Journal of Building Engineering 36 (2021) 102101

Contents lists available at ScienceDirect

Journal of Building Engineering

journal homepage: http://www.elsevier.com/locate/jobe

Optimization of mixture proportions by statistical experimental design

using response surface method - A review
Zhiping Li a, Dagang Lu a, b, c, *, Xiaojian Gao a, b, **
a
School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China
b
Key Lab of Structures Dynamic Behavior and Control of the Ministry of Education, Harbin Institute of Technology, Harbin, 150090, China
c
Key Lab of Smart Prevention and Mitigation of Civil Engineering Disasters of the Ministry of Industry and Information Technology, Harbin Institute of Technology,
Harbin, 150090, China

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

Keywords: A comprehensive review of the statistical experimental optimization problem concerning the mixture design of
Experimental design optimization various cement-based materials is presented herein. This review summarizes and discusses over 80 applications
Supplementary cementitious materials (SCMs) of optimum design regarding the basic test information under response surface method (RSM), including in­
Response surface methodology (RSM)
fluence factor and corresponding response, statistical method, and coefficient of determination. The statistical
Sustainable concrete
Ultra-high performance concrete (UHPC)
experimental design reported in previous studies has shown that RSM is a sequential procedure to provide a
suitable approximation for the mixture optimization. Then, linear, quadratic and interactive relationships of the
statistical model can be evaluated available. Especially, the multi-objective optimization issues with multiple or
competing performance requirements for various cement-based materials have also been reported, by consid­
ering fluidity, strength development, environmental impact, cost and durability. Overall, the results from
existing publications have demonstrated that statistical inference and analysis of variance (ANOVA) are suitable
for mix proportion design and process optimization of cement-based materials. The W/B ratio and mixture
components are the prevalent factors in experimental design optimization, and then the fluidity and strength as
the most popularly used response. Thus, theoretical optimum mixture proportioning can be used to predict
valuable fresh and hardened properties. Finally, a critical discussion of the selection of design strategy, inde­
pendent factors and their responses, and the experimental region involved in statistical experimental design, is
provided. Based on this review, we conclude that the multi-objective optimization approaches need a further
systematic study, and further studies of sustainable concrete optimization are needed by comparing the different
chemical composition and particle characteristics.

1. Introduction can be classified as prescriptive and performance-based approaches [1].

Prescriptive-based methods are often stepwise selection to provide a
The cement-based materials are prepared by using various types and mixture for a particular application, thereby satisfying the current mix
quantities of individual constituents. These mixture proportions play an proportion design standards and specifications, such as JGJ 55 [2] for
important role in fresh- and hardened-state performance, such as concrete, JGJ/T 98 [3] for mortar, and JGJ/T 233 [4] for cement. The
fluidity, rheological properties, strength development and durability. main advantage of these methods is that the mixture proportion is
Therefore, many research studies have been dedicated to experimental provided by the national or industry standard solely, not entirely
optimization of cement and concrete mixtures. depending on personal experience and subjective decision.
Experimental design optimization is an adjustment process of Performance-based techniques emphasize no strict requirements on the
selecting the available proportion of raw materials to prepare a cement- type and quantities of components, but are designed with many labo­
based mixture that satisfies specifiable requirements for a particular ratory trial experiments (defined as trial-and-error method).
application. Generally, conventional optimization for mixture design Trial-and-error or single variable method suffers from an exponential

* Corresponding author. School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China.
** Corresponding author. School of Civil Engineering, Harbin Institute of Technology, Harbin, 150090, China.
E-mail addresses: [email protected] (D. Lu), [email protected] (X. Gao).

https://doi.org/10.1016/j.jobe.2020.102101
Received 24 September 2020; Received in revised form 17 November 2020; Accepted 14 December 2020
Available online 17 December 2020
2352-7102/© 2020 Elsevier Ltd. All rights reserved.
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

growth in experimental times when many test factors are considered as term, βi for the first-order terms, βii for the quadratic terms and βij for the
independent variables in the optimization process. Furthermore, binary-interaction terms. A polynomial function cannot be a suitable
detailed optimization designs of concrete mixtures are often time- and approximation for all independent variable spaces. However, they usu­
resource-intensive [1]. Response surface method (RSM) is a combination ally work comparatively well for a relatively small area [12].
of mathematical and statistical techniques that are widely used in the The main purpose of experimental optimization is to move quickly to
area of concrete preparation optimization, where some nonlinear factors the actual optimum by using a simple and economically experimental
of concrete are added to obtain an optimum domain [5]. This method is process [13]. The general flow chart of RSM for experimental design
especially suitable for multiple performance requirements of concrete, optimization can be summarized in Fig. 1. The design procedure by
such as ultra-high performance concrete (UHPC) [6–9]. Over the past using RSM consists of the following sequential steps: (1) defining inde­
decade, the statistical experimental design of cement-based materials pendent factors and desired responses, (2) selecting appropriate design
has gained increasing attention with the sustainable development of the strategy to fit the response surfaces, (3) confirming the fitted model by
concrete industry. Among these, lots of researchers have investigated using analysis of variance (ANOVA) and statistical inference, and (4)
the optimization of mixture proportions by using RSM. determining the optimum set of operating conditions.
Recently, the multiple response problem of cement-based materials
has been widely reported in previous experimental studies. The simul­
2.2. Designs of the first-order model
taneous optimization process of several responses can be classified into
two steps, as follows: (1) a fitting response surface model is established
Designs for fitting the first-order model are called first-order designs.
for every response, and (2) operating constraints optimized by all re­
The most widely used first-order designs are 2k factorial design, Plack­
sponses are identified or maintained in the desired region. Some related
ett–Burman design and simplex design [14]. Among these designs,
optimization methods, such as D-optimal design [10], overlay of the
simplex lattice design has obtained considerable attention in the
contour plots and constrained optimization, have been used in previous
experimental design optimization of cement-based materials, which are
studies. Overlaying contour plots work effectively for a small number of
described briefly in the following section.
design variables. If more than three independent factors exist, then this
Simplex lattice design is used to investigate the effects of the com­
method is ineffective because the two-dimensional contour plot cannot
ponents or ingredients of a mixture on the response variable; it is also
obtain the best view of the response surface. The two other approaches
referred to as the mixture experiment. In general, the key feature of the
can be used for cases with more variables.
given mixtures is that the volume or mass fractions of these components
This paper summarizes and discusses the main achievements
must sum to one. Furthermore, the response of the given mixture de­
including the applications of different RSMs and optimization method­
pends only on the relative fraction but not on the total amount of the
ologies in the experimental design of cement-based materials. This re­
mixture constituents [15]. For instance, if x1 , x2 , ..., xk represent the
view is organized as follows. The basic procedure and certain theoretical
proportions of k ingredients of the given mixture, then
models and its evaluation and validation are reviewed briefly in Section
2. Then, in Section 3, the typical applications of central composite 0 ≤ xi ≤ 1 (i = 1, 2, ..., k) , (3)
design (CCD) and other optimization designs are summarized and
investigated to measure the feasibility and validity of the selected RSM, ∑
k
and xi = 1. (4)
especially for the sustainable concrete application. Finally, in Section 4, i=1
several related problems for further promising applications of RSMs in
cement-based materials are discussed. Moreover, some addition boundary constraints are found on the
components, thereby limiting the available region of the ingredients
2. Theoretical basis of RSM for cement-based materials between the lower limit (Li ) and the upper limit (Ti ). The general form of
the mixture optimization could be expressed as follows:
2.1. General procedure using RSM in experimental design optimization 0 ≤ Li ≤ xi ≤ Ti ≤ 1 (i = 1, 2, ..., k) . (5)

RSM has been used for various issues in the experimental optimiza­ The main types of simplex lattice designs in previous articles are
tion of cement-based materials [11]. This method aims to optimize shown in Fig. 2. The points presented in Fig. 2 denote experimental runs,
mixture design to consider several attributes, involving workability, and the three vertices, midpoints of the sides and the overall centroid of
strength development, cost, durability and environmental impact. These the triangle represent the pure blends, binary blends and ternary blends,
features are achieved with sequential experimentation including factors respectively. The controversy of the simplex lattice design is that most
such as water–binder ratio (W/B), mixture constituent, the proportion of test runs emerge in the boundary of the optimized area. Simplex lattice
supplementary cementitious materials (SCMs), preparation conditions and simplex centroid design should be added with points in the internal
and curing environment. In general, if the response is well expressed by region, as shown in Ref. [16].
a linear model of the independent factors, then the first-order regression
model can be expressed as follows: 2.3. Designs of the second-order model
∑
k
Y = β0 + βi Xi + ε, (1) Designs for fitting the second-order model are called second-order
i=1 designs. Applications of CCD and Box–Behnken design (BBD) to
cement and concrete have become more increasingly popular over the
where Y represents the response variable conforming to the regression past few decades.
coefficients (β); Xi represent the independent variables; k is the number CCD include 2k factorial runs, 2k star runs and k0 runs (centre-point
of optimized variables; ε denotes the random error of the estimated replications, usually 3 ≤ k0 ≤ 5); it is a good alternative to the 3k full
response. If a curvature is found in the local experiments, then a second- factorial design because it provides comparable experimental results
order regression model can be given as follows: with a small number of tests [17]. Fig. 3 shows a CCD for the case of k =
∑
k ∑
k k ∑
∑ k 2 and k = 3. In general, CCD is developed in a manner of the sequential
Y = β0 + β i Xi + βii Xi2 + βij Xi Xj + ε, (2) experiment to investigate a first-order design, followed by adding axial
i=1 i=1 i=1 j>1
runs to fit the second-order model. The first-degree model is used to
obtain initial information on the experimental programs and to assess
where the regression coefficients are expressed as β0 for the intercept
the importance of the component of the given mixture. Then, the

2
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

Fig. 1. General flow chart of RSM in experimental design optimization.

