Construction and Building Materials 170 (2018) 101–110

Contents lists available at ScienceDirect

Construction and Building Materials

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

Multi-objective optimization of influential factors on production process

of foamed concrete using Box-Behnken approach
Shervin Asadzadeh ⇑, Soheil Khoshbayan
Department of Industrial Engineering, North-Tehran Branch, Islamic Azad University, Tehran, Iran

h i g h l i g h t s

Box-Behnken design was applied to multi-objective optimization of foamed concrete.

Empirical models were derived to describe the relationship between factors and responses.

Maximum strength and minimum cost in the specified density ranges was considered.

Determining precise mix-designs based on specified density-strength was addressed.

The manufacturer can reach the best density and strength in line with minimum cost.

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

Article history: Foamed concrete is a type of lightweight concrete produced from cement, water, foam and sand; how-
Received 19 June 2017 ever, no sand is required for making low density foamed concrete. Foam is generated from mixing of
Received in revised form 23 February 2018 water and foaming agent under high pressured air. The two most important outputs of foamed concrete
Accepted 25 February 2018
production process are dry density and compressive strength in 28 days after the time of mixing which
are considered as the physical-mechanical properties. Furthermore, the cost of materials is a very critical
issue. The aim of this paper is to optimize the ratio of three effective factors for foamed concrete produc-
Keywords:
tion including cement mass, water and foam volume in order to achieve minimum density and maximum
Foamed concrete
Optimization
compressive strength along with minimum cost. To this end, the Box-Behnken design of response surface
Box-Behnken design (BBD) methodology has been used for optimization purposes. 15 experimental samples were prepared and
Response surface methodology (RSM) MINITAB statistical software was employed to analyze and determine the mix-design of materials in dif-
ferent ranges of densities based on their applications. The result reveals that Box-Behnken design leads to
optimized values of foamed concrete production outputs. Finally, the optimum values of each input fac-
tors for some critical densities and their corresponding compressive strengths were obtained.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction
In the process of concrete production, most significant input
variables are the quantities and ratios of main materials due to
their critical roles in the final production features. Once the ratios
of materials are optimized, the output variables including mechan-
ical and physical properties as well as cost will improve, which
boosts the quality of the finished product. Desirable values of these
mechanical and physical properties of concrete depend on the
intended place in the building, the structure type, the direction
and magnitude of different forces, exerted on it. Although other
factors including the type of used materials, curing conditions,
facilities together with equipment, environment, and the knowl-
⇑ Corresponding author.
E-mail address: Sh_asadzadeh@iau-tnb.ac.ir (S. Asadzadeh).
edge as well as the skill of engineers are important, the accurate
ratios of main contents should be determined more precisely. Thus,
the optimization of main materials and their mix-design has been
one of the most challenging tasks for civil engineers. The mix-
design of concrete is currently determined based on the producer’s
experience and speculation. This trial and error approach, is time-
consuming and inaccurate which does not rest on scientific fact. In
this research, non-autoclaved aerated concrete or foamed concrete,
which is one type of lightweight concrete, is chosen as a case study.
Foamed concrete is the result of mixing the foam obtained from
combining foam agent and water in foam generator with the mix-
ture of cement, water and sand. There is no standard procedure to
prepare foamed concrete. Due to the fact that applications of this
type of concrete have been increased over the last recent years,
improving the methodology of its preparation is an urgent need.
Compressive strength is a very critical characteristic of concrete

