Construction and Building Materials 362 (2023) 129722

Contents lists available at ScienceDirect

Construction and Building Materials

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

Optimization of fresh and mechanical properties of sustainable concrete

composite containing ARGF and fly ash: An application of response
surface methodology
Syyed Adnan Raheel Shah a, *, Nabil Ben Kahla b, c, Miniar Atig c, d, Muhammad Kashif Anwar a,
Marc Azab e, Atif Mahmood f
a
Department of Civil Engineering, Pakistan Institute of Engineering & Technology, Khanewal Road, Multan 66000, Pakistan
b
Department of Civil Engineering, College of Engineering, King Khalid University, Abha 61421, KSA
c
Laboratory of Systems and Applied Mechanics, Tunisia Polytechnic School, University of Carthage, La Marsa, Tunis 2078, Tunisia
d
Department of Civil Engineering, The Higher National Engineering School of Tunis, University of Tunis, Tunis, Tunisia
e
College of Engineering and Technology, American University of the Middle East, Egaila 54200, Kuwait
f
Department of Industrial and System Science Engineering, Binghamton University State University of New York, NY, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Globally, significant attempts have been made to properly utilize waste materials and various by-products as
Sustainability supplementary cementitious materials (SCMs) in the production of sustainable concrete. Fly ash and alkali
Eco-Friendly wastes resistance Glass Fiber (ARGF) are two ingredients that can help to improve the overall performance of concrete
Fly Ash
and mortars. The main objective of this study is to use Response Surface Methodology (RSM) to predict and
ARGF
Mechanical properties
optimize the fresh and mechanical properties of modified concrete made in combination with Fly Ash as a cement
Sustainable concrete alternative and ARGF as extra reinforcing material. The different percentages of both materials Fly ash and ARGF
were chosen as design factors to develop statistical models for the responses of fresh and mechanical parameters
at 7 and 28 days. The experimental runs were carried out utilizing Central Composite Design (CCD), and there
was a strong connection between the actual and estimated values. At a 95 % level of confidence, the effect of
multiple factors on the target response was studied using statistical analysis of ANOVA results. The coefficient of
determination (R2) values ranging from 0.82 to 0.99 show the reliability and performance of the proposed model.
The optimal nine responses values for the slump, 7D-CS, 28D-CS, 7D-STS, 28D-STS, 7D-FS, 28D-FS, 7D-UPV and
28D-UPV were 41.70, 1820.51, 2177.23,291.86, 378.85, 589.656, 756.58, 3.82, and 4.34 with desirability value
of 0.918. The linearity of the data points surrounding the line of fit for each mix blends indicating the model’s
efficiency in estimating the nine responses. This study will help in optimization of concrete mix design using
waste materials and fibers.

1. Introduction conventional cement is made, one ton of carbon dioxide (CO2) is pre­
dicted to be discharged into the air. The cement production line subse­
Both advanced and developing countries are facing issues in the quently produces around 7 % of world emissions in terms of CO2 [3].
removal and sustainable reuse of industrial by-products previously Carbon neutral concrete production is future of construction industry. It
defined as solid wastes in the last few years [1]. Concrete has become the is important to discover a less carbon or neutral alternative than Port­
most popular and frequently utilized material in construction on a global land cement. The acceleration of pollutant emissions caused by cement
basis due to progressive infrastructure construction. The concrete sector manufacturing can be rectified if only 30 % of world cement con­
requires over 1.5 BT (billion tonnes) of cement annually [2]. This has put sumption could be substituted with supplementary cementitious mate­
a lot of stress on the cement sector to produce a large amount of cement rials called as SCMs. SCMs have been an important aspect of concrete
to meet the increased consumption for infrastructural projects. If 1 ton of mixture in the past decade as an alternative to reduce energy usage and

* Corresponding author.
E-mail addresses: [email protected], [email protected], [email protected] (S.A. Raheel Shah), [email protected]
(N. Ben Kahla), [email protected] (A. Mahmood).

https://doi.org/10.1016/j.conbuildmat.2022.129722
Received 12 June 2022; Received in revised form 3 August 2022; Accepted 8 November 2022
Available online 18 November 2022
0950-0618/© 2022 Elsevier Ltd. All rights reserved.
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

CO2 footprints [4,5]. Fly ash has been considered to be more efficient materials have numerous advantages, including higher fiber reinforcing
than other additional ingredients like slag, waste glass, and rice husk ash effectiveness due to SCMs which improved the bonding between fiber
(RHA) in meeting most of the standards for high-strength concrete [6]. particles and the binder matrix [22]. SCMs can help with FRC work­
Authorities and other professional agencies are progressively con­ ability issues by using FA and slag [23]. SCMs such as slag and silica
cerned with developing more effective strategies for reducing the scar­ fume (SF) can also help to improve the dispersion of fiber particles [24].
city and depletion of natural materials to attain the noble goal of As a result, the combination of such ingredients demonstrates improved
sustainable development. Managing the quantity of waste produced and concrete characteristics. Several researchers have successfully investi­
waste disposal is one such approach. The production and management gated the behavior of commercial fibers in SCMs [25,26]. However,
of waste Fly Ash (FA) are one of the most significant environmental numerous studies have used various types of fibers such as GF, SF, and
concerns. FA is a residue of burning coal produced in a power station synthetic or industrial fibers with normal concrete to improve their
that is employed as a cement replacement material in concrete mixtures mechanical characteristics. Numerous researchers studied the mechan­
[7]. However, the pace at which it is produced is far surpassing its ical characteristics of ordinary concrete at various percentage levels of
current application. It is regarded as a waste substance with over 600 MT GF [27]. Similarly, Vijai et al. [28] evaluated the mechanical charac­
produced yearly out of which 80 % of it being thrown in ash dams and teristics of geopolymer concrete made with FA at various GF proportions
dumps [8]. The environment is endangered because of this. Depending ranging from 0.01 to 0.03 % of concrete volume. They discovered that
on their chemical content, ASTM C-168 defined two categories (low and adding GF to FA-based geopolymer concrete reduces the mechanical
high calcium) of FA for usage in Cement concrete. Due to its strong characteristics when compared to GPC incorporated FA content without
reactivity and productivity as a pozzolana, low calcium-based FA is the GF. Hence, it is vital to determine the influence of such added in­
most extensively utilized category of FA in cementitious and geopolymer gredients on the mixed cementitious material to produce concrete with
mixtures. On the other hand, FA of class C is linked to a high CaO desired characteristics. Due to the obvious considerable complexity of
component, which induces cement uncertainty and so limits its appli­ the interaction seen between multiple factors and outputs of concrete,
cation [9]. The number of research has been increased rapidly due to a typical regression analysis could not be sufficient to develop a precise
recent increase in the formation of FA (class C) as a by-product of the estimation. Recently, Response Surface Methodology (RSM) has become
coal burning in power generation stations [9]. FA was recommended for popular in the Civil Engineering field for predicting concrete properties
20–25 % of cement replacement by the overall weight of binder in quickly and accurately.
concrete production [10]. Identifying strategies to use more FA content RSM is a well-known mathematical or statistical predictive tool. It
is thus necessary to dramatically reduce trash dumping and its associ­ provides an adequate experimental study of non-linear 3d response
ated problems [11]. As a result, the Canadian center for mineral and surfaces to identify the effects of multiple factors. It determines linear
energy technology called CANMET recommended the first application of relationships and quadratic participation of various components to
HVFA in the manufacturing of concrete structures in the earlier 1980s make necessary changes in the concrete design mix to meet design
[10]. The cementitious ingredients in high volume fly ash (HVFA) criteria [29]. This strategy facilitates effective material consumption,
concrete must have an allowable min FA content of 50 % by overall lowering construction costs and reducing the possibility of structural
weight [12]. According to the preceding study [13], researchers have material failure [30]. RSM has been used in several cement concrete
recently used up to 70 % of HVFA in cement materials. The researcher investigations [30–34], but its application in SCMs with FA and ARGF
studied the effect of HVFA content ranging between 40 and 70 % on the has yet to be investigated.
concrete strength and hydration behavior. Rashad [14] studied the The objective of this study is to determine the effects of FA and ARGF
behavior of HVFA mixture comprising 70 % of FA substitution and on concrete properties, construct and validate statistical models for the
metakaolin of micro-sized that has been exposed to extreme temperature appropriate responses, and then maximize mortar blends for infra­
conditions. Posi et al. [15] studied HVFA based jointing mortar having structural applications. Moreover, the effects of the particle percentages
0 to 70 % of FA cement substitution. According to prior studies [15], of FA and ARGF as predictor factors on the fresh and hardened char­
using HVFA in cement improves workability, decreases shrinkage, re­ acteristics of concrete mixtures using RSM design are investigated in this
duces hydration heat, and improves durability at high temperatures. study. This method can speed up the use of fiber-reinforced fly ash based
Moreover, it reduces CO2 footprints and the cost of concrete concrete in real-world applications.
manufacturing by decreasing the amount of cement used [16].
On the other hand, the lower tensile characteristics of normal con­ 2. Background
crete are indeed a disadvantage as they restrict structural opportunities
in the construction sector. Optimum use of waste materials in combi­ 2.1. Design of experiments (DOE)
nation with fibres can help in developing sustainable concrete. The
splitting tensile strength (STS) of standard concrete is typically ranging The Design of Experiment commonly known as “DoE” is a useful
between 12 and 15 percent of its compressive strength [17]. Several approach for dealing with multiple factors in decision-making mecha­
researchers provide an appropriate solution for this issue and recom­ nisms [35]. Engineers, services, and industrial sectors have all embraced
mend the use of different fibers in normal concrete. Concrete researchers the approach to increase experimental efficiency [36]. Previously,
have proposed a variety of fibers including glass (GF) [18], steel (SF) problems regarding many variables have been addressed by using one
[19], carbon (CF), polypropylene (PPF) [20], and polyvinyl fibers (PVA) variable at a time called as OVAT method, which keeps all factors fixed
and among others for improving the crack resistance and tensile prop­ except one and performs tests until the best outcomes for the individual
erties of conventional concrete. However, the selection of such fibers is manipulated parameter is achieved. The technique is iterated for every
mostly determined by their intended use. Conversely, only a few aca­ factor until the correct solution is obtained while addressing the issue
demics have concluded that using commercial fibers is not a cost- with several variables. While the procedure is straightforward and
effective approach [21]. They have the disadvantage of reducing the precise, it still demands a large dataset and experiments, which takes a
concrete workability and increasing its cost in ordinary concrete. The lot of time, cost, and effort [37]. This problem is quite important in those
inclusion of a suitable plasticizer, depending on the quantity and type of investigations utilizing concrete materials. While such research focuses
fiber, has been recommended to help with low workability. As a result, on a single responsive parameter, such as the effect and proportion of
resolving such challenging issues is essential for the advancement of eco- substitute material on concrete characteristics [38,39]. A large variety
friendly sustainable concrete. A novel type of durable, inexpensive, of variables, some of which are significantly interconnected, influence
ductile, and eco-sustainable concrete can be made by combining fiber the overall performance of high strength and self-consolidated concrete
with SCMs. Researchers have discovered that combining both two (SCC) [40]. Sustainable concrete design using waste materials needs