Fig. 2. Simplex lattice designs for three-component mixture plans: (a) {2,3} lattice, (b) {3,3} lattice, and (c) simplex centroid.

Fig. 3. Central composite designs for (a) k = 2 variables and (b) k = 3 variables of experimental optimization (The red dot is the centre-point replication, generally,
3 ≤ k0 ≤ 5). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

quadratic terms are chosen to obtain additional information to deter­ BBD consists of 2k factorial three-level designs with incomplete block
mine the desired properties of the given constraints. The value of α and to afford as either rotatable or nearly rotatable properties and to avoid
k0 depend on the number of runs in the factorial region of the given the vertices of the cubic region, as shown geometrically in Fig. 4a. All
√̅̅̅
experiment to ensure that CCD can achieve either the orthogonality points of BBD located at a spherical region of radius 2, to avoid the
behaviour or uniform precision behaviour. upper and lower limits of the given constraints. In addition, this would

3
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

Fig. 4. Spherical designs for three variables: (a) Box-Behnken design, and (b) face-centred central composite design (The red dot is the centre-point replication,
generally, 3 ≤ k0 ≤ 5). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

be available for BBD when the extreme vertices are prohibitively responses simultaneously. Thus, this approach tends to satisfy each
expensive or impossible to complete owing to the constraints of the desirable response as soon as possible without excessively compro­
experimental conditions. Face-centred design (CCF) is a useful variation mising any performance specifications. In general, every response Yi is
of CCD, where α = 1. Fig. 4b shows the star points of CCF located at the transformed into an individual desirability function as:
centre of the surface of the cube region, instead of the spherical area as
0 ≤ di (Yi ) ≤ 1, (9)
in CCD. Using CCF often leads to a reasonable assessment of experi­
mental errors because of more centre runs. where the value of di (Yi ) ranges between 0 and 1. For the combination of
the single responses near to the target values, the value of di (Yi ) should
2.4. Evaluation and validation of the fitting model be close to 1. The composite desirability function D can be expressed as
follows:
ANOVA is most often used to validate the predictive ability of the
fitted model before prediction, to ensure that the mathematical model ∏
k
(10)
1 1
D = (d1 (Y1 ), d2 (Y2 )....dk (Yk ))k = di (Yi )k ,
provides an adequate approximation of the actual response behaviour. i=1
The ANOVA expressions for regression model assessment and validation
are summarized in Table 1. In general, the overall accuracy of the pre­ where k represents the total responses involved in the optimization
dicted model is often described by the coefficient of determination R2, process.
which is calculated as follows:
SSmod SSres 3. Literature survey of RSM in mixture design optimization
R2 = =1− . (6)
SStot SStot
In a review article published in 1999 [11], RSM is the first time
The value of R2 varies between 0 and 1. For the predicted model with
systematically discussed and compared in mixture design optimization
good accuracy, the value of R2 is close to 1. After considering the number of high-performance concrete, and the multi-objective optimization by
of model terms, a related statistic parameter of adjusted R2 can be ob­ using material science-based statistical models is also presented to pre­
tained, as follows: dict the concrete properties. Then, a comprehensive review of linear
MSres SSres /(k − p) combination, statistical models, artificial intelligence method, and
R2adj = 1 − =1− . (7) physics-based models was provided to optimize the design and propor­
MStot SStot /(k − 1)
tioning of the concrete mixture [1]. Based on the previously surveyed,
The value of R2adj decreases as statistically insignificant variables in this paper attempted to evaluate the advances in cement and concrete
the model increase. The differences between the predicted and the mixture optimization by using RSM over the past two decades. Symbols
actual values are defined as residual errors, which play a critical role in used in this review are listed in Appendix A. Applications of RSM of
evaluating the model accuracy. Another statistic used to measure the mixture optimization of cement-based materials are shown in Appendix
predictive ability of the model, is expressed as follows: B.
SSpre Since the experimental results of Appendix B were obtained by
R2pre = 1 − . (8) various characteristics of raw materials and under various preparing
SStot
conditions, this paper only collected the basic test information (influ­
The value of R2pre and R2adj should be within 0.2. ence factor and corresponding response, statistical method and coeffi­
Desirability function is another useful method to optimize multiple cient of determination) for further discussion. For more corresponding

Table 1
Basic structure of the ANOVA test in the RSM-based experimental design.
Source of variation Degrees of freedom Sum of squares Mean square F-value
∑k
Total corrected k-1 SStot = − y) 2
i=1 (yi
Model p-1 SSmod = SStot − SSres MSmod = SSmod /(p − 1) MSmod /MSres
∑
Residual k-p SSres = ki=1 (yi − ̂
y i )2 MSres = SSres /(k − p)
Lack of fit m− p SSlof = SSres − SSpe MSlof = SSlof /(m − p) MSlof /MSpe
∑m ∑ki
Pure error k− m SSpe = 2 MSpe = SSpe /(k − m)
i=1 i=1 (yij − yi )

Note: k = total number of experiments in the set; p = total number of parameters in the model; m = number of distinct level of factor combinations; ki = number of
replications of the ith level; Adapted from Ref. [18].

4
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

details of the tests, one can refer to the references. conducted by CCD combined with canonical analysis to maximize the
toughness and modulus of elasticity of the fibre metakaolin-based geo­
polymer. Zahid et al. [25] applied CCD technique to establish the effect
3.1. Optimization designs in cement-based materials applications of independent factors (NaOH molarity, NaOH–Na2SiO3 ratio and curing
temperature) to evaluate several responses (such as setting time,
Over the past two decades, many studies on cement-based materials modulus of elasticity, compressive strength, flexural strength, flexural
have focused on using RSM as a secondary analysis in multi-objective toughness and ductility index) of FA-based engineered geopolymer
optimization that can be achieved with a series of separate experi­ composite. CCD was used to confirm the optimal mixture parameter of
ments. This tool has been used successfully by previous researchers to alkali-activated slag mortar with the maximum flexural strength and
optimize fresh and hardened properties for cement and concrete fields. compressive strength, by considering the influence of usage of waste
However, mixture designs of some advanced cement-based materials are glass powder [26]. Revathi et al. [27] used CCD to establish the
always difficult to standardize and reproduce owing to lack of available regression model of three factors (modulus of sodium silicate, liquid–FA
guidelines [19]. Herein, the focus is on summarizing the RSM applica­ ratio and mineral admixture) and these interactions with mechanical
tions; and the existing methods, including CCD, BBD and CCF are dis­ strength with 15 experimental trials.
cussed in the following sub-sections. UHPC is characterized by dense microstructures that possess ultra-
high mechanical, ductility and durability performance. The optimiza­
3.1.1. Central composite design (CCD) tion approach often starts with a combination of particle packing and
CCD is the most commonly used method of experimental optimiza­ statistical design method to obtain a mixture proportion of UHPC. The
tion in the cement-based material field, which is used for fitting the effects of three factors (distribution modulus, SCM and W/B ratio) on the
second-order model. CCD is often used as a screening design to deter­ rheological and mechanical properties of strain-hardening UHPC were
mine the critical factors and their interactions. As an example, optimized by combining CCD and modified Andreasen and Andersen
Mohammend et al. [20] used CCD in modelling the fresh and hardened particle packing model [28]. Sun et al. [29] used CCD to evaluate the
performance of rubbercrete mixture to develop available mix propor­ effect of porous aggregate and shrinkage-reducing admixture on
tion. Two factors (W/B and crumb rubber) with five levels were selected autogenous shrinkage of UHPC by using the modified dense
and 45 runs were performed in this research. The response surface with particle-packing model. Wang et al. [30] used the modified Andreasen
three slump levels for compressive strength is presented in Fig. 5. and Andersen particle packing models to achieve a compacting binder
Based on the previous CCD applications shown in Appendix B. the matrix of eco-friendly UHPC. Then, CCF was applied by maximum use of
existing studies can be classified into three groups of research charac­ combined micro-coral sand and coral sand. The developed eco-friendly
teristics, as follows: (1) optimizing the raw materials and preparation
condition to achieve the optimal performance or the most economical
mix design results, (2) adding new components to investigate the per­
formance range, and (3) combining with other modelling techniques
and then evaluating the feasibility. Especially, geopolymer/alkali-
activated materials have acquired wide attention as promising con­
struction and maintenance materials due to their superior performance
[21]. Venkatesan et al. [22] applied CCD to determine the optimal
conditions of geopolymer concrete by using partial replacement of fine
aggregate with waste foundry sand and fly ash (FA). Then, D-optimal
design was used to conduct the proportion of mixture components to
acquire the desired responses. Mohammed et al. [23] optimized the
experimental parameters of ingredients, such as anhydrous sodium
metasilicate, ground granulated blast-furnace slag (GGBS) and FA to
produce cast in situ alkali-activated binders. The optimal condition was
provided using CCF to evaluate the three responses (split tensile
strength, compressive strength and water absorption). Da Silva Alves
et al. [24] investigated the effect of sisal fibre, activator–metakaolin
Fig. 6. Ecological evaluation of eco-friendly ultra-high performance concrete
mass ratio, and curing time on toughness and modulus of elasticity. In
with environmental impact indicator. Adapted from Ref. [30].
addition, the optimization of the experimental parameters was

Fig. 5. Response surface with three slump levels for compressive strength: (a) low slump, (b) medium slump and (c) high slump. Adapted from Ref. [20].