https://doi.org/10.1016/j.conbuildmat.2018.02.189
0950-0618/Ó 2018 Elsevier Ltd. All rights reserved.
102 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110
and the quality of concrete and its structure are all in direct rela-
tionship. The compressive strength could be considered as a repre-
sentative for all properties of concrete such as permeability,
durability and abrasion resistance [1]. An ideal foamed concrete
should have high compressive strength and low density and pro-
duction cost. Box-Behnken Design (BBD), as one of the crucial tech-
niques in Response Surface Method (RSM), and MINITAB software
introduce a statistical, systematic and precise method for the opti-
mization and determination of foamed concrete’s mix-design.
Therefore, in this model, input variables are considered as the con-
tent ratio of foamed concrete in volume 1 m3 comprising of
cement, water and foam while output variables are considered
compressive strength, volumetric density and production (mate-
rial) cost.
Moreover, DOE-based statistical methods have been imple-
mented to optimize different types of concrete characteristics.
For instance, RSM technique and design-expert software were
applied to create a relationship between input factors and
responses of fly-ash lightweight concrete’s production [2]. High
strength self-compacting concrete (HSSCC) samples were obtained
in a multi-objective approach by the Taguchi method [3]. Taguchi
method was applied to optimize mechanical and physical proper-
ties of pervious concrete [4]. Furthermore, the properties of
Alkali-slag concrete (ASC) under freeze-thaw cycles were opti-
mized using Box-Behnken design (BBD) [5]. Similarly, the effect
of four factors on the compressive strength and abrasion resistance
of Alkali-Activated slag (AAS) concrete was evaluated by the Tagu-
chi method [6]. Likewise, Taguchi design of experiments was used
for analyzing the effect of main factors on the compressive
strength and ultrasonic pulse velocity of geo-polymer concrete
production in high temperature [7]. To our best knowledge, no
research has been done dealing with multi-objective optimization
production with the aid of BBD. Thus, the aim of this paper is to
optimize and determine influential factors on foamed concrete
production process. Although the Taguchi model is a suitable
method to optimize factors, RSM approaches are suitable for fac-
tors with continuous values, simple to train and learn, and more
accurate to be presented. Moreover, the RSM set of methods make
possible to differentiate between whether the related equation is
linear or whether it appears quadratic. In addition, in order to
determine triple factors based on input responses, BBD method is
a unique solution.
The rest of the paper has been organized as follows: Section 2
presents the concept of RSM and its well-known design called
BBD. Section 3 elaborates on the experimented procedure. The
result and the optimum values of factors are given in Section 4.
Final remarks and conclusion are provided in Section 5.
2. Theoretical basis
In many cases, organized experiments are needed to create
some changes in system’s controllable parameters in order to ana-
lyze the changes in the process outputs. The most important object
of design of experiment (DOE) is the determination of effective fac-
tors in the way that outputs are close to their nominal values. Since
there is no proved and scientific relationship between factors and
responses in each problem, the empirical models are required to
describe this relationship. To this end, a regression model is sug-
gested to justify response changes based on input values. The sur-
face of this equation in k-dimensional space shows the order of the
equation. In First-order models, there is no sign of curvature and
the equation is written as below:
X
k X
k X
k
y ¼ b0 þ bi X i þ bij X i X j þ e0
i¼1 i¼1 j¼1
where bi and bij are regression coefficients, X i represents the
factors and X i X j demonstrates interaction between factors and
e0 is a measurement error in the process. However, in most
cases, a curvature is observed in the surface. Therefore, a sec-
ond or higher order model would be suitable for predicting the
surface curve:
X
k X
k X
k X
k
Y ¼ b0 þ bi X i þ bii X 2i þ bij X i X J þ e0 ð2Þ
i¼1 i¼1 i¼1 j¼1
Response surface methodology (RSM) is a technique that pro-
vides methods to determine the curve of this surface. Box-
Behnken design was suggested to present a second-order model
in a k-dimensional space. For instance, in case of three factors,
the edge points in the middle of cube sides along with center
points are tested [8–10] (see Fig. 1).
In this study, Box-Behnken model with three factors is used as
statistical approach to predict a multi-objective surface which
describes the relationship between three factors of foamed con-
crete production, and compressive strength, density and cost are
listed as three responses. It should be noted that the criterion
used for evaluating the response optimization is the desirability
function. If the goal is to optimize a single response, the term
individual desirability function is used, while in case of more than
one response, composite desirability function is commonly
applied. The definition of desirability function is different depend-
ing on whether the response should be maximized, minimized or
whether the target value should be determined. For instance, if
maximizing a response is desired, the desirability is calculated
as below [10]:
di ¼ 0 Y i < Li
 r
ðY i  Li Þ i
di ¼ Li < Y i < T i ð3Þ
ðT i  Li Þ
di ¼ 1 Y i < T i
where T i is the target value which is selected large enough for the
ith response, Li is the lowest acceptable value for the ith response,
di is called desirability and ri is the weight of desirability function
corresponding to the ith response.
If the importance of all responses is equal, the composite (over-
all) desirability is calculated via the below equation:
1=n
D ¼ ðd1  d2  . . . dn Þ ð4Þ
where the n is the number of responses.
ð1Þ
Fig. 1. Box-Behnken design for k = 3.
 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110 103