2
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

even more research. In such scenarios, large sample size will be regarded construction materials.
as a prerequisite for all the important strength and durability factors.
The term “nr” denotes the number of options feasible for “r” factors that 2.3. Application of RSM
can take “n” integers. For example, an investigation with five factors and
five outcomes involves 3125 data sources, requiring extensive investi­ The DM and BBD procedures are the most efficient, according to
gation of concrete qualities a time- and expensive process that cannot be studies of data capture methodologies [57], however, most concrete past
accomplished only by laboratory tests [41]. papers employed the CCD approach to conducting the trials. In prior
DoE reduces the amount of data essential for optimal statistical research, the justification for choosing this strategy was not stated.
modeling, clarifying the effect of each factor on the response variables However, the CCD model enables equal accuracy in all dimensions as per
under consideration. On the other hand, the versatility of DoE analysis this study [58]. Other scholars [59,60] employed CCD because of its
may be a challenge for academics with a lack of appropriate field rotatability and capability to accurately forecast results within the
knowledge and an unclear approach to implementing the problem- studied range. Several past studies employed the Minitab tool to perform
solving job [42]. In recent years, the use of DoE in the research of the RSM trial analysis and data processing. Several other researchers
construction materials has increased in number. DoE is used in concrete preferred Design Expert. In one investigation, the experiment was car­
research ranging from predicting the strength of concrete to developing ried out using the statistical analysis system or SAS. Because BBD and
concrete design mixes. Non-destructive testing (NDT) of conventional CCD are two of Minitab’s approaches, this could explain why they are so
concrete frequently involves mathematical equations since the concrete popular. Whereas the RSM can handle more factors, many academics use
strength cannot be deduced from the findings of the NDT method [43]. it to construct trials with two [61] or three factors, probably to get most
Simple linear regression (SLR) with a scatter plot is typically used in this of the contour graph. It is widely utilized in design codes including
situation. Presently, novel DoE approaches, such as Response Surface Eurocode 2 to indicate the influence of two factors.
Methodology [44] or even deep learning [45], have been efficient in The workability of fresh concrete was chosen as a dependent (x)
achieving more accurate and precise predictions. DoE techniques are factor of RSM research in some research, as evidenced by slump lab tests.
extremely favorable for investigations relevant to suitable material Whereas the workability can be assessed early in the production process,
substitution [46], with a considerable number of latest works employing the addition of the slump factor is primarily used in the RSM analysis to
a more advanced approach to analyze test data [47]. Tire rubber [48], investigate the impact of substitute material on the workability of con­
palm oil fuel ash (POFA) [49], fly ash [49,50], and other substitute crete. The workability of fresh concrete with the w/b ratio was explored
materials have all been examined using different DoE approaches. In using an RSM plot in studies of concrete with E-wastes [1], rubber
another study, response surface methodology [51], curve fitting wastes [33], and fly ash [62].
methods [52], or ANNs-based techniques are used to maximize new
concrete design processes, which often include the addition of typical 3. Materials and methodology
eco-friendly waste materials.
3.1. Basic materials
2.2. Response surface Methodology (RSM)
Fiber-reinforced fly ash-based concrete is composed of OPC, fly ash,
The RSM is the DoE approach, which is used to study the impact and glass fiber, sand, coarse aggregates, and water. The proportions of all
interactions of multiple factors on a response variable. The main used material ingredients are determined using M20 concrete grade.
objective is to streamline the testing process and maximize the out­ According to ASTM C150 specifications[63], cement of grade 53 was
comes, similar to the conventional Taguchi approach. As per Bradley used as a pozzolanic material. The initial and final setting time of used
[53], the RSM method involves decoding the response surface geometry, cement was 127 min and 205 min. A Fly ash of class F produced by
along with the local max (+1), min (− 1), and ridge paths, as well as burning coal was utilized and it was collected from the local power
determining the location where the most suitable response occurs. To station[64]. The chemical compositions of the used binders (FA and
develop a response surface design that defines the optimal conditions for OPC) are presented in Table 1. The physical properties of the used in­
the predictor variables, the RSM model examines the 1st order, 2nd gredients are described in Table 2. The sand with a size range of <4.75
order, and combined effect in between factors. The RSM approach like mm was used whereas, the coarse aggregate of size ranging between
other DoE methods offers a mathematical approach to an issue, mini­ 4.75 and 12.5 mm was utilized in this study. The fineness modulus of
mizes the number of laboratory tests, as well as saves cost and time in sand was 2.39. The physical characteristics of used aggregates were
the performed investigation. It also considers the interaction of factors to tested under ASTM standards and specifications [65]. The bulk density
increase the accuracy and reliability of the model. However, one of sand and coarse aggregates was 1440 kg/m3 and 1530 kg/m3
drawback of the RSM model is that the observational results are fit to a respectively. Potable water was used for making all concrete mixtures.
quadratic polynomial order, which may not be the appropriate design The gradation curve of both aggregates is plotted in Fig. 1.
for representing all curvature systems[54]. Numerous types of fibers are commercially available; glass fiber (GF)
Although the RSM provides a comprehensive analytic tool for test is preferred due to its better surface area to weight ratio. The fabrication
results, it does not prescribe the data collection approach. However, of fiber-based concrete was done with alkali resistance glass fiber
unlike the Taguchi approach, which uses the Orthogonal Array, there (ARGF), which is unaffected by the alkaline state of the OPC [66]. They
are alternative collecting data strategies. Box-Behnken design (BBD), 3 are simple to incorporate into concrete mixtures, have a high modulus
levels (3 k) full factorial designs, central composite design (CCD), property and tensile strength, and do not corrode like a steel. The used
Doehlert design matrix (DM), and other methods are among the tech­ ARGF was available at commercial level. The physical characteristic of
niques used. As a result, it has few uses in comparison to RSM, as the the used fiber is presented in Table 2.
series of tests involved becomes too vast as the number of levels rises,
resulting in low data gathering performance [55]. The efficiency of the 3.2. Mix proportion and samples fabrications
other three approaches has been performed in several studies [56,57].
The CCD is one of the most effective approaches. The CCD design can be FA was used as a cementitious material in concrete mixtures at four
described for any study which may cause an error if conducted under varying percentages ranging between 0 and 40 % with different per­
severe conditions. Because such analysis does not involve pairings in centages of ARGF levels ranging from 0 % to 1.5 %. In this study, the
which all variables are at their max or min values [5]. This advantage is constant water-binder ratio of 0.60 was used for M20 grade concrete.
particularly important, and it should be considered when exploring The FA, OPC as well as fine and coarse aggregates were mixed all