5
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

UHPC was evaluated by using the environmental impact indicator with

the radar map (Fig. 6). On the other hand, the optimum design of UHPC
usually diminishes the energy consumption and emissions of CO2 with
the reduction of cement content. Ferdosian and Camoes [31] used CCD
to investigate the effect of SF, ultra-fine FA and sand of UHPC on fluidity
and compressive strength. Then, a multi-objective optimization was
conducted, and the cost and environmental influences were optimized
by the overall desirability (Fig. 7). Furthermore, CCD shows an excellent
fitting effect on other experiments [32–36].

3.1.2. Other optimization designs

Factorial design is another method to optimize the mixture propor­
tion of cement-based materials. It is often classified into two categories:
full factorial design and fractional factorial design. Long et al. [37]
applied fractional factorial design to build statistical models to investi­
gate the influence of mixture proportion and raw material properties on
workability, strength development, and visco-elastic performance of
self-compacting concrete. Then eleven additional SCC mixtures were Fig. 8. Additional SCC mixtures used to validate the derived statistical models.
used to validate the statistical models for fresh properties. Including Adapted from Ref. [37].
eight runs within the range of the factorial design to develop the wide
range, three central points were used to evaluate the error in the 90%
confidence limit (Fig. 8). Jiao et al. [38] applied simplex centroid design
to optimize the paste consisting of cement, FA and slag for a given
strength grade, then optimized the paste, fine aggregate and coarse
aggregate based on rheological properties of SCC, and at last, over­
lapped the contour plots to acquire the multiple performance re­
quirements (Fig. 9).
As for mix optimization of geopolymer/alkali-activated materials, Li
et al. [39] proposed a mixture proportioning methodology according to
the performance requirements of alkali-activated concrete and used the
simplex centroid design for optimizing three types of aggregates to
obtain the optimized bulk density. Mermerdas et al. [40] applied sim­
plex lattice design to optimize three independent variables (curing age,
curing temperature, and volume of binder) of geopolymer mortars and
to maximize the compressive strength of FA and GGBS. Shi et al. [16]
used simplex lattice design to correlate the ingredients of ternary cement
blends (cement, slag and FA) on ASR expansion with only seven
experimental trials. Then, the ternary contour diagram was used to
analyse the composition effect on ternary composite blends (Fig. 10). Li
et al. [41] used BBD to investigate the effect of the degree of sol ratio, the Fig. 9. Optimization of cementitious materials composition by overlapping the
content of slag and age on fracture toughness and their interaction on contour plots. Adapted from Ref. [38].
fracture properties before and after freeze-thaw resistance of alkali–slag
concrete. Bektas et al. [42] used BBD to investigate the influence of three solution–slag ratio, sand ratio and slag content and their interaction on
critical mix factors (alkali content, W/B ratio and ground clay brick the freeze-thaw cycles of the alkali–slag concrete. Then, the predicted
content) in three-levels to measure four responses (alkali–slag reaction model was built to evaluate the effect of air bubble characteristic on
expansion, Fc, Ft and modulus of elasticity) in two replicates of 15 runs. freeze-thaw cycles in cold regions.
Cai et al. [43] applied BBD to analyse the influence of activator As for mix optimization of UHPC, Ghafari et al. [15] present an ac­
curate analytical approach based on simplex lattice design to optimize
the component of UHPC. The main strategy of this method can be
described in seven steps, as follows: (1) constructing the main optimum
objective to obtain the highest compressive strength, acceptable scope of
workability and economical cost of raw materials; (2) selecting the
mixture design method, where D-optimal techniques are recommended;
(3) defining the constraint bounds of mixture components, parameters
and these variation ranges in the defined experiments; (4) developing
the design matrix based on the D-optimal mixture trials; (5) collecting
the experimental data; (6) building the analytical model to predict the
properties of UHPC; (7) optimizing the mixture proportion of UHPC to
satisfy the desirable value of the response variable. Soliman and
Tagnit-Hamou [44] proposed a modified approach combining a
full-factorial design approach and particle-packing model to optimize
UHPC as follows: (1) particle packing of aggregates, and (2) building the
optimized model by investigating the combined effect of W/B ratio and
high range water-reducing admixture.
Fig. 7. Desirability of ultra-high performance concrete for its main variable
constituents. Adapted from Ref. [31].

6
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

the concrete industry [22,26,30,31,33,40,45,65,85,94,98,101,102],

which are summarized in Appendix B. However, further study of the
sustainable concrete application is needed by comparing the different
chemical composition and particle characteristics. De Brito et al. [46]
presented a ternary phase diagram to provide the chemical composition
of various binder types from 81 publications. As shown in Fig. 11, the
chemical composition of industrial wastes are diversities and signifi­
cantly determined on the source of the raw materials, and it cannot
directly be replaced with the equivalent mass of cement because of the
amorphous particles is different from the cement. Furthermore, certain
experimental studies of sustainable optimization were focused on cost
and environmental impact [53,76,84,88,91].

4. Summary and discussion

Based on the 80 applications of RSMs in the existing literature and its

analysis, some critical information of statistical models to optimize the
mixture proportioning of cement-based materials, including RSM
method, test specimen, factors (independent variables) and its responses
(dependent variables), were collected. The summary is listed in Ap­
pendix B.

Fig. 10. Ternary contour diagram of composition design for composite cement.
4.1. Selection of design strategy
Adapted from Ref. [16].

As it is shown in Appendix B, CCD is the most popular method for

3.2. Optimization designs for sustainable concrete applications
mixture proportioning optimization in cement-based materials. CCD
comprises a two-level factorial design, centre point and a star design in
Some industrial wastes are blended with cement clinker to prepare
which test points with a distance α from the centre point. CCD provides a
Portland cement or used as concrete constituents for sustainable appli­
considerable high efficiency with up to six factors if all optimizations are
cation, which are widely investigated by academics and engineers.
carried in parallel instead of sequentially experiments. It is often used to
Existing experimental design of industrial wastes applications has
be regarded as a better alternative of the full factorial design because it
attempted to explore the alternative of SCM and their performances in
can offer similar results with a smaller number of experiments [17]. In

Fig. 11. CaO–SiO2–Al2O3 ternary phase diagram for the cement blends of sustainable concrete. Adapted from Ref. [46].

7
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

addition, both linear and quadratic regression models are permitted to such as artificial neural networks [48–50], fuzzy classification [51].
be determined by these design strategies, and the interactive effects of These combined approaches have been demonstrated experimentally by
various independent factors and critical points (minimum, maximum providing well precision in data learning and prediction. Although these
and saddle points) can be evaluated. solutions have been used in several other fields, little research has been
Another popular method is simplex design, including the simplex- reported of these applications in mixture proportion optimization.
lattice design and simplex-centroid design. The factors of these strate­
gies are the component of a mixture, and the factor levels are not in­ 4.4. Current challenges for the applications of sustainable concrete
dependent. If there are three ingredients of the mixture, the constrained
experimental region is constructed to a trilinear coordinate system as In general, reducing the environmental impact and resources con­
shown in Refs. [16,38]. Each of the three sides of the triangle represents sumption of sustainable concrete is related to replace cement clinker
a mixture that has only two components, and the missing component with solid wastes, which contains many ingredients and are always
labelled on the opposite corner. subject to multi-performance requirements. Statistical experimental
However, BBD has not been employed as extensive as the above- design has been developed to optimize the mixture proportion of sus­
mentioned strategies in cement-based materials. While in BBD strat­ tainable concrete. However, target performance during the optimization
√̅̅̅
egy, all points located at a spherical region of radius 2. And also, BBD process may be mutually exclusive, which leads to numerous redundant
does not contain any corner points of the cubic region to avoid the upper works. The combined desirability of various weighted values and their
and lower limits of the given constraints. This would be available for corresponding solutions has been developed a multi-objective optimi­
BBD when the extreme vertices are prohibitively expensive or impos­ zation [5,31,40,84]. The simultaneous nonlinear optimization with
sible to complete owing to the constraints of the experimental desirability function should be further studied in the future.
conditions. So far, many multi-variable problems for sustainable concrete opti­
Anyway, the prevalence of CCD usage in cement-based materials is mization have become increasingly common. It is difficult to coordinate
partly attributed to that it is easy to follow the other researcher’s steps. the raw materials properties and their dosage are often lacks a theo­
As for geopolymer/alkali-activated materials and UHPC with various retical basis. Furthermore, little attention has been focused on the in­
ingredients and several performance requirements, D-optimal design or dependent factors and their interactions of sustainable concrete
Doehlert design [47] or BBD might be a better beneficial strategy. applications.

