat.2019.03.189
Wang, Q., “A Mix Design Method of Fly Ash Geopolymer Concrete Based 39. Bezabih, T.; Kanali, C.; and Thuo, J., “Effects of Teff Straw Ash
on Factors Analysis,” Construction and Building Materials, V. 272, 2021. on the Mechanical and Microstructural Properties of Ambient Cured Fly
doi: 10.1016/j.conbuildmat.2020.121612 Ash-Based Geopolymer Mortar for Onsite Applications,” Results in Engi-
20. Dinesh, A.; Anitha Selvasofia, S. D.; Datcheen, K. S.; and Rakhesh neering, V. 18, 2023.
Varshan, D., “Machine Learning for Strength Evaluation of Concrete 40. John, S. K.; Nadir, Y.; and Girija, K., “Effect of Source Materials,
Structures – Critical Review,” Materials Today: Proceedings, 2023. doi: Additives on the Mechanical Properties and Durability of Fly Ash and
10.1016/j.matpr.2023.04.090 Fly Ash-Slag Geopolymer Mortar: A Review,” Construction and Building
21. Ahmed, A. H.; Jin, W.; and Hussein, M. A., “Artificial Intelligence Materials, V. 280, 2021. doi: 10.1016/j.conbuildmat.2021.122443
Models for Predicting Mechanical Properties of Recycled Aggregate 41. Toniolo, N., and Boccaccini, A. R., “Fly Ash-Based Geopolymers
Concrete (RAC): Critical Review,” Journal of Advanced Concrete Tech- Containing Added Silicate Waste. A Review,” Ceramics International,
nology, V. 20, No. 6, 2022, pp. 404-429. doi: 10.3151/jact.20.404 V. 43, No. 17, 2017, pp. 14545-14551. doi: 10.1016/j.ceramint.2017.07.221
22. Hu, X.; Li, B.; Mo, Y.; and Alselwi, O., “Progress in Artificial Intel- 42. Li, X.; Bai, C.; Qiao, Y.; Wang, X.; Yang, K.; and Colombo, P.,
ligence-Based Prediction of Concrete Performance,” Journal of Advanced “Preparation, Properties and Applications of Fly Ash-Based Porous
Concrete Technology, V. 19, No. 8, 2021, pp. 924-936. doi: 10.3151/ Geopolymers: A Review,” Journal of Cleaner Production, V. 359, 2022.
jact.19.924 doi: 10.1016/j.jclepro.2022.132043
23. Zhang, C.; Zhu, Z.; Liu, F.; Yang, Y.; Wan, Y.; Huo, W.; and Yang, L., 43. Klima, K. M.; Schollbach, K.; Brouwers, H. J. H.; and Yu, Q.,
“Efficient Machine Learning Method for Evaluating Compressive Strength “Thermal and Fire Resistance of Class F Fly Ash Based Geopolymers –
of Cement Stabilized Soft Soil,” Construction and Building Materials, A Review,” Construction and Building Materials, V. 323, Mar. 2022. doi:
V. 392, 2023. doi: 10.1016/j.conbuildmat.2023.131887 10.1016/j.conbuildmat.2022.126529
24. Parathi, S.; Nagarajan, P.; and Pallikkara, S. A., “Ecofriendly 44. Atabey, İ. İ.; Karahan, O.; Bilim, C.; and Atiş, C. D., “The Influence
Geopolymer Concrete: A Comprehensive Review,” Clean Technologies and of Activator Type and Quantity on the Transport Properties of Class F Fly
Environmental Policy, V. 23, No. 6, 2021, pp. 1701-1713. doi: 10.1007/ Ash Geopolymer,” Construction and Building Materials, V. 264, 2020. doi:
s10098-021-02085-0 10.1016/j.conbuildmat.2020.120268
25. Pather, B.; Ekolu, S. O.; and Quainoo, H., “Effects of Aggregate 45. Onoue, K.; Sagawa, Y.; Atarashi, D.; and Takayama, Y., “Optimiza-
Types on Acid Corrosion Attack Upon Fly-Ash Geopolymer and Portland tion of Mix Proportions and Manufacturing Conditions of Fly Ash-Based
Cement Concretes – Comparative Study,” Construction and Building Mate- Geopolymer Mortar by Parameters Design with Dynamic Characteris-
rials, V. 313, 2021. doi: 10.1016/j.conbuildmat.2021.125468 tics,” Cement and Concrete Composites, V. 133, 2022. doi: 10.1016/j.
26. Luo, Y., Jiang, Z., Wang, D., Lv, Y.; Gao, C.; and Xue, G., “Effects of cemconcomp.2022.104645
Alkaline Activators on Pore Structure and Mechanical Properties of Ultrafine 46. Jiao, Z.; Li, X.; Yu, Q.; Yao, Q.; and Hu, P., “Sulfate Resistance of
Metakaolin Geopolymers Cured at Room Temperature,” Construction and Class C/Class F Fly Ash Geopolymers,” Journal of Materials Research and
Building Materials, V. 361, 2022. doi: 10.1016/j.conbuildmat.2022.129678 Technology, V 23, 2023, pp. 1767-1780. doi: 10.1016/j.jmrt.2023.01.131
27. Nakum, A. V., and Arora, N. K., “The Impact of Alkaline Solu- 47. Liu, M.; Hu, R.; Zhang, Y.; Wang, C.; and Ma, Z., “Effect of
tion to Fly Ash Ratio with Different Molarities of Sodium Hydroxide on Ground Concrete Waste as Green Binder on the Micro-Macro Properties of