3
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Table 1
Chemical composition of Fly ash and ordinary Portland cement (OPC).
Binder LOI SiO2 Al2 O3 Fe2 O3 CaO MgO SO3 K2 O Na2 O Others

FA (%) 1.79 47.65 23.27 2.16 9.98 1 0.79 _ _ 12.26

OPC (%) 3.55 20.80 5.02 3.15 62.02 1.86 2.64 0.66 0.42

[68]. The mix design proportion of each ingredient is described in

Table 2
Table 3.
Physical Properties of used ARGF.
Fiber Properties Values

ARG-Fiber Dia (μm) 12–20 3.3. Experimental testing methods

Tensile Strength (MPa) 1700
Tensile Strain (%) 2 The mechanical characteristics such as compressive, splitting tensile
Modulus of Elasticity (GPa) 72
strength, and flexural strength were evaluated and performed on the
Bulk Density (kg/m3 ) 410
Adhesion to Matrix Excellent
universal testing machine as shown in Fig. 2. The samples were cured for
Alkali Resistance Good 7 and 28 days, respectively. A cylindrical specimen with dimensions of
150 mm (dia) and 300 mm (height) was prepared for conducting testing
of compressive (CS) and splitting tensile strength (STS). For the flexural
together for around 4 min in a mechanical mixer. After that, the strength (FS) testing, beams with dimensions of 100 × 100 × 500 mm
calculated amount of water is added to it and mixed with an additional 2 were produced. The samples were cured and tested following ASTM
min. Finally, the fiber percentage was added to the mixture and stirred specifications [69]. The CS and STS tests were performed according to
for >5 min only. Special attention should be given to ensuring the ho­ ASTM standards at a constant loading rate of 5 kN per sec [70,71].
mogeneous mixing of all ingredients. The workability of freshly made Similarly, the constant loading rate of 1 kN per sec is followed for testing
concrete was measured using slump cone testing as per ASTM standards flexural strength as per ASTM standards [72]. Three samples at each
[67]. After that, samples were cast and demolded after 24 hr. and placed percentage’s levels of both FA and ARGF for each mechanical property
in a curing tank at room temperature as followed by ASTM standard were prepared and testing. Then, the average values were taken for the

Fig. 1. Gradation curve for fine and coarse aggregates.

Table 3
Mix Design Proportion for ARGF based FA concrete.
Mix ID OPC FA FA ARGF ARGF Sand Water Aggregate
3 3 3 3 3
Unit kg/m % kg/m % kg/m kg/m kg/m kg/m3

M1-C 415 0 0 0 0 616 249 1320

M2 415 0 0 0.5 3.24 616 249 1320
M3 415 0 0 1 6.49 616 249 1320
M4 415 0 0 1.5 9.73 616 249 1320
M5 373 10 21.64 0 0 616 236.4 1320
M6 373 10 21.64 0.5 3.24 616 236.4 1320
M7 373 10 21.64 1 6.49 616 236.4 1320
M8 373 10 21.64 1.5 9.73 616 236.4 1320
M9 331 20 43.09 0 0 616 224 1320
M10 331 20 43.09 0.5 3.24 616 224 1320
M11 331 20 43.09 1 6.49 616 224 1320
M12 331 20 43.09 1.5 9.73 616 224 1320
M13 289 30 65 0 0 616 212.4 1320
M14 289 30 65 0.5 3.24 616 212.4 1320
M15 289 30 65 1 6.49 616 212.4 1320
M16 289 30 65 1.5 9.73 616 212.4 1320
M17 249 40 86.36 0 0 616 202 1320
M18 249 40 86.36 0.5 3.24 616 202 1320
M19 249 40 86.36 1 6.49 616 202 1320
M20 249 40 86.36 1.5 9.73 616 202 1320

4
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 2. (a) Universal Testing Machine by Matest brand; (b) Compressive cylinder (c) Flexural Beam; (d) Splitting Tensile Cylinder at 7- and 28-days curing.

outcomes.
Table 5
Actual and Coded Design Factors with their space type.
3.4. Experimental setup using RSM
No. Actual Factors Coded Factors (+1, − 1)

RSM is performed to design trials and optimize factors to anticipate Space Type X1: FA (%) X2: ARGF (%) A B
desired results to establish links between several factors [73]. A com­ 1 Axial 20 0 0 − 1
bination of sixteen trial runs was created followed by a central com­ 2 Factorial 40 0 1 − 1
posite design (CCD). The CCD technique is a partial factorial way of 3 Factorial 0 1.5 − 1 1
determining a functional interaction among input variables and re­ 4 Axial 20 1.5 0 1
5 Factorial 0 0 − 1 − 1
sponses [74]. The Design Expert (13 version) software was used for 6 Factorial 40 0 1 − 1
experimental design setup, and computational and statistical analysis. 7 Factorial 0 1.5 − 1 1
The ANOVA results were used to explore the relationships between the 8 Factorial 40 1.5 1 1
input factors such as FA and ARGF contents and responses such as slump, 9 Factorial 40 0 1 − 1
10 Axial 20 0 0 − 1
CS, STS, FS, and UPV after 7- and 28-days curing.
11 Factorial 40 0 1 − 1
The factors were chosen as the volumetric percentages of FA and 12 Factorial 0 0 − 1 − 1
ARGF, which were designated as “A” and “B”, respectively. The higher 13 Factorial 0 1.5 − 1 1
(+1) and lower (− 1) levels of each factor are presented in Table 4. The 14 Factorial 40 1.5 1 1
actual and codded values of experimental runs are described in Table 5. 15 Factorial 0 0 − 1 − 1
16 Factorial 40 1.5 1 1
The experimental runs were calculated using the following Equations
(1).
4. Results and discussions
N = 2k + 2k + c (1)

Where k represents the number of factors involved, the factorial points 4.1. ANOVA analysis
are represented by 2 k, 2 k is the axial space type, and “c” indicates the
number of trials in the center location [75]. The effects and interactions of two factors Fly ash (A) and ARGF (B)
After conducting experiments, the CCD design dataset was evaluated on the fresh and mechanical properties of the modified concrete were
to develop multivariate regression calculations as a statistical method tested and predicted using the RSM model. The multiple regression
for estimating concrete parameters. The 2nd-order polynomial or analysis was performed on the experimental dataset using the RSM
quadratic equation (Eq. (2)) has been established to characterize the model. The ANOVA results by the predicted model based on the CCD
correlation between the variables. method are presented in Table 3. At the significance level of 0.05, the
study presents the sum of squares (SS), F-value, and P-value. The
Y = β0 + β1 A + β2 B + β12 AB + β11 A2 + β22 B2 (2) importance of each factor is evaluated using P-values of 0.05 and 0.01 as
a measure, suggesting an acceptable and acceptable match between
Where Y is the projected response value, β0 is the intercept, β1 , and β2 predicted and actual values [5,76]. The ANOVA results of our study
are called coefficients of linear effects, and β12 is the coefficient of show model P-values<0.005 for each response variable which indicates
combined effect[1]. β11 and β22 are the 2nd order coefficient of effects. a strong significant relationship between coefficients (A and B) and re­
To assess the performance of the quadratic projections, the sum of sponses as described in Table 3. The F-value of the model for the slump,
squares (SS), F value, probability, or P-value with a 95 % of confidence compressive strength (CS), splitting tensile strength (STS), flexural
level, and R-squared (R2) were calculated using RSM. strength (FS), and ultrasonic pulse velocity (UPV) after 7 and 28 days of
curing are observed as 190.48, 97.57, 34.21, 352.71, 89.54, 157.44,
167.15, 57.42 and 18.59 respectively as presented in Tables 6-10. It
demonstrates the importance of the resulting models. The efficiency and
Table 4 validity of the model are further tested based on their Lack of Fits and F
Factors and their levels using RSM based CCD Model.
values. The lack of Fits shows the data variance near the fitted model
Factors Units Experimental Levels in RSM based CCD [77]. The Lack of Fit is insignificant if their P-value is > 0.005 and it
Low Level (-1) High Level (+1) showed to be greater than for acceptable results.
The coefficient of determination (R2) is another statistical measure to
A: Fly Ash Content % 0 40
B: ARGF Content % 0 1.5 assess the reliability and performance of the proposed model. The R2