5. Conclusions and prospective

4.2. Selection of factors and responses
Based on the review and discussions in this paper, the conclusions
The W/B ratio and mixture components are the prevalent factors in
can be drawn as below:
experimental design optimization, then fluidity and strength as the most
popularly used response. Each response of mixture optimization is often
(1) The RSM is a sequential procedure to provide a suitable
expressed with a polynomial function of factors such as W/B ratio,
approximation for the fitting functional models between various
cement content, admixture dosage and SCM replacement. Changing of
independent factors and their responses. Then, linear, quadratic
W/B ratio leads to a remarkable variation of concrete properties. In
and interactive relationships of these models can be evaluated.
general, selection of the factors and its level should be according to the
And also, the minimum, the maximum and the saddle points of
preliminary tests or practical experience and not depending on the re­
optimization region can be evaluated available. So, many appli­
searcher’s convenience. Furthermore, performing heavy single-variable
cations in modelling a variety of cement-based materials field
studies with the purpose to optimize with three or more factors should
have been attempted, as shown in Appendix B.
be avoided [14].
(2) CCD is the most commonly used method in cement and concrete
Recently, the multiple response problem of cement-based materials
mixture design. Most studies considered four or less independent
has become a concern. Jiao et al. [38] overlapped several critical con­
variables. The W/B ratio and mixture components are the prev­
tour lines of each response to acquire the multiple performance re­
alent factors in experimental design optimization, and then the
quirements. Ferdosian and Camoes [31] employed D-optimal design to
fluidity and strength as the most popularly used response. How­
develop a combined desirability with different important weights for
ever, D-optimal design or BBD or Doehlert design might be better
their corresponding solutions. These multi-objective optimization ap­
for geopolymer/alkali-activated materials and UHPC with
proaches need a further systematic study.
various ingredients and several performance requirements.
Obviously, the choice of the variable levels in the optimization
(3) The choice of factors and their levels is very important for
process is more important than the design itself. Every level of RSM must
experimental optimization by using RSM. Each level should be
be appropriate and provide valuable information. When the design
appropriate and provide valuable information, and it is necessary
points are too close together, it will not result in the obvious influence of
to assure the responses within the acceptable region of the opti­
the corresponding response. On the contrary, if design points are at the
mum value.
extreme point of a reasonable region, the responses are often hard to
(4) The multiple or competing performance requirement of mixture
adopt.
design has become a concern. The in-depth investigations are
needed to combine and compare with other modelling tech­
4.3. Selection of experimental domain niques. Previous studies investigated the combination of D-
optimal method and particle packing models. However, artificial
Although RSM has many outstanding characteristics and has been neural network and fuzzy logic for modelling mixture optimiza­
widely used for mixture design and process optimization of various ex­ tion issues may be a promising research direction.
periments, the fitting models can be only suitable for the experimental (5) Further study of sustainable concrete optimization is needed by
domain and are not accurate for extrapolation. In addition, discrete comparing the different chemical composition and particle
variables cannot be selected for experimental optimization. For characteristics. However, target performance during the optimi­
example, a specific type of SCM or any other mixture components cannot zation process may be mutually exclusive, which leads to
be considered in the mixture optimization problems. numerous redundant works. So, the simultaneous nonlinear
In order to overcome the defects of RSM strategy, some researchers optimization with desirability function should be further studied
attempt to integrate of RSM with other machine learning algorithms, in the future.

8
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

(6) Although plenty of studies have been reported in the previous interests or personal relationships that could have appeared to influence
literature with the laboratory experiment in the cement-based the work reported in this paper.
materials field, little attention has focused on the applications
in engineering practice. Thus, more attempts relating to the Acknowledgements
practical project by using RSM are needed.
The authors would like to thank for the financial support from the
Declaration of competing interest National Science Foundation of China (Grant Nos. 51678209,
41861134010).
The authors declare that they have no known competing financial

Appendix A. List of symbols

Symbol Description

RSM response surface method

CCD central composite design
BBD Box- Behnken design
CCF face-centred central composite design
ANOVA analysis of variance
UHPC ultra-high performance concrete
HPC high performance concrete
SCC self-compacting concrete
ASR alkali-silica reaction
W/B Water-binder ratio
SCM supplementary cementitious material
SF silica fume
FA fly ash
QP quartz powder
GBFS Granulated blast furnace slag
Vca volume of coarse aggregate
Vfa volume of fine aggregate
SP Superplasticizer
Fc 28-day compressive strength
Ft flexure strength

Appendix B. Applications of RSM for mixture optimization of cement-based materials

RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.

experiments

CCD SCC W/B ratio, Binder, SP, fluidity, rheology, Fc slump = 0.95, Fc = 0.83 15 2000 [52]
Vca
Factorial design Mortar SF, FA, GBFS SP, setting time, drying 13/26* 2002 [53]
shrinkage,fc,cost
CCD SCC cement, limestone filler, fluidity, Fc slump = 0.98, Fc = 0.97 21 2002 [54]
SP, W/B ratio
BBD mineral aggregate six types of silica sand void content 0.96 54 2003 [55]
Full factorial steel fibre aspect ratio, volume of fracture energy, characteristic 10 2004 [56]
design reinforced steel fibre length
concrete
CCD SCC pulverised fuel ash, SP, fluidity, rheology, slump = 0.99, rheology = 0.98, 21 2004 [57]
cement, W/B ratio segregation ratio, Fc segregation ratio = 0.99, Fc = 0.99
CCD foam concrete filler-cement ratio, FA, Fc, dry density Fc = 0.958, dry density = 0.987 20 2006 [58]
foam volume
BBD bridge deck SF, FA, slag Fc, Ft, chloride permeability, 15 2006 [59]
overlay concrete abrasion resistance
Bucher–Bourgund frost-resistant W/B ratio, entrained air residual strain 7 2007 [60]
design concrete pore, number of cycles
simplex centroid blends of industrial red clay, granite waste, water absorption, shrinkage, 10/40 2008 [61]
design wastes kaolin waste modulus of rupture
simplex centroid HPC cement, FA, GBFS, SP, fluidity, Fc 78 2009 [49]
design Vca, Vfa
CCD SCC cement, W/B ratio, FA, Fc, modulus of elasticity Fc = 0.823 31 2009 [62]
SP
simplex centroid drilling fluid Bentonite, low molar apparent viscosity, plastic 10/30 2010 [63]
design carboxymethyl viscosity
cellulose, high molar
carboxymethyl cellulose
BBD high-strength temperature, binder specific gravity, water 18 2011 [64]
lightweight content, binder type absorption, crushing strength
concrete
fractional factorial SCC fluidity, Fc, shrinkage, creep 19 2012 [37]
design
(continued on next page)

9
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

(continued )
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments

binder, W/B, binder

type, SP, sand-aggregate
ratio
(continued on next page)

RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.

experiments
CCD recycled masonry cement, degree of moisture content, dry density 13 2012 [65]
and concrete compaction
CCD High flowing GBFS, FA, W/B ratio, SP fluidity, Fc, Durability slump = 0.9841, Fc = 0.9315, 21 2012 [66]
concrete carbonaion = 0.9717
CCD SCC cement, W/B ratio, FA, fluidity, Fc, modulus of fluidity = 0.905, Fc = 0.920, 31 2012 [67]
SP elasticity modulus of elasticity = 0.818
CCD oil well cement SP, SCM, temperature yield stress, plastic viscosity 40 2012 [68]
slurry
BBD alkali–slag solution-slag ratio, slag, Air bubble spacing 17 2013 [43]
concrete sand coefficient, Air bubble
specific surface area, Grades
of freeze-thaw
Resistance
Factorial design pervious concrete W/B ratio, cement, Vca fresh density, hardened fresh density = 0.79, hardened 13 2013 [69]
density, void ratio, Fc density = 0.97, void ratio = 0.98,
Fc = 0.87
fractional factorial Concrete W/B, cement, fineness Fc 46/92 2013 [70]
design modulus of aggregate,
SP
CCD warm mix asphalt binder, resident, air void, bulk specific gravity, air void = 0.94, bulk specific 20 2013 [71]
compaction voids filled with asphalt gravity = 0.96, stability = 0.77,
temperature binder, stability, fluidity fluidity = 0.89
CCD UHPC cement, SF fluidity, Fc slump = 0.9949, Fc = 0.9913 13 2013 [36]
simplex lattice HPC cement, grinded dune fluidity, Fc slump = 0.78, Fc = 0.91 21 2014 [72]
design sand, limestone filler
2014
Factorial design Concrete FA, metakaolin, testing Fc, chloride permeability, 9 2014 [73]
age sorptivity, water absorption
Full factorial Concrete binder, W/B, Vfa/Vca Fc 0.8 27 2014 [74]
design
BBD silicate cement water-soluble polymer, Fc 27 2014 [75]
chemical additive, SP
(continued on next page)
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments
BBD alkali–slag alkali, W/B ratio, expansion, Fc, Ft, modulus of expansion = 0.886, Fc = 0.889, Ft 30 2014 [42]
concrete ground clay elasticity = 0.900, modulus of elasticity =
0.847
CCD normal weight W/B ratio, Vca, SP fluidity, Fc, splitting tensile slump = 0.992, Fc = 0.837, 20 2014 [76]
concrete strength, cost splitting tensile strength = 0.825,
cost = 1.000
CCD self-compacting steel fibre, powder- Ft, fluidity Ft = 0.91, fluidity = 0.92 20 2014 [35]
UHPC aggregate ratio
CCD modified asphalt asphalt, polyethene fluidity, void, stability, bulk slump = 0.9880, void = 0.9980, 13 2015 [77]
mixture terephthalate modifier specific gravity stability = 0.9853, bulk specific
gravity = 0.9883
CCD SCC binder, W/B ratio, SP fluidity, Fc, filling capacity, slump = 0.96, Fc = 0.86, filling 20 2015 [78]
sieve segregation capacity = 0.95, sieve segregation
= 0.94
CCD SCC binder, W/B ratio, SCM fluidity, Fc, segregation factor 27 2015 [79]
CCF cement paste W/B ratio, FA/B ratio, fluidity, Fc slump = 0.855, Fc = 0.852 20 2015 [80]
nano-iron oxide-to-
binder
simplex lattice UHPC cement, sand, SF, QP, fluidity, Ft slump = 0.74, Ft = 0.90 53 2015 [15]
design SP, steel fibre
BBD pervious concrete three admixture paste thickness, slump, film paste thickness = 0.92, slump = 18 2015 [81]
drying time 0.89, film drying time = 0.69
BBD alkali–slag sol ratio, slag, age on initiation fracture toughness, 17 2015 [41]
concrete fracture toughness unstable fracture toughness,
crack mouth opening
displacement, critical
effective crack
simplex lattice Mortar cement, SF, nano-silica fluidity, Fc, Ft, splitting 13 2016 [82]
design strength, density, absorption,
capillary water
simplex lattice alkali-activated cement, FA, slag ASR expansion 17 2016 [16]
design cement
(continued on next page)
(continued on next page)