3. Experimental procedure mcement

þ V water þ V foam 6 1000 ð7Þ
3:16
3.1. Determination of variables’ value based on BBD where the mass of cement is shown by mcement . Inequality (5) would
convert to an equation when the foamed concrete is only produced
According to foam concrete production process and obtaining from cement, water and foam. In such state, foamed concrete would
experts’ points of view, the minimum and maximum values for be so light (density less than 400 kg/m3) because no sand is applied.
the three factors were determined as follow: The difference of the above three factors from 1000 dm3 shows the
amount of sand. In case all three factors have their maximum val-

Cement mass (X1) from 300 kg to 450 kg ues, the aggregation of their volume is equal to:

Water volume (X2) from 150 dm3 to 250 dm3
450 3

Foam volume (X3) from 200 dm3 to 600 dm3 þ 250 þ 600 ¼ 992:40 dm
3:16
The fourth factor is considered to be sand. It should be noted Hence, the aggregation of three factors never exceeds 1000 dm3.
that the two variables of sand and foam are considered as dense The regression equation describing the relation between response
factors. In foamed concrete production, the role of foam and variable and main effects is:
sand is contradictory, since the foam is a void creator, and sand
Y ¼ b0 þ b1 X 1 þ b2 X 2 þ b3 X 3 þ b11 X 21 þ b22 X 22 þ b33 X 23
acts as a kind of filler. In other words, reducing the sand mass
and increasing the foam volume can reduce density-strength þ b12 X 1 X 2 þ b13 X 1 X 3 þ b23 X 2 X 3 ð8Þ
while reducing foam in line with increasing sand leads to high
where Y is a response variable that can be related to compressive
density-strength. Thus, the sand mass can be determined based
strength, density or cost. Based on the above-mentioned values,
on other materials and three main factors specify the amount
variables’ coding would be regarded according to Table 1 and Fig. 2.
of sand automatically. One of the considerations in the above-
According to BBD for three factors, 15 experiments are totally
mentioned values is the ratio of water to cement (W/C). The
carried out containing three center points. These experiments are
Allowed limitation of this ratio is 0.4–0.6. If this ratio reduced,
described in Table 2.
the cement would not absorb enough water needed for hydra-
tion process, so dried parts remain un-reactied. On the other
3.2. Determination of factors’ values for producing specimens
hand, W/C > 0.6 would result in creating void in cement after
hydration. Although producing specimens with more and less
Prior to producing specimens, we should know the amount of
W/C ratio is possible in laboratory scale, it has no place in man-
sand. Subsequently, the value of factors is determined in one cube
ufactured and practical scale. As the values of foamed concrete
meter, the mass of sand (msand ) in 1000 dm3 is given by:
effective factors are determined in a volume of 1000 L, the  
aggregation of these three factors shall not exceed over 1000 mcement
msand ¼ qsand 1000  V water  V foam  ð9Þ
dm3. It means: qcement
Vcement þ Vwater þ Vfoam 6 1000 dm
3
ð5Þ where:
mcement
The volume of cement, water and foam are shown by Vcement, 1000  V water  V foam  ¼ V sand ð10Þ
Vwater, and Vfoam respectively. Moreover, the density of cement
qcement
(qcement ) is equal to: Sand volume is shown by V sand . The density of the used sand or
lime powder (qsand Þ is equal to:
kg
qcement ¼ 3:16 3
ð6Þ
kg
dm qsand ¼ 2:55 3
ð11Þ
So, the inequality (5) could be written as: dm

Table 1
Coded and uncoded values for factors.