5
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Table 6
Statistical measures to check RSM model accuracy and validity.
Responses Std. Dev Mean C.V. % R2 Adj. R2 Pred. R2 Adeq. Precision

Slump (mm) 2.59 45.07 5.74 0.9858 0.9806 0.9595 35.4187

7-CS (PSI) 69.74 158.23 4.41 0.9858 0.9508 0.9442 24.8517
28-CS (PSI) 209.68 1906.50 11.00 0.8404 0.8158 0.7878 13.9696
7-STS (PSI) 8.55 216.07 3.96 0.9888 0.9860 0.9840 41.3948
28-STS (PSI) 17.74 271.32 6.54 0.9572 0.9465 0.9380 20.7733
7-FS (PSI) 23.03 375.03 6.14 0.9752 0.9690 0.9638 31.1395
28-FS (PSI) 24.21 510.86 4.74 0.9768 0.9758 0.9666 32.3147
7-UPV (Km/s) 0.0721 3.46 2.08 0.9349 0.9186 0.9060 28.8810
28-UPV (Km/s) 0.1544 3.87 3.99 0.8229 0.7786 0.7509 10.1440
Standard <30 0–1 Difference < 2 >4

Table 7
ANOVA Results for the Response variable Slump.
Responses Parameters Sum of Sq. dF Mean Sq. F-value p-value Significant

Model 5098.04 4 1274.51 190.48 <0.0001 Yes

Y1: Slump (mm) A-Fly Ash Content 1305.08 1 1305.08 195.05 <0.0001 Yes
B-ARGF Content 3722.73 1 3722.73 556.38 <0.0001 Yes
AB 192.48 1 192.48 28.77 0.0002 Yes
A2 5.10 1 5.10 0.7616 0.4015
B2 0.0000 0
Residual 73.60 11 6.69
Lack of Fit 73.60 1 73.60
Pure Error 0.0000 10 0.0000
Cor Total 5171.64 15

Table 8
ANOVA Results for the Response variables: 7D and 28D- Compressive strength.
Responses Parameters Sum of Sq. dF Mean Sq. F-value p-value Significant

Model 1.424E + 06 3 4.746E + 05 97.57 <0.0001 Yes

Y2: 7D-CS (PSI) A-Fly Ash Content 1.198E + 06 1 1.198E + 06 246.36 <0.0001 Yes
B-ARGF Content 2.549E + 05 1 2.549E + 05 52.41 <0.0001 Yes
AB 47686.97 1 47686.97 9.80 0.0087
Residual 58369.40 12 4864.12
Lack of Fit 58369.40 2 29184.70
Pure Error 0.0000 10 0.0000
Cor Total 1.482E + 06 15
Y3: 28D-CS (PSI) Model 3.008E + 06 2 1.504E + 06 34.21 <0.0001 Yes
A-Fly Ash Content 2.609E + 06 1 2.609E + 06 59.34 <0.0001 Yes
B-ARGF Content 5.328E + 05 1 5.328E + 05 12.12 0.0041
Residual 5.715E + 05 13 43963.90
Lack of Fit 5.715E + 05 3 1.905E + 05
Pure Error 0.0000 10 0.0000
Cor Total 3.580E + 06 15

Table 9
ANOVA Results for the Response variable 7D and 28D- Splitting Tensile strength (STS).
Responses Parameters Sum of Sq. dF Mean Sq. F-value p-value Significant

Model 77343.06 3 25781.02 352.71 <0.0001 Yes

Y4: 7D-STS (PSI) A-Fly Ash Content 76457.71 1 76457.71 1046.00 <0.0001 Yes
B-ARGF Content 1987.53 1 1987.53 27.19 <0.0001 Yes
AB 605.65 1 605.65 8.29 0.0087
Residual 877.14 12 73.10
Lack of Fit 877.14 2 438.57
Pure Error 0.0000 10 0.00
Cor Total 78220.30 15
Y5: 28D-STS (PSI) Model 84507.77 3 28169.26 89.54 <0.0001 Yes
A-Fly Ash Content 80878.84 1 80878.84 257.09 <0.0001 Yes
B-ARGF Content 1561.37 1 1561.37 4.96 0.0458
AB 2071.58 1 2071.58 6.58 0.0247
Residual 3775.11 12 314.59
Lack of Fit 3775.11 2 1887.55
Pure Error 0.0000 10 0.0000
Cor Total 88282.88 15

6
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Table 10
ANOVA Results for the Response variable 7D and 28D- Flexural strength (FS).
Responses Parameters Sum of Sq. dF Mean Sq. F-value p-value Significant

Model 2.505E + 05 3 83513.43 157.44 <0.0001 Yes

Y6: 7D-FS (PSI) A-Fly Ash Content 82096.14 1 82096.14 154.76 <0.0001 Yes
B-ARGF Content 1.542E + 05 1 1.542E + 05 290.71 <0.0001 Yes
AB 4685.86 1 4685.86 8.83 0.0117
Residual 6365.55 12 530.46
Lack of Fit 6365.55 2 3182.78
Pure Error 0.0000 10 0.0000
Cor Total 2.569E + 05 15
Y7: 28D-FS (PSI) Model 2.938E + 05 3 97940.78 167.15 <0.0001 Yes
A-Fly Ash Content 1.231E + 05 1 1.231E + 05 210.07 <0.0001 Yes
B-ARGF Content 1.490E + 05 1 1.490E + 05 254.24 <0.0001 Yes
AB 14853.29 1 14853.29 25.35 0.0003
Residual 7031.50 12 585.96
Lack of Fit 7031.50 2 3515.75
Pure Error 0.0000 10 0.0000
Cor Total 3.009E + 05 15

Fig. 3. Actual and predicted values for nine different responses: (a) Slump; (b) 7D-CS; (c) 28D-CS; (d) 7D-STS; (e) 28D-STS; (f) 7D-FS; (g) 28D-FS; (h) 7D-UPV; (i)
28D-UPV.

7
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 3. (continued).

values show how closely our data matches the fitted model. In general, responses. The conformity of the data points across the fitted line for
the higher the R-square value, the better will be the model performance. each of the responses in the actual vs predicted plots demonstrates how
It ranges from 0 to 1. In this study, the R-square values of model factors close the estimated responses were to the actual responses. Moreover,
such as slump, 7D-CS, 28D-CS, 7D-STS, 28D-STS, 7D-FS, 28D-FS, 7D- the linear dispersion of the sample points near the fit line in the normal
UPV, and 28D-UPV were 0.986, 0.986,0.84, 0.99, 0.96, 0.97, 0.98, displays of residuals for each response indicates that the designs were
0.93 and 0.82 respectively. The high coefficients of determination sug­ significant and the error terms were dispersed normally [77]. In
gested that the models fit the dataset well. Moreover, the predicted and conclusion, the statistical results for each response are within range and
adjusted R2 values with their difference <0.2 is preferred [78] and in our satisfactory.
case, all values of the responses were within the acceptable difference
range as seen in Tables 6-10. The ratio of standard deviation to the mean
4.2. Effects of FA and ARGF on fresh properties
is termed as coefficient of variation (CV) which is the measure of
dispersion of dataset from the mean. It is measured in percentage and
The ANOVA results were obtained for slump values (response vari­
<30 is acceptable. In our cases, all values of CV were within ranged and
able) to study the effects of FA and ARGF combinations based on P-
showed acceptable fits of the proposed model. The value >30 % is not
values as presented in Table 7. The F-value of 190.48 for the model
consider desirable. Likewise, the signal-to-noise ratio is determined by
indicates that it is significant. An F-value of this magnitude has a 0.01 %
the appropriate precision level or Adeq. precision, which must be >4
chance of occurring due to noise. Model terms having P-values <0.050
[79]. In this study, these values were 35.42, 24.85,13.96, 41.39,
are significant. The regression coefficients are not significant if this
20.77,31.14, 32.31, 28.88 and 10.14 for the slump, CS, STS, FS and UPV
value is >0.100. In this scenario, remove the insignificant model terms
at 7 and 28 days of curing. These statistics reflect that there was a
to improve model efficiency if there are so many useless model terms
positive indication and so the models may be utilized to guide the design
(except those needed to aid hierarchy). In this study, both model terms
process.
are significant. A, B, and AB are important model factors in our case as
The Slump, CS, STS, FS, and UPV models at 7 and 28 days were also
described in ANOVA Table 7. The max and min levels of input factors are
diagnosed using the actual vs predicted plots illustrated in Fig. 3,
codded in +1 and − 1 respectively. By comparing the factor coefficients
respectively. The linearity of the data points surrounding the line of fit in
(i.e., A & B), the coded equation can be used to determine the relative
all scenarios indicated that the models were accurate in estimating the
contribution of the factors. The final equation of the model in the form of