10
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

(continued )
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments

RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.

experiments
CCD Concrete crumb rubber, Fc, water absorption, unit Fc = 0.9703, water absorption = 9 2016 [83]
metakaolin weight 0.8751, unit weight = 0.8321
CCD SCC W/B ratio, cement, Vfa, fluidity, Fc, cost fluidity = 0.9604, Fc = 0.9547, 52 2016 [84]
FA, SP cost = 1
CCD UHPC SF, SP, fibre, cement, flexural toughness flexural toughness = 0.85 45 2016 [34]
W/B
simplex lattice geopolymer mortar FA, GBFS binder, curing time, curing 7/14 2017 [40]
design temperature, Fc
simplex lattice UHPC cement, SF, QP, quartz fluidity, Fc, air void slump = 0.99, Fc = 0.99, air void = 10 2017 [44]
design sand 0.80
CCF eco-friendly UHPC micro-coral sand, coral Fc 0.97 10 2017 [30]
sand
CCD SCC W/B, marble powder- fluidity, Fc 33 2017 [85]
cement ratio
CCF Mortar clinker, FA, debit Fc 0.98 15 2017 [86]
grinding agent
CCD warm mix asphalt compaction adhesion failure, direct 11/22 2017 [87]
temperature, test tensile strength, fracture
temperature energy, broken aggregate
CCD high-strength SCC W/B, cement, FA, SP, Fc, fluidity, cost Fc = 0.955, fluidity = 0.960, cost 52 2017 [88]
Vfa =1
CCF UHPC QP, quartz sand, water Fc, Ft Fc = 0.984, Ft = 0.830 16 2017 [32]
curing
CCD UHPC SF, sand, ultra-fine fly fluidity, Fc slump = 0.9596, fc = 0.9568 28 2017 [31]
ash
simplex centroid low carbon cement, mineral fluidity, Fc, hydration heat, 7 2018 [89]
design cementitious admixture, hydrated porosity, non-evaporable
material lime water
simplex centroid alkali-activated gravel, sand bulk density 7 2018 [39]
design concrete
simplex centroid Concrete Vca, Vfa, paste, cement, rheology, Fc 16 2018 [38]
design FA, slag
BBD grout material cement, FA, microsilica, fluidity, Fc, Ft, shrinkage slump = 0.9647, Fc = 0.9810, Ft = 16 2018 [90]
metakaolin 0.7966, shrinkage = 0.8053
BBD foamed concrete cement, foam Fc, dry density, cost 15 2018 [91]
(continued on next page)
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments
simplex centroid SCC SP, stone powder, fluidity, Fc 42 2018 [92]
design gravel, sand, cement
CCD geopolymeric modulus of sodium Fc 0.9736 15 2018 [27]
binder silicate, liquid, mineral
admixture
CCD alkali-activated Na2O, glass powder Fc, Ft Fc = 0.9678, Ft = 0.9754 13 2018 [26]
slag mortar
CCD geopolymer NaOH molarity, Fc, elastic modulus, Ft, Fc = 0.9951, elastic modulus = 20 2018 [25]
composite Na2SiO3, curing flexural toughness, ductility 0.9977, Ft = 0.9924, flexural
temperature index, tensile first crack toughness = 0.9837, ductility
strength, ultimate tensile index = 0.9731, tensile first crack
strength, tensile strain strength = 0.9876, ultimate tensile
capacity strength = 0.9791, tensile strain
capacity = 0.9850
CCD rubbercrete W/B ratio, crumb fluidity, unit weight, void, Fc 45 2018 [20]
mixture rubber
CCD self-consolidating SF, slag, SP, W/B ratio fluidity, Fc, segregation slump = 0.9589, Fc = 0.8561, 30 2018 [93]
mortar segregation = 0.8141
CCD normal concrete two types of plastic fluidity, Fc slump = 0.8198, Fc = 0.9750 13 2018 [94]
waste aggregate
CCF polymer nanosilica additive, complex modulus, phase complex modulus = 0.9995, phase 13 2019 [95]
nanocomposite- temperature angle, viscosity angle = 0.9989, viscosity = 0.9995
modified asphalt
Full factorial SCC cement, FA, W/B, SP fluidity, Fc slump = 0.9319, Fc = 0.9343 18 2019 [51]
design
CCD recycled concrete cement, slump, recycled Fc 0.9881 17 2019 [96]
aggregate coarse aggregate
CCD Geopolymer sisal fibre, activator, Fc, toughness, modulus of 18 2019 [24]
curing time elasticity
(continued on next page)
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments
CCRD fibre reinforced aspect ratio, cement, W/ fluidity, Fc, Ft, split tensile 20 2019 [97]
concrete B ratio strength, water absorption
(continued on next page)

11
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

(continued )
RSM method Specimen Factors Reponse(s) R2 Number of Year Ref.
experiments

slump = 0.98, Fc = 0.95, Ft = 0.98,

split tensile strength = 0.93, water
absorption = 0.8640
CCF alkali-activated slag, anhydrous sodium Fc, Ft, water absorption Fc = 0.9856, Ft = 0.9913, water 15 2019 [23]
paste metasilicate activator absorption = 0.8994
CCD geopolymer Vfa, FA, waste foundry Fc 0.99 14 2019 [22]
concrete sand
CCD UHPC porous aggregate, autogenous shrinkage 0.9296 11 2019 [29]
shrinkage reducing
admixture
CCF UHPC nano-silica, waste glass fluidity, Fc, drying shrinkage slump = 0.93, Fc = 0.98, drying 10 2019 [33]
powder shrinkage = 0.96
CCD eco-efficient SCC limestone powder, FA, fluidity, Fc slump = 0.9679, Fc = 0.9695 20 2019 [98]
SP
BBD pervious concrete aggregate size bulk density, apparent 24 2020 [99]
density, void
BBD blended paste cement, SF, FA, QP fluidity, rheology, hydration slump = 0.9613, rheology = 16 2020 [5]
heat, Fc, drying shrinkage 0.9818, hydration heat = 0.9975,
Fc = 0.9955, drying shrinkage =
0.9459
CCF Concrete manufactured sand, Fc, permeability coefficient, permeability coefficient = 0.9745, 17 2020 [100]
metakaolin, waste paper sorptivity
sludge ash
CCD cementitious cement, curing time Fc 0.97 14 2020 [101]
composites
CCD strain-hardening W/B ratio, SCM, fluidity, rheology, Fc, fracture 20/60 2020 [28]
UHPC distribution modulus toughness
CCD geopolymer mortar molarity, binder, Fc, drying shrinkage Fc = 0.9063, drying shrinkage = 27 2020 [102]
sodium silicate to 0.9296
sodium hydroxide ratio
Note: * two groups, 11 mixtures for each one.