Coded value Natural value Coded value Natural value Coded value Natural value
Cement 1 300 0 375 1+ 450
Water 1 150 0 200 1+ 250
Foam 1 200 0 400 1+ 600

Fig. 2. Box-Behnken design for triple materials of foamed concrete production based on (A) coded values and (B) uncoded values.
104 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110

Table 2
The values of three variables for 15 experiments based on BBD.

Experiment Cement mass coded Cement mass uncoded Water volume coded Water volume uncoded Foam volume coded Foam volume uncoded
No. (X1) (X1) (X2) (X2) (X3) (X3)
1 1 300 1 150 0 400
2 +1 450 1 150 0 400
3 1 300 +1 250 0 400
4 +1 450 +1 250 0 400
5 1 300 0 200 1 200
6 +1 450 0 200 1 200
7 1 300 0 200 +1 600
8 +1 450 0 200 +1 600
9 0 375 1 150 1 200
10 0 375 +1 250 1 200
11 0 375 1 150 +1 600
12 0 375 +1 250 +1 600
13 0 375 0 200 0 400
14 0 375 0 200 0 400
15 0 375 0 200 0 400

Table 3
The amount of four variables while producing specimens.

Specimen No. Cement mass (kg) Water volume (dm3) Foam volume (dm3) Sand mass (kg)
1 1.0 0.5 1.33 3.0
2 1.5 0.5 1.33 2.6
3 1.0 0.84 1.33 2.2
4 1.5 0.84 1.33 1.8
5 1.0 0.67 0.67 4.3
6 1.5 0.67 0.67 0.4
7 1.0 0.67 2.0 0.9
8 1.5 0.67 2.0 0.5
9 1.25 0.5 0.67 4.6
10 1.25 0.84 0.67 3.7
11 1.25 0.5 2.0 1.1
12 1.25 0.84 2.0 0.3
13 1.25 0.67 1.33 2.4
14 1.25 0.67 1.33 2.4
15 1.25 0.67 1.33 2.4

Fig. 3. Grading curve of sand particles.