8
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

codded factors A and B is presented in Eq. (3) for the response variable compressive strength (CS) values at 7 and 28 days are presented in
(slump). Fig. 5. These response plots describe the effect and interaction of Fly ash
and GF factors upon the compressive strength values (response vari­
Slump (Y1) = + 44.70 − 10.09A − 15.51B − 3.88AB − 1.46A2 (3)
able). The 7- and 28-days CS values of the response variables ranged
The 2D and 3D surface plots of the Fly ash and GF combinations for between the 1068.92–1997.51 PSI (7.37 MPa to 13.77 MPa) and
slump values are presented in Fig. 4. These response plots describe the 1319.37–2557.35 (9.09 to 17.64 MPa) respectively. For comparison
effect and interaction of Fly ash and GF factors upon the slump values purposes, a control specimen was cast at 0 % of both FA and GF content.
(response variable). The min and max slump values of modified concrete The highest CS values were represented by the red region whereas, the
is lies between 12.7 and 66.04. The high slump values of the modified lowest CS values were reported above 40 % FA and 1.5 % of GF as can be
concrete are shown in red color. However, the green and blue plot re­ observed from the contour line plot and region.
gions show the minimum and intermediate values of slump respectively.
The optimum response can be seen by the red region bounded by slump 4.3.2. Response surface for splitting tensile strengths (STS)
values of 50–60 mm in the contour line plot. Any pairing of these factors The ANOVA analysis was studied for splitting tensile strength (STS)
FA and ARGF below these threshold values (up to 20 % of FA and 0.75 % values (response variable) after 7 and 28 days to study the effects of FA
of ARGF) will result in higher slump values. However, any pairing of and GF combinations based on P-values as presented in Table 9. For 7D-
factors just above indicated range resulted in a lower surface response. STS, the F-value of 352.71 for the model indicates that it is significant.
The slump values were significantly reduced as the percentage of FA and Model terms having P-values <0.050 are significant. The regression
GF increased. The slump values are dramatically lowered as the pore coefficients (A and B) are not significant if this value is >0.100. In this
volume of the concrete matrix decreased with increasing in FA content case, the models are significant for STS at 7 and 28 days. In this case A, B,
(%). Similar results were reported in these studies [80,81]. Because FA is AB are significant model terms at 7 and 28 days of STS. The values
a porous material having macro and micro levels both within and even >0.1000 signify that model terms are not significant. Similarly, the
on the particle surface, a certain quantity of adding water will be Model F-value of 89.54 for 28D-STS implies the model is significant. By
consumed on its surface. comparing the factor coefficients (i.e., A & B), the coded equation can be
used to determine the relative contribution of the factors. The final
equation of the model in the form of codded factors A and B is presented
4.3. Effects of FA and GF on mechanical properties
Eq. (6) and (7) for both response variables (STS) at 7 and 28 days.

4.3.1. Response surface for compressive strengths (CS) 7D − STS (Y4) = + 219.06 − 77.20A − 11.27B − 6.87AB (6)
The ANOVA analysis was studied for compressive strength values
(response variable) after 7 and 28 days to study the effects of FA and GF 28D − STS (Y5) = + 276.74 − 79.40A + 9.99B − 12.71AB (7)
combinations based on P-values as presented in Table 8. For 7D-CS, the The 2D and 3D surface plots for the combinations of Fly ash and GF
F-value of 97.57 for the model indicates that it is significant. Model contents on the splitting tensile strength (STS) properties of FA-ARGF-
terms having P-values <0.050 are significant. The regression co­ based concrete at 7 and 28 days are presented in Fig. 6. The splitting
efficients (A and B) are not significant if this value is >0.100. In this tensile strength were ranging from 120.86 to 305.46and 195.29 to
case, the models are significant for CS at 7 and 28 days. In this case, A, B, 387.065 PSI after 7 and 28 days respectively. When compared with
and AB are significant model terms. If these values are >0.1000 signify standard concrete mix, the STS of concrete incorporating 10 % of FA
the model terms are not significant. Similarly, the Model F-value of improves by about 6 % at 28 days as shown in the red region of the
34.21 for 28D-CS implies the model is significant. In this case, A, and B contour plot (Fig. 6). After 28 days of curing, the splitting tensile
are significant model terms. By comparing the factor coefficients (i.e., A strength of the modified concrete decreases as more cement is replaced
& B), the coded equation can be used to determine the relative contri­ with FA. The 28D-STS rises between 195.29 and 3387.065 PSI with the
bution of the factors. The final equation of the model in the form of increase of ARGF content and 0 % of FA, and the maximum STS was
codded factors A and B is presented in Eq. (4) and (5) for both response found when 1.5 % of ARGF content was used. The strength of 7D-mix
variables (CS) after 7 and 28 days. including 10 % FA in addition to 1.5 % of ARGF is over 70 % greater
7D − CS (Y2) = + 1581.57 − 305.64A − 127.67B − 60.97AB (4) than conventional concrete. The increased tensile strengths were due to
improved bonding between cement paste or matrix and fiber addition.
28D − CS (Y3) = + 1911.59 − 449.90A − 184.26B (5) The STS were improved at 28 days in comparison with the samples at 7
days. Moreover, the STS were decreased with an increase in FA contents,
The 2D and 3D surface plots of the Fly ash and GF combinations for

Fig. 4. Surface plots of response variable Slump: (a) 2D Surface; (b) 3D surface.

9
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 5. Surface plots of response variables compressive strength (a) 2D Surface of 7D-CS; (b) 3D surface of 7D-CS; (c) 2D Surface of 28D-CS; (d) 3D surface of
28D-CS.

but ARGF addition improves their overall strengths as compared to terms having P-values <0.050 are significant. The regression co­
control specimens. According to this study, the binder matrix r improved efficients (A and B) are not significant if this value is >0.100. In this
the pozzolanic effect and played an important role in improving the case, the models are significant for FS at 7 and 28 days. In this case, A, B,
strength of ARGF concrete. The STS were significantly improved at 10 % and AB are significant model terms at 7 and 28 days of FS. Similarly, the
FA due to the falling action of FA material. However, using>10 % FA Model F-value of 167.15 for 28D-STS implies the model is significant. By
reduces the concrete STS, but it is still higher than the standard mixture. comparing the factor coefficients (i.e., A, B, and AB), the coded equation
The minimum STS was observed after using 40 % of FA content at 7 and can be used to determine the relative contribution of the factors. The
28 days. Due to the reduced cement matrix quality of fibers at an final equation of the model in the form of codded factors A and B is
increased proportion of FA, the strength increases owing to fiber addi­ presented in Eq. (8) and (9) for both response variables (FS) at 7 and 28
tion at 40 %. According to a study[82], the poor cement matrix has days.
inferior fiber strength development to the strong cementitious matrix.
7D − FS (Y6) = + 391.24 − 80.0A + 99.30B − 19.11AB (8)
To conclude this, a 10 % substitute of FA is advised to achieve optimum
ductility advantages.
28D − FS (Y7) = + 527.0 − 97.96A + 97.60B − 34.03AB (9)

4.3.3. Response surface for flexural strengths (FS) The 2D and 3D surface plots for the combinations of Fly ash and GF
The ANOVA analysis was studied for Flexural strength (FS) values contents on the flexural strength (FS) properties of FA-ARGF-based
(response variable) after 7 and 28 days to study the effects of FA and GF concrete at 7 and 28 days are presented in Fig. 7. The flexural
combinations based on P-values as presented in Table 10. For 7D-FS, the strength was ranging from 234.19 to 577.75 and 355.56 to 748.165 PSI
F-value of 157.44 for the model indicates that it is significant. Model after 7 and 28 days respectively. In comparison to each concrete blend,