References [14] A. Khuri, S. Mukhopadhyay, Response surface methodology, Wiley

interdisciplinary reviews: Comput. Stat. 2 (2010) 128–149, https://doi.org/
10.1002/wics.73.
[1] M.A. DeRousseau, J.R. Kasprzyk, W.V. Srubar, Computational design
[15] E. Ghafari, H. Costa, E. Júlio, Statistical mixture design approach for eco-efficient
optimization of concrete mixtures: a review, Cement Concr. Res. 109 (2018)
UHPC, Cement Concr. Compos. 55 (2015) 17–25, https://doi.org/10.1016/j.
42–53, https://doi.org/10.1016/j.cemconres.2018.04.007.
cemconcomp.2014.07.016.
[2] Chinese National Institute of Standardization, Specification for Mix Proportion
[16] Z. Shi, C. Shi, R. Zhao, D. Wang, F. He, Factorial design method for designing
Design of Ordinary Concrete, vols. 55–2011, JGJ, Beijing, 2011.
ternary composite cements to mitigate ASR expansion, J. Mater. Civ. Eng. 28
[3] Chinese National Institute of Standardization, Specification for Mix Proportion
(2016) 4016064, https://doi.org/10.1061/(ASCE)MT.1943-5533.0001568.
Design of Masonry Mortar, JGJ/T 98-2010, Beijing, 2010.
[17] K.M. Sharif, M.M. Rahman, J. Azmir, A. Mohamed, M.H.A. Jahurul, F. Sahena, I.
[4] Chinese National Institute of Standardization, Specification for Mix Proportion
S.M. Zaidul, Experimental design of supercritical fluid extraction - a review,
Design of Cement Soil, JGJ/T 233-2011, Beijing, 2011.
J. Food Eng. 124 (2014) 105–116, https://doi.org/10.1016/j.
[5] Z. Li, D. Lu, X. Gao, Multi-objective optimization of gap-graded cement paste
jfoodeng.2013.10.003.
blended with supplementary cementitious materials using response surface
[18] S. Bouzalakos, A.W.L. Dudeney, B.K.C. Chan, Formulating and optimising the
methodology, Construct. Build. Mater. 248 (2020) 118552, https://doi.org/
compressive strength of controlled low-strength materials containing mine
10.1016/j.conbuildmat.2020.118552.
tailings by mixture design and response surface methods, Miner. Eng. 53 (2013)
[6] Q. Chen, H. Wang, H. Li, Z. Jiang, H. Zhu, J.W. Ju, Z. Yan, Multiscale modelling
48–56, https://doi.org/10.1016/j.mineng.2013.07.007.
for the ultra-high performance concrete: from hydration kinetics to macroscopic
[19] J.S.J. van Deventer, J.L. Provis, P. Duxson, Technical and commercial progress in
elastic moduli, Construct. Build. Mater. 247 (2020), https://doi.org/10.1016/j.
the adoption of geopolymer cement, Miner. Eng. 29 (2012) 89–104, https://doi.
conbuildmat.2020.118541.
org/10.1016/j.mineng.2011.09.009.
[7] C. Shi, Z. Wu, J. Xiao, D. Wang, Z. Huang, Z. Fang, A review on ultra high
[20] B.S. Mohammed, V.C. Khed, M.F. Nuruddin, Rubbercrete mixture optimization
performance concrete: Part I. Raw materials and mixture design, Construct. Build.
using response surface methodology, J. Clean. Prod. 171 (2018) 1605–1621,
Mater. 101 (2015) 741–751, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.jclepro.2017.10.102.
conbuildmat.2015.10.088.
[21] P. Zhang, K. Wang, Q. Li, J. Wang, Y. Ling, Fabrication and engineering properties
[8] H. Wei, A. Zhou, T. Liu, D. Zou, H. Jian, Dynamic and environmental
of concretes based on geopolymers/alkali-activated binders - a review, J. Clean.
performance of eco-friendly ultra-high performance concrete containing waste
Prod. 258 (2020) 120896, https://doi.org/10.1016/j.jclepro.2020.120896.
cathode ray tube glass as a substitution of river sand, Resour. Conserv. Recycl.
[22] M. Venkatesan, Q. Zaib, I.H. Shah, H.S. Park, Optimum utilization of waste
162 (2020) 105021, https://doi.org/10.1016/j.resconrec.2020.105021.
foundry sand and fly ash for geopolymer concrete synthesis using D-optimal
[9] Z. Mo, X. Gao, A. Su, Mechanical performances and microstructures of metakaolin
mixture design of experiments, Resour. Conserv. Recycl. 148 (2019) 114–123,
contained UHPC matrix under steam curing conditions, Construct. Build. Mater.
https://doi.org/10.1016/j.resconrec.2019.05.008.
(2020) 121112, https://doi.org/10.1016/j.conbuildmat.2020.121112.
[23] B.S. Mohammed, S. Haruna, M. Mubarak Bn Abdul Wahab, M.S. Liew,
[10] J.H. Son, J.W. Baek, A.E. Sy Choi, H.S. Park, Thiomer solidification of an ASR
Optimization and characterization of cast in-situ alkali-activated pastes by
bottom ash: optimization based on compressive strength and the characterization
response surface methodology, Construct. Build. Mater. 225 (2019) 776–787,
of heavy metal leaching, J. Clean. Prod. 166 (2017) 12–20, https://doi.org/
https://doi.org/10.1016/j.conbuildmat.2019.07.267.
10.1016/j.jclepro.2017.07.113.
[24] L.C. Da Silva Alves, R.A. Dos Reis Ferreira, L. Bellini Machado, L.A. de Castro
[11] M. Simon, K. Snyder, G. Frohnsdorff, Advances in Concrete Mixture Optimization,
Motta, Optimization of metakaolin-based geopolymer reinforced with sisal fibers
Concrete Durability and Repair Technology Conference, Scotland, 1999,
using response surface methology, Ind. Crop. Prod. 139 (2019) 111551, https://
pp. 21–32.
doi.org/10.1016/j.indcrop.2019.111551.
[12] D.C. Montgomery, Design and analysis of experiments (6nd ed), J. Am. Stat.
[25] M. Zahid, N. Shafiq, M.H. Isa, L. Gil, Statistical modeling and mix design
Assoc. 16 (2) (2000).
optimization of fly ash based engineered geopolymer composite using response
[13] S. Karimifard, M.R. Alavi Moghaddam, Application of response surface
surface methodology, J. Clean. Prod. 194 (2018) 483–498, https://doi.org/
methodology in physicochemical removal of dyes from wastewater: a critical
10.1016/j.jclepro.2018.05.158.
review, Sci. Total Environ. 640–641 (2018) 772–797, https://doi.org/10.1016/j.
[26] L. Zhang, Y. Yue, Influence of waste glass powder usage on the properties of
scitotenv.2018.05.355.
alkali-activated slag mortars based on response surface methodology, Construct.