ACI Materials Journal/September 2024 103

Eco-Friendly Metakaolin-Based Geopolymer Mortar,” Journal of Building 53. Li, X.; Wang, Y.; Basu, S., Kumbier, K.; and Yu, B., “A Debiased
Engineering, V. 68, 2023. doi: 10.1016/j.jobe.2023.106191 MDI Feature Importance Measure for Random Forests,” 33rd Conference
48. Oyejobi, D. O.; Adewuyi, A. P.; Hassan, I. A.; Suleiman, I.; on Neural Information Processing Systems, 2019, pp. 1-11.
Oyebanji, Y. O.; and Yusuf, S. O., “Performance Evaluation of Fly-Ash 54. Belgiu, M., and Drăguţ, L., “Random Forest in Remote Sensing: A
Based Geopolymer Mortar,” Materials Today: Proceedings, V. 86, 2023, Review of Applications and Future Directions,” ISPRS Journal of Photo-
pp. 88-95. doi: 10.1016/j.matpr.2023.03.292 grammetry and Remote Sensing, V. 114, 2016, pp. 24-31. doi: 10.1016/j.
49. Karakaş, H.; İlkentapar, S.; Durak, U.; and Örklemez, E., “Proper- isprsjprs.2016.01.011
ties of Fly Ash-Based Lightweight-Geopolymer Mortars Containing Perlite 55. Otchere, D. A.; Ganat, T. O. A.; Nta, V.; Brantson, E. T.; and
Aggregates: Mechanical, Microstructure, and Thermal Conductivity Coef- Sharma, T., “Data Analytics and Bayesian Optimised Extreme Gradient
ficient,” Construction and Building Materials, V. 362, 2023.doi: 10.1016/j. Boosting Approach to Estimate Cut-Offs From Wireline Logs for Net
conbuildmat.2022.129717 Reservoir and Pay Classification,” Applied Soft Computing, V. 120, 2022.
50. Tenepalli, J. S., and Neeraja, D., “Properties of Class F Fly Ash Based 56. Chicco, D.; Warrens, M. J.; and Jurman, G., “The Coefficient of
Geopolymer Mortar Produced with Alkaline Water,” Journal of Building Determination R-Squared is more Informative than SMAPE, MAE, MAPE,
Engineering, V. 19, 2018, pp. 42-48. doi: 10.1016/j.jobe.2018.04.031 MSE and RMSE in Regression Analysis Evaluation,” PeerJ Computer
51. Zhang, R.; Li, Y.; Goh, A. T. C.; Zhang, W.; and Chen, Z., “Analysis Science, V. 7, 2021, pp. 1-24. doi: 10.7717/peerj-cs.623
of Ground Surface Settlement in Anisotropic Clays Using Extreme Gradient 57. Emmanuel, T.; Maupong, T.; Mpoeleng, D.; Semong, T.; Mphago, B.;
Boosting and Random Forest Regression Models,” Journal of Rock and Tabona, O., “A Survey on Missing Data in Machine Learning,” Journal
Mechanics and Geotechnical Engineering, V. 13, No. 6, 2021, pp. 1478- of Big Data, V. 8, No. 1, 2021, p. 140. doi: 10.1186/s40537-021-00516-9
1484. doi: 10.1016/j.jrmge.2021.08.001 58. Sangoju, B.; Ramesh, G.; Bharatkumar, B. H.; and Ramanjane-
52. Soutsos, M.; Boyle, A. P.; Vinai, R.; Hadjierakleous, A.; and yulu, K., “Evaluation of Durability Parameters of Concrete with Manu-
Barnett, S. J., “Factors Influencing the Compressive Strength of Fly Ash facture Sand and River Sand,” Journal of the Institution of Engi-
Based Geopolymers,” Construction and Building Materials, V. 110, 2016, neers (India): Series A, V. 98, No. 3, 2017, pp. 267-275.
pp. 355-368. doi: 10.1016/j.conbuildmat.2015.11.045 59. Khale, D., and Chaudhary, R., “Mechanism of Geopolymerization
and Factors Influencing Its Development: A Review,” Journal of Materials
Science, V. 42, No. 3, 2007, pp. 729-746. doi: 10.1007/s10853-006-0401-4

104 ACI Materials Journal/September 2024