10
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 6. Surface plots of Response variables Splitting Tensile strength (a) 2D Surface of 7D-STS; (b) 3D surface of 7D-STS; (c) 2D Surface of 28D-STS; (d) 3D surface of
28D-STS.

the maximal strength was reached with a 1.5 % fiber content in absence and 28D curing is presented in 2D and 3D surface plots (Fig. 8). The
of FA content at 7 and 28 days. The strength was significantly reduced effect of each factor on the UPV is comprehensively evaluated at both
with the increase of FA content as displayed in the contour plot. How­ curing ages 7 and 28. The UPV values ranged between 3.09 and 3.85 and
ever, the improved strength was achieved in the case of each combi­ 3.46 to 4.3 for 7 and 28 days of UPV samples. The highest UPV values are
nation of FA and ARGF as compared to reference samples. The FS of described in red color as shown in contour plots. Whereas the minimum
modified concrete is improved by a firm cement matrix with the addition UPV values are represented by dark blue colors. According to the find­
of fiber particles. As discussed in the introduction section, this fiber- ings, the 10 % of FA concrete achieves optimum strength at both curing
based cement bonding is beneficial to bear the tensile load. As a ages in comparison to all FA mix designs. The ultrasonic pulse velocity
result, our findings are consistent with past research [82,83]. decreased as the percentage levels of ARGF and FA increased. The 28D-
UPV strength values are nearly identical in the case of both 30 and 40 %
of FA-based concrete mixtures. The lower content of FA material gives
4.4. Response surface for ultrasonic pulse velocity (UPV)
higher strength as compared to their other percentages. The higher UPV
values were obtained up to 10 % of FA and 1.5 % of ARGF content.
The ANOVA analysis was studied for Ultrasonic pulse velocity (UPV)
values (response variable) after 7 and 28 days to study the effects of FA
5. Optimization and model validation
and GF combinations based on P-values as presented in Table 11. For
7D-UPV, the F-value of 57.42 for the model indicates that it is signifi­
It is difficult to get the highest value for multiple responses at the
cant. Model terms having P-values <0.050 are significant. The regres­
same time. As a result, numerous responses are optimized using multi-
sion coefficients (A and B) are not significant if this value is >0.100. In
objective optimization techniques. In this research, a compromise way
this case, the models are significant for UPV at 7 and 28 days. In this
of optimization was used for nine multiple responses. Two factors Fly
case, A, and AB are significant model terms at 7 days of UPV. Similarly,
ash and ARGF content was used as independent factors as discussed
the Model F-value of 18.59 for 28D-UPV implies the model is significant.
earlier. Nine responses such as Slump (mm), 7-Day CS, 28-Day CS, 7-Day
In this case, only coefficient AB are significant model terms at 28 days of
STS, 28-Day STS, 7-Day FS, 28-Day FS, 7-Day UPV, and 28-Day UPV
UPV. By comparing the factor coefficients (i.e., A, B, and AB), the coded
were used as a dependent variable. To optimize the nine parameters,
equation can be used to determine the relative contribution of the fac­
RSM was employed to determine the ideal content of the two factors. In
tors. The final equation of the model in the form of codded factors A and
this study, Design Expert 13-version software was used to apply process
B is presented in Eq. (10) and (11) for both response variables (UPV) at 7
optimization. Each variable and response have an important level. A
and 28 days.
pretty close result fulfilling the predefined upper and lower bounds is
7D − UPV (Y8) = + 3.46 − 0.1495A − 0.0125B − 0.2266AB (10) achieved using a multi-objective way of optimization. Desirability re­
lates to how near the solution is to the actual return. The desirability
28D − UPV (Y9) = + 3.86 − 0.0861A + 0.0825B − 0.3054AB (11) nearer to 1 is recommended to discover better results. Fig. 9 depicts the
effect of optimization algorithms on the predicted value function and the
The ultrasonic pulse velocity of each modified concrete batch at 7D

11
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 7. Surface plots of Response variables Flexural Strength (a) 2D Surface of 7D-FS; (b) 3D surface of 7D-FS; (c) 2D Surface of 28D-FS; (d) 3D surface of 28D-FS.

Table 11
ANOVA Results for the Response variable 7D and 28D- Ultrasonic Pulse Velocity (UPV).
Responses Parameters Sum of Sq. dF Mean Sq. F-value p-value Significant

Model 0.8942 3 0.2981 57.42 <0.0001 Yes

Y8: 7D-UPV (Km/s) A-Fly Ash Content 0.2867 1 0.2867 55.24 <0.0001 Yes
B-ARGF Content 0.0025 1 0.0025 0.4721 0.5051
AB 0.6587 1 0.6587 126.89 <0.0001 Yes
Residual 0.0623 12 0.0052
Lack of Fit 0.0623 2 0.0311
Pure Error 0.0000 10 0.0000
Cor Total 0.9565 15
Y7: 28D-UPV (Km/s) Model 1.33 3 0.4430 18.59 <0.0001 Yes
A-Fly Ash Content 0.0950 1 0.0950 3.99 0.0691
B-ARGF Content 0.1064 1 0.1064 4.47 0.0562
AB 1.20 1 1.20 50.21 <0.0001 Yes
Residual 0.2860 12 0.0238
Lack of Fit 0.2860 2 0.1430
Pure Error 0.0000 10 0.0000
Cor Total 1.61 15

12
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

Fig. 8. Surface plots of Response variables Ultrasonic Pulse Velocity (a) 2D Surface of 7D-UPV; (b) 3D surface of 7D-UPV; (c) 2D Surface of 28D-UPV; (d) 3D surface
of 28D-UPV.

Fig. 9. 2D and 3D surface plots for desirability for fly ash and ARG-fiber.

desirability value of 0.918 was obtained which indicates that response 6. Conclusion
optimization is possible. The optimal nine responses values for the
slump, 7D-CS, 28D-CS, 7D-STS, 28D-STS, 7D-FS, 28D-FS, 7D-UPV and In this study, Fly ash was employed as a supplementary cementitious
28D-UPV were 41.70, 1820.51, 2177.23,291.86, 378.85, 589.656, material (SCMs), while ARGF was added as an extra reinforcing ingre­
756.58, 3.82, and 4.34 with desirability value of 0.918. dient to produce a sustainable concrete. RSM has been used to forecast
the desired qualities of concrete by obtaining an optimal mixed pro­
portion of the main constituent based on computational and statistical
models. The ANOVA statistics are used to examine the effect of an