12
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

Build. Mater. 181 (2018) 527–534, https://doi.org/10.1016/j. [51] S. Selvaraj, S. Selvaraj, S. Sivaraman, S. Sivaraman, Prediction model for
conbuildmat.2018.06.040. optimized self-compacting concrete with fly ash using response surface method
[27] T. Revathi, R. Jeyalakshmi, N.P. Rajamane, Geopolymeric binder: the effect of based on fuzzy classification, Neural Comput. Appl. 31 (2019) 1365–1373,
silica fume addition on Fly activation by using response surface methodology, https://doi.org/10.1007/s00521-018-3575-1.
Mater. Today: Proceedings 5 (2018) 8727–8734, https://doi.org/10.1016/j. [52] K.H. Khayat, A. Ghezal, M.S. Hadriche, Utility of statistical models in
matpr.2017.12.299. proportioning self-consolidating concrete, Mater. Struct. 33 (2000) 338–344,
[28] K. Ragalwar, W.F. Heard, B.A. Williams, R. Ranade, Significance of the particle https://doi.org/10.1007/BF02479705.
size distribution modulus for strain-hardening-ultra-high performance concrete [53] M.L. Nehdi, J. Summer, Optimization of ternary cementitious mortar blends using
(SH-UHPC) matrix design, Construct. Build. Mater. 234 (2020) 117423, https:// factorial experimental plans, Mater. Struct. 35 (2002) 495–503, https://doi.org/
doi.org/10.1016/j.conbuildmat.2019.117423. 10.1007/BF02483137.
[29] Y. Sun, R. Yu, Z. Shui, X. Wang, D. Qian, B. Rao, J. Huang, Y. He, Understanding [54] A. Ghezal, K.H. Khayat, Optimizing self-consolidating concrete with limestone
the porous aggregates carrier effect on reducing autogenous shrinkage of Ultra- filler by using statistical factorial design methods, ACI Mater. J. 99 (2002)
High Performance Concrete (UHPC) based on response surface method, 264–272.
Construct. Build. Mater. 222 (2019) 130–141, https://doi.org/10.1016/j. [55] M. Muthukumar, D. Mohan, M. Rajendran, Optimization of mix proportions of
conbuildmat.2019.06.151. mineral aggregates using Box Behnken design of experiments, Cement Concr.
[30] X. Wang, R. Yu, Z. Shui, Q. Song, Z. Zhang, Mix design and characteristics Compos. 25 (2003) 751–758, https://doi.org/10.1016/S0958-9465(02)00116-6.
evaluation of an eco-friendly Ultra-High Performance Concrete incorporating [56] F. Bayramov, C. Taşdemir, M.A. Taşdemir, Optimisation of steel fibre reinforced
recycled coral based materials, J. Clean. Prod. 165 (2017) 70–80, https://doi. concretes by means of statistical response surface method, Cement Concr.
org/10.1016/j.jclepro.2017.07.096. Compos. 26 (2004) 665–675, https://doi.org/10.1016/S0958-9465(03)00161-6.
[31] I. Ferdosian, A. Camoes, Eco-efficient ultra-high performance concrete [57] M. Sonebi, Medium strength self-compacting concrete containing fly ash:
development by means of response surface methodology, Cement Concr. Compos. modelling using factorial experimental plans, Cement Concr. Res. 34 (2004)
84 (2017) 146–156, https://doi.org/10.1016/j.cemconcomp.2017.08.019. 1199–1208, https://doi.org/10.1016/j.cemconres.2003.12.022.
[32] M.A. Mosaberpanah, O. Eren, Effect of quartz powder, quartz sand and water [58] E.K.K. Nambiar, K. Ramamurthy, Models relating mixture composition to the
curing regimes on mechanical properties of UHPC using response surface density and strength of foam concrete using response surface methodology,
modelling, Advances in Concrete Construction 5 (2017) 481–492, https://doi. Cement Concr. Compos. 28 (2006) 752–760, https://doi.org/10.1016/j.
org/10.12989/acc.2017.5.5.481. cemconcomp.2006.06.001.
[33] M.A. Mosaberpanah, O. Eren, A.R. Tarassoly, The effect of nano-silica and waste [59] J.P. Won, J.M. Seo, C.G. Park, J.H. Kim, Statistical optimisation and durability
glass powder on mechanical, rheological, and shrinkage properties of UHPC using characteristics of bridge deck overlay concrete, Mag. Concr. Res. 58 (2006)
response surface methodology, Journal of Materials Research and Technology 8 601–608, https://doi.org/10.1680/macr.2006.58.9.601.
(2019) 804–811, https://doi.org/10.1016/j.jmrt.2018.06.011. [60] T. Cho, Prediction of cyclic freeze–thaw damage in concrete structures based on
[34] M.A. Mosabepranah, O. Eren, Statistical flexural toughness modeling of ultra high response surface method, Construct. Build. Mater. 21 (2007) 2031–2040, https://
performance concrete using response surface method, Comput. Concr. 17 (2016) doi.org/10.1016/j.conbuildmat.2007.04.018.
477–488, https://doi.org/10.12989/cac.2016.17.4.477. [61] R.R. Menezes, H.G.M. Neto, L.N.L. Santana, H.L. Lira, H.S. Ferreira, G.A. Neves,
[35] E. Ghafari, H. Costa, E. Júlio, RSM-based model to predict the performance of Optimization of wastes content in ceramic tiles using statistical design of mixture
self-compacting UHPC reinforced with hybrid steel micro-fibers, Construct. Build. experiments, J. Eur. Ceram. Soc. 28 (2008) 3027–3039, https://doi.org/10.1016/
Mater. 66 (2014) 375–383, https://doi.org/10.1016/j.conbuildmat.2014.05.064. j.jeurceramsoc.2008.05.007.
[36] M.A.A. Aldahdooh, N.M. Bunnori, M.A.M. Johari, Evaluation of ultra-high- [62] A.N.S. Al Qadi, K. Nasharuddi, H. Al-Mattarn, Q.N.S. AL-Kadi, Statistical models
performance-fiber reinforced concrete binder content using the response surface for hardened properties of Self-Compacting concrete, Am. J. Eng. Appl. Sci. 2
method, Mater. Des. 52 (2013) 957–965, https://doi.org/10.1016/j. (2009) 764–770, https://doi.org/10.3844/ajeassp.2009.764.770.
matdes.2013.06.034. [63] R.R. Menezes, L.N. Marques, L.A. Campos, H.S. Ferreira, L.N.L. Santana, G.
[37] W. Long, G. Lemieux, S. Hwang, K.H. Khayat, Statistical models to predict fresh A. Neves, Use of statistical design to study the influence of CMC on the
and hardened properties of self-consolidating concrete, Mater. Struct. 45 (2012) rheological properties of bentonite dispersions for water-based drilling fluids,
1035–1052, https://doi.org/10.1617/s11527-011-9815-9. Appl. Clay Sci. 49 (2010) 13–20, https://doi.org/10.1016/j.clay.2010.03.013.
[38] D. Jiao, C. Shi, Q. Yuan, X. An, Y. Liu, Mixture design of concrete using simplex [64] N.U. Kockal, T. Ozturan, Optimization of properties of fly ash aggregates for high-
centroid design method, Cement Concr. Compos. 89 (2018) 76–88, https://doi. strength lightweight concrete production, Mater. Des. 32 (2011) 3586–3593,
org/10.1016/j.cemconcomp.2018.03.001. https://doi.org/10.1016/j.matdes.2011.02.028.
[39] N. Li, C. Shi, Z. Zhang, D. Zhu, H. Hwang, Y. Zhu, T. Sun, A mixture proportioning [65] D.X. Xuan, L.J.M. Houben, A.A.A. Molenaar, Z.H. Shui, Mixture optimization of
method for the development of performance-based alkali-activated slag-based cement treated demolition waste with recycled masonry and concrete, Mater.
concrete, Cement Concr. Compos. 93 (2018) 163–174, https://doi.org/10.1016/j. Struct. 45 (2012) 143–151, https://doi.org/10.1617/s11527-011-9756-3.
cemconcomp.2018.07.009. [66] M. Li, Y. Chen, Y. Chan, L.H. Vihn, A study of statistical models application for
[40] K. Mermerdas, Z. Algin, S.M. Oleiwi, D.E. Nassani, Optimization of lightweight mixture of high-flowing concrete, Journal of Marine Science and Technology-
GGBFS and FA geopolymer mortars by response surface method, Construct. Build. Taiwan 20 (2012) 325–335.
Mater. 139 (2017) 159–171, https://doi.org/10.1016/j. [67] A.N.S. Alqadi, K.N.B. Mustapha, S. Naganathan, Q.N.S. Al-Kadi, Uses of central
conbuildmat.2017.02.050. composite design and surface response to evaluate the influence of constituent
[41] Q. Li, L. Cai, Y. Fu, H. Wang, Y. Zou, Fracture properties and response surface materials on fresh and hardened properties of self-compacting concrete, KSCE
methodology model of alkali-slag concrete under freeze–thaw cycles, Construct. Journal of Civil Engineering 16 (2012) 407–416, https://doi.org/10.1007/
Build. Mater. 93 (2015) 620–626, https://doi.org/10.1016/j. s12205-012-1308-z.
conbuildmat.2015.06.037. [68] A. Shahriar, M.L. Nehdi, Optimization of rheological properties of oil well cement
[42] F. Bektas, B.A. Bektas, Analyzing mix parameters in ASR concrete using response slurries using experimental design, Mater. Struct. 45 (2012) 1403–1423, https://
surface methodology, Construct. Build. Mater. 66 (2014) 299–305, https://doi. doi.org/10.1617/s11527-012-9841-2.
org/10.1016/j.conbuildmat.2014.05.055. [69] M. Sonebi, M.T. Bassuoni, Investigating the effect of mixture design parameters
[43] L. Cai, H. Wang, Y. Fu, Freeze–thaw resistance of alkali–slag concrete based on on pervious concrete by statistical modelling, Construct. Build. Mater. 38 (2013)
response surface methodology, Construct. Build. Mater. 49 (2013) 70–76, https:// 147–154, https://doi.org/10.1016/j.conbuildmat.2012.07.044.
doi.org/10.1016/j.conbuildmat.2013.07.045. [70] M.T. Cihan, A. Güner, N. Yüzer, Response surfaces for compressive strength of
[44] N.A. Soliman, A. Tagnit-Hamou, Using particle packing and statistical approach concrete, Construct. Build. Mater. 40 (2013) 763–774, https://doi.org/10.1016/j.
to optimize Eco-Efficient Ultra-High-Performance concrete, ACI Mater. J. 114 conbuildmat.2012.11.048.
(2017) 847–858, https://doi.org/10.14359/51701001. [71] M.O. Hamzah, B. Golchin, C.T. Tye, Determination of the optimum binder content
[45] L. Qin, X. Gao, A. Su, Q. Li, Effect of carbonation curing on sulfate resistance of of warm mix asphalt incorporating Rediset using response surface method,
cement-coal gangue paste, J. Clean. Prod. 278 (2021) 123897, https://doi.org/ Construct. Build. Mater. 47 (2013) 1328–1336, https://doi.org/10.1016/j.
10.1016/j.jclepro.2020.123897. conbuildmat.2013.06.023.
[46] J. de Brito, R. Kurda, The past and future of sustainable concrete: a critical review [72] R. Zaitri, M. Bederina, T. Bouziani, Z. Makhloufi, M. Hadjoudja, Development of
and new strategies on cement-based materials, J. Clean. Prod. (2020) 123558, high performances concrete based on the addition of grinded dune sand and
https://doi.org/10.1016/j.jclepro.2020.123558. limestone rock using the.mixture design modelling approach, Construct. Build.
[47] D.B. Hibbert, Experimental design in chromatography: a tutorial review, Journal Mater. 60 (2014) 8–16, https://doi.org/10.1016/j.conbuildmat.2014.02.062.
of Chromatography B-Analytical Technologies in the Biomedical and Life [73] E. Güneyisi, M. Gesoğlu, Z. Algın, K. Mermerdaş, Optimization of concrete
Sciences 910 (2012) 2–13, https://doi.org/10.1016/j.jchromb.2012.01.020. mixture with hybrid blends of metakaolin and fly ash using response surface
[48] W.E. Elemam, A.H. Abdelraheem, M.G. Mahdy, A.M. Tahwia, Optimizing fresh method, Composites Part B 60 (2014) 707–715, https://doi.org/10.1016/j.
properties and compressive strength of self-consolidating concrete, Construct. compositesb.2014.01.017.
Build. Mater. 249 (2020), https://doi.org/10.1016/j.conbuildmat.2020.118781. [74] S. Ahmad, S.A. Alghamdi, A statistical approach to optimizing concrete mixture
[49] I. Yeh, Optimization of concrete mix proportioning using a flattened design, TheScientificWorldJOURNAL (2014) 561537–561539, https://doi.org/
simplex–centroid mixture design and neural networks, Eng. Comput. 25 (2009) 10.1155/2014/561539, 2014.
179–190, https://doi.org/10.1007/s00366-008-0113-2. [75] J. Zheng, G. Shao, X. Shen, Synergistic interactions of chemical additives on the
[50] Z.M. Sbartai, S. Laurens, S.M. Elachachi, C. Payan, Concrete properties evaluation strength development of silicate cement by a box-behnken model optimization,
by statistical fusion of NDT techniques, Construct. Build. Mater. 37 (2012) J. Appl. Polym. Sci. 131 (2014), https://doi.org/10.1002/app.41071.
943–950, https://doi.org/10.1016/j.conbuildmat.2012.09.064. [76] B. Simsek, Y.T. Ic, E.H. Simsek, A.B. Guvenc, Development of a graphical user
interface for determining the optimal mixture parameters of normal weight