13
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

individual factor on the responses. The following main conclusions were References
derived from this study.
[1] K. Senthil Kumar, K. Baskar, Response surfaces for fresh and hardened properties of
concrete with e-waste (HIPS), J. Waste Manage. (2014).
• The suggested predictive models based on the P-value metric were [2] P. Shafigh, et al., Oil palm shell lightweight concrete containing high volume
found to be statistically significant and useful in estimating the ground granulated blast furnace slag, Constr. Build. Mater. 40 (2013) 231–238.
studied fresh and mechanical characteristics of ARGF-based Fly ash [3] T. Blankendaal, P. Schuur, H. Voordijk, Reducing the environmental impact of
concrete and asphalt: a scenario approach, J. Cleaner Prod. 66 (2014) 27–36.
concrete. The statistical results for each response are within range [4] M.M.A. Elahi, et al., Improving the sulfate attack resistance of concrete by using
and satisfactory. supplementary cementitious materials (SCMs): a review, Constr. Build. Mater. 281
• The R-square values of model factors such as slump, 7D-CS, 28D-CS, (2021), 122628.
[5] M. Haque, et al., Prediction and optimization of the fresh and hardened properties
7D-STS, 28D-STS, 7D-FS, 28D-FS, 7D-UPV, and 28D-UPV were of concrete containing rice husk ash and glass fiber using response surface
0.986, 0.986,0.84, 0.99, 0.96, 0.97, 0.98, 0.93 and 0.82 respectively. methodology, Case Stud. Constr. Mater. 14 (2021), e00505.
• The slump values of modified concrete decrease as the percentage of [6] H. Alhazmi, et al., Utilization of polymer concrete composites for a circular
economy: a comparative review for assessment of recycling and waste utilization,
Fly ash and ARGF proportions rise due to a decrease in available
Polymers 13 (13) (2021) 2135.
pores volume. [7] Z. Yao, et al., A comprehensive review on the applications of coal fly ash, Earth Sci.
• The mechanical properties such as Compressive strength (CS), Rev. 141 (2015) 105–121.
splitting tensile strength (STS), and flexural strength (FS) at 7 and 28 [8] E.-H. Yang, Y. Yang, V.C. Li, Use of high volumes of fly ash to improve ECC
mechanical properties and material greenness, ACI Mater. J. 104 (6) (2007) 620.
days of curing have shown the optimum responses at the Fly ash [9] G. Xu, X. Shi, Characteristics and applications of fly ash as a sustainable
content of up to 20 % and 1.5 % of ARGF addition. Moreover, the construction material: a state-of-the-art review, Resour. Conserv. Recycl. 136
inclusion of ARGF as a reinforcing material reduces the brittleness of (2018) 95–109.
[10] L. Lam, Y. Wong, C.S. Poon, Degree of hydration and gel/space ratio of high-
modified concrete and improves the bonding of ARG fibers and volume fly ash/cement systems, Cem. Concr. Res. 30 (5) (2000) 747–756.
binder, resulting in increased concrete strength. [11] J. Shekhovtsova, et al., Estimation of fly ash reactivity for use in alkali-activated
• The target responses were optimized by developing a mathematical cements-a step towards sustainable building material and waste utilization,
J. Cleaner Prod. 178 (2018) 22–33.
system, and the best mixture was discovered with 10–20 % Fly ash [12] M. Reiner, K. Rens, High-volume fly ash concrete: analysis and application, Pract.
and 1.5 % ARGF predicated upon the RSM model, which was Period. Struct. Des. Constr. 11 (1) (2006) 58–64.
experimentally confirmed with <5 % error. [13] J. Sun, et al., Compressive strength and hydration characteristics of high-volume
fly ash concrete prepared from fly ash, J. Therm. Anal. Calorim. 136 (2) (2019)
• The optimal nine responses values for the slump, 7D-CS, 28D-CS, 7D- 565–580.
STS, 28D-STS, 7D-FS, 28D-FS, 7D-UPV and 28D-UPV were 41.70, [14] A.M. Rashad, Investigation on high-volume fly ash pastes modified with micro-size
1820.51, 2177.23,291.86, 378.85, 589.656, 756.58, 3.82, and 4.34 metakaolin subjected to high temperatures, J. Central South Univ. 27 (1) (2020)
231–241.
with desirability value of 0.918.
[15] P. Posi, et al., Effect of fly ash fineness on compressive, flexural and shear strengths
• The structural concrete developed in this study uses significant of high strength-high volume fly ash jointing mortar, GEOMATE J. 16 (54) (2019)
amounts of both FA and ARGF, which is a successful outcome that 36–41.
can aid in addressing the issues of waste management and resource [16] M. Murali, et al., Utilization of crumb rubber and high-volume fly ash in concrete
for environmental sustainability: RSM-based modeling and optimization, Materials
depletion and will contribute to acheive long term sustainability. 14 (12) (2021) 3322.
Moreover, it was evident from this study that utilization of FA and [17] M.F.M. Zain, et al., Prediction of splitting tensile strength of high-performance
ARGF has promising applications in terms of improved durability, concrete, Cem. Concr. Res. 32 (8) (2002) 1251–1258.
[18] B. Ali, L.A. Qureshi, S.U. Khan, Flexural behavior of glass fiber-reinforced recycled
better service life and cost effectiveness. aggregate concrete and its impact on the cost and carbon footprint of concrete
pavement, Constr. Build. Mater. 262 (2020), 120820.
CRediT authorship contribution statement [19] V. Afroughsabet, L. Biolzi, P.J. Monteiro, The effect of steel and polypropylene
fibers on the chloride diffusivity and drying shrinkage of high-strength concrete,
Compos. B Eng. 139 (2018) 84–96.
Syyed Adnan Raheel Shah: Conceptualization, Conceptualization, [20] I. Hussain, et al., Comparison of mechanical properties of concrete and design
Data curation, Methodology, Resources, Software, Supervision, Writing thickness of pavement with different types of fiber-reinforcements (steel, glass, and
polypropylene), Case Stud. Constr. Mater. 13 (2020), e00429.
– original draft. Nabil Ben Kahla: Conceptualization, Project adminis­ [21] B. Ali, L.A. Qureshi, R. Kurda, Environmental and economic benefits of steel, glass,
tration, Supervision. Miniar Atig: Formal analysis, Investigation. M. and polypropylene fiber reinforced cement composite application in jointed plain
Kashif Anwar: Data curation, Formal analysis, Investigation, Method­ concrete pavement, Compos. Commun. 22 (2020), 100437.
[22] Z. Wu, C. Shi, K. Khayat, Influence of silica fume content on microstructure
ology, Software, Writing – original draft. Marc Azab: Validation,
development and bond to steel fiber in ultra-high strength cement-based materials
Writing – review & editing. Atif Mahmood: Validation, Writing – re­ (UHSC), Cem. Concr. Compos. 71 (2016) 97–109.
view & editing. [23] C.D. Atiş, O. Karahan, Properties of steel fiber reinforced fly ash concrete, Constr.
Build. Mater. 23 (1) (2009) 392–399.
[24] J.-K. Kim, et al., Tensile and fiber dispersion performance of ECC (engineered
Declaration of Competing Interest cementitious composites) produced with ground granulated blast furnace slag,
Cem. Concr. Res. 37 (7) (2007) 1096–1105.
The authors declare that they have no known competing financial [25] L.A. Qureshi, B. Ali, A. Ali, Combined effects of supplementary cementitious
materials (silica fume, GGBS, fly ash and rice husk ash) and steel fiber on the
interests or personal relationships that could have appeared to influence hardened properties of recycled aggregate concrete, Constr. Build. Mater. 263
the work reported in this paper. (2020), 120636.
[26] M. Papachristoforou, E. Anastasiou, I. Papayianni, Durability of steel fiber
reinforced concrete with coarse steel slag aggregates including performance at
Data availability elevated temperatures, Constr. Build. Mater. 262 (2020), 120569.
[27] Y. Choi, R.L. Yuan, Experimental relationship between splitting tensile strength
Data will be made available on request. and compressive strength of GFRC and PFRC, Cem. Concr. Res. 35 (8) (2005)
1587–1591.
[28] K. Vijai, R. Kumutha, B. Vishnuram, Investigation on properties of glass fiber
Acknowledgment reinforced fly ash based geopolymer concrete, Int. J. Earth Sci. Eng. 5 (4) (2012)
817–824.
[29] Y. Moodi, et al., Using response surface methodology and providing a modified
The Authors extend their appreciation to the Deanship of Scientific
model using whale algorithm for estimating the compressive strength of columns
Research at King Khalid University for funding this work through Large confined with FRP sheets, Constr. Build. Mater. 183 (2018) 163–170.
Groups Project under grant number RGP. 2/104/43. [30] T. Awolusi, et al., Application of response surface methodology: predicting and
optimizing the properties of concrete containing steel fibre extracted from waste
tires with limestone powder as filler, Case Stud. Constr. Mater. 10 (2019), e00212.