13
Z. Li et al. Journal of Building Engineering 36 (2021) 102101

concretes: a response surface methodology based quadratic programming Build. Mater. 189 (2018) 170–180, https://doi.org/10.1016/j.
approach, Chemometr. Intell. Lab. Syst. 136 (2014) 1–9, https://doi.org/ conbuildmat.2018.08.156.
10.1016/j.chemolab.2014.05.001. [91] S. Asadzadeh, S. Khoshbayan, Multi-objective optimization of influential factors
[77] T. Baghaee Moghaddam, M. Soltani, M.R. Karim, H. Baaj, Optimization of asphalt on production process of foamed concrete using Box-Behnken approach,
and modifier contents for polyethylene terephthalate modified asphalt mixtures Construct. Build. Mater. 170 (2018) 101–110, https://doi.org/10.1016/j.
using response surface methodology, Measurement 74 (2015) 159–169, https:// conbuildmat.2018.02.189.
doi.org/10.1016/j.measurement.2015.07.012. [92] A. Habibi, J. Ghomashi, Development of an optimum mix design method for self-
[78] A. Lotfy, K.M.A. Hossain, M. Lachemi, Statistical models for the development of compacting concrete based on experimental results, Construct. Build. Mater. 168
optimized furnace slag lightweight aggregate self-consolidating concrete, Cement (2018) 113–123, https://doi.org/10.1016/j.conbuildmat.2018.02.113.
Concr. Compos. 55 (2015) 169–185, https://doi.org/10.1016/j. [93] M. Aziminezhad, M. Mandikhani, M.M. Memarpour, RSM-based modeling and
cemconcomp.2014.09.009. optimization of self-consolidating mortar to predict acceptable ranges of
[79] A.A. Abouhussien, A.A.A. Hassan, Optimizing the durability and service life of rheological properties, Construct. Build. Mater. 189 (2018) 1200–1213, https://
self-consolidating concrete containing metakaolin using statistical analysis, doi.org/10.1016/j.conbuildmat.2018.09.019.
Construct. Build. Mater. 76 (2015) 297–306, https://doi.org/10.1016/j. [94] M.A.A. Aldahdooh, A. Jamrah, A. Alnuaimi, M.I. Martini, M.S.R. Ahmed, A.S.
conbuildmat.2014.12.010. R. Ahmed, Influence of various plastics-waste aggregates on properties of normal
[80] L. Soto-Perez, V. Lopez, S.S. Hwang, Response Surface Methodology to optimize concrete, Journal of Building Engineering 17 (2018) 13–22, https://doi.org/
the cement paste mix design: time-dependent contribution of fly ash and nano- 10.1016/j.jobe.2018.01.014.
iron oxide as admixtures, Mater. Des. 86 (2015) 22–29, https://doi.org/10.1016/ [95] N. Bala, I. Kamaruddin, M. Napiah, M.H. Sutanto, Polymer Nanocomposite-
j.matdes.2015.07.049. Modified asphalt: Characterisation and optimisation using response surface
[81] B.E. Jimma, P.R. Rangaraju, Chemical admixtures dose optimization in pervious methodology, Arabian J. Sci. Eng. 44 (2019) 4233–4243, https://doi.org/
concrete paste selection – a statistical approach, Construct. Build. Mater. 101 10.1007/s13369-018-3377-x.
(2015) 1047–1058, https://doi.org/10.1016/j.conbuildmat.2015.10.003. [96] A. Hammoudi, K. Moussaceb, C. Belebchouche, F. Dahmoune, Comparison of
[82] M.F.A. El Hameed, M.F. Ghazy, M.A.A.A. Elaty, Cement mortar with nanosilica: artificial neural network (ANN) and response surface methodology (RSM)
experiments with mixture design method, ACI Mater. J. 113 (2016) 43–53, prediction in compressive strength of recycled concrete aggregates, Construct.
https://doi.org/10.14359/51688632. Build. Mater. 209 (2019) 425–436, https://doi.org/10.1016/j.
[83] O. Rezaifar, M. Hasanzadeh, M. Gholhaki, Concrete made with hybrid blends of conbuildmat.2019.03.119.
crumb rubber and metakaolin: optimization using Response Surface Method, [97] T.F. Awolusi, O.L. Oke, O.O. Akinkurolere, A.O. Sojobi, Application of response
Construct. Build. Mater. 123 (2016) 59–68, https://doi.org/10.1016/j. surface methodology: Predicting and optimizing the properties of concrete
conbuildmat.2016.06.047. containing steel fibre extracted from waste tires with limestone powder as filler,
[84] A. Khan, J. Do, D. Kim, Cost effective optimal mix proportioning of high strength Case Studies in Construction Materials 10 (2019) e212, https://doi.org/10.1016/
self compacting concrete using response surface methodology, Comput. Concr. 17 j.cscm.2018.e00212.
(2016) 629–648, https://doi.org/10.12989/cac.2016.17.5.629. [98] M.K. Mohammed, A.I. Al-Hadithi, M.H. Mohammed, Production and optimization
[85] K.E. Alyamac, E. Ghafari, R. Ince, Development of eco-efficient self-compacting of eco-efficient self compacting concrete SCC with limestone and PET, Construct.
concrete with waste marble powder using the response surface method, J. Clean. Build. Mater. 197 (2019) 734–746, https://doi.org/10.1016/j.
Prod. 144 (2017) 192–202, https://doi.org/10.1016/j.jclepro.2016.12.156. conbuildmat.2018.11.189.
[86] N.H. Mtarfi, Z. Rais, M. Taleb, K.M. Kada, Effect of fly ash and grading agent on [99] Q. Zhang, X. Feng, X. Chen, K. Lu, Mix design for recycled aggregate pervious
the properties of mortar using response surface methodology, Journal of Building concrete based on response surface methodology, Construct. Build. Mater. 259
Engineering 9 (2017) 109–116, https://doi.org/10.1016/j.jobe.2016.12.004. (2020) 119776, https://doi.org/10.1016/j.conbuildmat.2020.119776.
[87] M.O. Hamzah, S.Y. Teh, B. Golchin, J. Voskuilen, Use of imaging technique and [100] S. Vasudevan, V. Poornima, M. Balachandran, Influence of admixtures on
direct tensile test to evaluate moisture damage properties of warm mix asphalt properties of concrete and optimization using response surface methodology,
using response surface method, Construct. Build. Mater. 132 (2017) 323–334, Mater. Today: Proceedings 24 (2020) 650–661, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.conbuildmat.2016.11.092. matpr.2020.04.319.
[88] A. Khan, J. Do, D. Kim, Experimental optimization of High-Strength Self- [101] C. Yan, H. Zhao, J. Zhang, S. Liu, Z. Yang, The cementitious composites using
Compacting concrete based on D-Optimal design, J. Construct. Eng. Manag. 143 calcium silicate slag as partial cement, J. Clean. Prod. (2020) 256, https://doi.
(2017), https://doi.org/10.1061/(ASCE)CO.1943-7862.0001230. org/10.1016/j.jclepro.2020.120514.
[89] M. Wu, Y. Zhang, G. Liu, Z. Wu, Y. Yang, W. Sun, Experimental study on the [102] K. Mermerdaş, Z. Algın, Ş. Ekmen, Experimental assessment and optimization of
performance of lime-based low carbon cementitious materials, Construct. Build. mix parameters of fly ash-based lightweight geopolymer mortar with respect to
Mater. 168 (2018) 780–793, https://doi.org/10.1016/j. shrinkage and strength, Journal of Building Engineering 31 (2020) 101351,
conbuildmat.2018.02.156. https://doi.org/10.1016/j.jobe.2020.101351.
[90] M. Shamsuddoha, G. Hüsken, W. Schmidt, H. Kühne, M. Baeßler, Ternary mix
design of grout material for structural repair using statistical tools, Construct.

14