14
S.A. Raheel Shah et al. Construction and Building Materials 362 (2023) 129722

[31] K.E. Alyamac, E. Ghafari, R. Ince, Development of eco-efficient self-compacting [57] S.C. Ferreira, et al., Box-Behnken design: an alternative for the optimization of
concrete with waste marble powder using the response surface method, J. Cleaner analytical methods, Anal. Chim. Acta 597 (2) (2007) 179–186.
Prod. 144 (2017) 192–202. [58] E.K. Nambiar, K. Ramamurthy, Models relating mixture composition to the density
[32] B. Şimşek, et al., Analysis of the effects of dioctyl terephthalate obtained from and strength of foam concrete using response surface methodology, Cem. Concr.
polyethylene terephthalate wastes on concrete mortar: a response surface Compos. 28 (9) (2006) 752–760.
methodology based desirability function approach application, J. Cleaner Prod. [59] B. Şimşek, Y.T. İç, E.H. Şimşek, A RSM-based multi-response optimization
170 (2018) 437–445. application for determining optimal mix proportions of standard ready-mixed
[33] B.S. Mohammed, V.C. Khed, M.F. Nuruddin, Rubbercrete mixture optimization concrete, Arab. J. Sci. Eng. 41 (4) (2016) 1435–1450.
using response surface methodology, J. Cleaner Prod. 171 (2018) 1605–1621. [60] P. Mrudul, et al., Study on silica infused recycled aggregate concrete using design
[34] M. Tyagi, et al., Adsorptive removal of cyanide from coke oven wastewater onto of experiments, J. Eng. Sci. Technol. 12 (4) (2017) 958–971.
zero-valent iron: optimization through response surface methodology, isotherm [61] S. Vasudevan, V. Poornima, M. Balachandran, Influence of admixtures on
and kinetic studies, J. Cleaner Prod. 178 (2018) 398–407. properties of concrete and optimization using response surface methodology,
[35] G.D. Bowden, B.J. Pichler, A. Maurer, A design of experiments (DoE) approach Mater. Today:. Proc. 24 (2020) 650–661.
accelerates the optimization of copper-mediated 18F-fluorination reactions of [62] I. Abdulkadir, et al., Modelling and multi-objective optimization of the fresh and
arylstannanes, Sci. Rep. 9 (1) (2019) 1–10. mechanical properties of self-compacting high volume fly ash ECC (HVFA-ECC)
[36] Zahraee, S.M., et al. Teaching the Design of Experiment and Response Surface using response surface methodology (RSM), Case Stud. Constr. Mater. 14 (2021),
Methodology Using Paper Helicopter Experiment. in Proceedings of the The 4th e00525.
International Graduate Conference on Engineering Science & Humanity, Johor Bahru, [63] A. Pourkhorshidi, et al., Applicability of the standard specifications of ASTM C618
Johor, Malaysia. 2013. for evaluation of natural pozzolans, Cem. Concr. Compos. 32 (10) (2010) 794–800.
[37] B. Dejaegher, Y. Vander Heyden, Experimental designs and their recent advances in [64] ASTM, A., C332-17, Standard specification for lightweight aggregates for insulating
set-up, data interpretation, and analytical applications, J. Pharm. Biomed. Anal. 56 concrete. West Conshohocken: ASTM International, 2017.
(2) (2011) 141–158. [65] Designation, A., C33-03 “Standard Specification for Concrete Aggregate”. Annual
[38] B. Chong, et al., Properties of concrete with eggshell powder: a review, Phys. Book of ASTM Standard, American Society for Testing and Material, Philadelphia
Chem. Earth, Parts A/B/C 120 (2020), 102951. Pennsylvania, Section. 4: p. 10-16.
[39] C.B. Wei, et al., Properties of Concrete with Eggshell Powder and Tyre Rubber [66] A.-C.-I.-A. Committee, ACI 544.1 R-96: State-of-the-Art Report on Fiber Reinforced
Crumb. Key Engineering Materials, Trans Tech Publ, 2021. Concrete, ACI Committee, Detroit, 1996.
[40] S. Wu, et al., Predictive modeling of high-performance concrete with regression [67] Astm, C., 143, Standard Test Method for Slump of Hydraulic Cement Concrete. ASTM
analysis. In 2010 IEEE International Conference on Industrial EngIneerIng and International, 2003.
EngIneerIng Management, IEEE, 2010. [68] ASTM, A., C192/C192M-13a-Standard practice for making and curing concrete test
[41] B.W. Chong, et al., Design of experiment on concrete mechanical properties specimens in the laboratory. 2013, Technical report, ASTM International.
prediction: a critical review, Materials 14 (8) (2021) 1866. [69] Astm, C., Standard test method for density, absorption, and voids in hardened concrete.
[42] M. Tanco, et al., Is design of experiments really used? A survey of Basque C642-13, 2013.
industries, J. Eng. Des. 19 (5) (2008) 447–460. [70] Concrete, A.I.C.C.o. and C. Aggregates, Standard Test Method for Splitting Tensile
[43] F.A. Silva, et al., Use of nondestructive testing of ultrasound and artificial neural Strength of Cylindrical Concrete Specimens1. 2017: ASTM international.
networks to estimate compressive strength of concrete, Buildings 11 (2) (2021) 44. [71] Concrete, A.I.C.C.o. and C. Aggregates, Standard test method for compressive strength
[44] A. Poorarbabi, M. Ghasemi, M.A. Moghaddam, Concrete compressive strength of cylindrical concrete specimens. 2014: ASTM international.
prediction using non-destructive tests through response surface methodology, Ain [72] ASTM, C. Standard test method for flexural strength of concrete (using simple
Shams Eng. J. 11 (4) (2020) 939–949. beam with third-point loading). in American society for testing and materials.
[45] L.T. Ngo, Y.-R. Wang, Y.-M. Chen, Applying adaptive neural fuzzy inference system 2010.
to improve concrete strength estimation in ultrasonic pulse velocity tests, Adv. Civ. [73] D.C. Montgomery, Design and Analysis of Experiments, John wiley & sons, 2017.
Eng. 2018 (2018). [74] B.E. Jimma, P.R. Rangaraju, Chemical admixtures dose optimization in pervious
[46] L. Li, et al., Development of nano-silica treatment methods to enhance recycled concrete paste selection–A statistical approach, Constr. Build. Mater. 101 (2015)
aggregate concrete, Cem. Concr. Compos. 118 (2021), 103963. 1047–1058.
[47] G. Cibilakshmi, J. Jegan, A DOE approach to optimize the strength properties of [75] E. Ghafari, H. Costa, E. Júlio, RSM-based model to predict the performance of self-
concrete incorporated with different ratios of PVA fibre and nano-Fe2O3, Adv. compacting UHPC reinforced with hybrid steel micro-fibers, Constr. Build. Mater.
Compos. Lett. 29 (2020), 2633366X20913882. 66 (2014) 375–383.
[48] C.B. Wei, et al., Properties of Mortar with Waste Tyre Rubber as Partial Sand [76] A. Khodaii, H. Haghshenas, H.K. Tehrani, Effect of grading and lime content on
Replacement. Key Engineering Materials, Trans Tech Publ, 2021. HMA stripping using statistical methodology, Constr. Build. Mater. 34 (2012)
[49] A.N. Rizalman, C.C. Lee, Comparison of artificial neural network (ANN) and 131–135.
response surface methodology (RSM) in predicting the compressive strength of [77] A.I. Nassar, N. Thom, T. Parry, Optimizing the mix design of cold bitumen
POFA concrete, Appl. Model. Simul. 4 (2020) 210–216. emulsion mixtures using response surface methodology, Constr. Build. Mater. 104
[50] W.N.F.W. Hassan, et al., Mixture optimization of high-strength blended concrete (2016) 216–229.
using central composite design, Constr. Build. Mater. 243 (2020), 118251. [78] B.E. Achara, B.S. Mohammed, M. Liew, Bond behaviour of nano-silica-modified
[51] N.I. Rahim, et al., Effect of crumb rubber, fly ash, and nanosilica on the properties self-compacting engineered cementitious composite using response surface
of self-compacting concrete using response surface methodology, Materials 15 (4) methodology, Constr. Build. Mater. 224 (2019) 796–814.
(2022) 1501. [79] I. Abdulkadir, B. Mohammed, RSM study and analysis on the 6 months compressive
[52] W. Lokuge, et al., Design of fly ash geopolymer concrete mix proportions using strength development and shrinkage behavior of high volume fly ash ECC (HVFA-
Multivariate Adaptive Regression Spline model, Constr. Build. Mater. 166 (2018) ECC), Int. J. Adv. Res. Eng. Technol. (IJARET) 11 (9) (2020) 965–980.
472–481. [80] B.S. Mohammed, O.C. Fang, Assessing the properties of freshly mixed concrete
[53] N. Bradley, The Response Surface Methodology, Indiana University South Bend, containing paper-mill residuals and class F fly ash, J. Civ. Eng. Constr. Technol. 2
2007. (2) (2011) 17–26.
[54] Aydar, A.Y., Utilization of response surface methodology in optimization of extraction of [81] A.S.M. Akid, et al., Assessing the influence of fly ash and polypropylene fiber on
plant materials. Statistical approaches with emphasis on design of experiments fresh, mechanical and durability properties of concrete, J. King Saud Univ.-Eng.
applied to chemical processes, 2018: p. 157-169. Sci. (2021).
[55] G. Anitha, V. Pandey, Review on: Statistical designs and response surface [82] B. Ali, et al., A step towards durable, ductile and sustainable concrete:
methodology (RSM) as a tool for the optimization of HPLC Methods, Anal. Res 5 simultaneous incorporation of recycled aggregates, glass fiber and fly ash, Constr.
(2016) 552–569. Build. Mater. 251 (2020), 118980.
[56] S.L. Ferreira, et al., Doehlert matrix: a chemometric tool for analytical chemistry, [83] T. Akhtar, et al., Experimental investigation of eco-friendly high strength fiber-
Talanta 63 (4) (2004) 1061–1067. reinforced concrete developed with combined incorporation of tyre-steel fiber and
fly ash, Constr. Build. Mater. 314 (2022), 125626.

15