Recalling that the density of cement was expressed in Eqs. (6) The above-given equation shows the mass of required sand for
and (9) becomes: an experiment. In order to calculate dry density and 28-day com-
 mcement  pressive strength, some specimens should be produced in labora-
msand ¼ 2:55 1000  V water  V foam  ð12Þ tory. The volume of these 15 cube-shaped specimens is equal to
3:16
 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110
15  15  15 cm3 or 0.00375 m3. As the values in Table 2 are
defined in 1 m3, the volume values corresponding to water, foam
and cement mass are as follows:
The volume of water and foam in specimen = 0.00375  the vol-
ume of water and foam in 1 m3
The mass of cement in specimen = 0.00375  the volume of
cement in 1 m3
The values for cement, water and foam in specimens are speci-
fied in Table 3.
3.3. Materials
The type of cement applied in specimens is Portland cement
type 1–525 produced in Tehran cement manufacturing. The chem-
ical and physical properties of this type of cement is determined
according to the ASTM C150 [11].
The quality of water for producing concrete should be drink-
able. This is the most important standard for the water [12]. The
water, used in specimens such other type of concrete, must be
clear, fresh, without any oil, sugar, salt, acid, alkali and other
destructive and harmful materials. The municipal water was suit-
able for this intention.
Foaming agent is a water-soluble, protein-hydrolyzate based,
black-brown colored liquid that forms high stability foams in com-
bination with special additives. These additives are composed of
chemical solvents which never react with cement and sand chem-
Fig. 4. (A) Equipment for generating foam in laboratory (B) Generated foam pumped out from the lancer (C) Mixer (D) Paddle-shaped mixing blade inside the mixer.
105
ically. The PH is between 6.5 and 7.5 and the density is between
1.12 and 1.18 gr/cm3. The size of air bubbles which is the result
of mixing this substance with water should be 0.1–1.5 mm.
Formed bubbles become stabilized in 48 h after creating foamed
concrete.
The type of sand for making specimens is lime powder (Carbon-
ated calcium) which is obtained by crushing cobbles in stone
crusher and could be used as a filler to decrease compressive
strength. Particles size range is between 0.075 and 1 mm. The
result of sieve analysis test is described based on grading curve
shown in Fig. 3.
3.4. Preparing specimens in laboratory
In order to make specimens based on Table 3, firstly, the desired
mass of cement and sand were measured and mixed with the
determined volume of water in a metal graduated cylinder to form
initial grout. Then, the reservoir in the foam generator was filled
with water and foaming agent in ratio of 1–20. Afterwards, the
pressure of 0.5–0.7 MPa was applied in the reservoir by the com-
pressor. The under pressure mixture then passed from the reser-
voir to a foam gun (lancer) that is equipped with some eyelets to
inject formed-foam out. High quality and stable foam is a cream
white colored substance with the density of about 0.06 kg/ dm3.
If the ratio of water to foaming agent is not adjusted, the exiting
substance would be in shape of a liquid with bubbles excessive
in size. As a result, the process of making foam needs so much trial
and error. The foam, created in the foam generator, was injected
106 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110

into the metal graduated container. Measuring the volume of grout

and foam was precisely achieved using a graduated container.
Although reaching an accurate volume of foam was a difficult
work, the closest possible values were determined. The final mix-
ture was poured into turned-on mixer (Fig. 4). Lasting 90 s, mixer
was turned off and the foamed concrete was transferred to 15 
15  15 cm3 test cubes composite mold [13]. After 24 h, the spec-
imens were brought out from the opened molds.
It is noteworthy that specimen No.3 had low adhesion and vis-
cosity on account of extreme high water to cement ratio. For cur-
ing, specimens were put in the water tank in 19 ± 2 °C
temperatures. Cured samples were prepared to measure the den-
sity and compressive strength. (Fig. 5).

3.5. Determination of dry density, 28-day compressive strength and

cost

Since 28-day compressive strength was intended, measuring

strength of foamed concrete specimens was done 4 weeks after
forming samples. As mentioned before, the specimen’s volume is
0.003375 m3. However, dividing the measured mass (msample ) on
the laboratory balance by specimen’s volume gives the dry density
[14]:

msample
qðDryÞ ¼ ð13Þ
0:003375
Owing to lightness and low strength, it was not possible to mea-
sure compressive strength in the concrete compression strength
testing apparatus with the minimum capacity of normally more
than 200 tons. Therefore, we were obliged to take test specimens
in grout compression strength testing machine instead (Fig. 6).
The samples were tested based on standard ISO1920-4 [15]. The
values for strength, shown on the equipment screen by kg/cm2, Fig. 6. Measuring 28-day compressive strength in grout compression strength
testing equipment.
should be converted to MPa.

Fig. 5. (A) Curing specimens in the water tank (B) Specimens brought out from the water tank (C) One set of specimens dried in fresh air.
 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110 107

For each experiment, specimens were replicated three times. Table 5

The compressive strength for an experiment was equal to an aver- The regression coefficients for density based on coded values.

age of three samples of compressive strength. Considering the cost Term Coef SE Coef T P
of cement (kg), sand (kg), and foam (dm3) according to the Iranian Constant 11166.46 2.304 506.254 0.000
currency unit (ICU), the production cost for specimens were esti- Cement 30.65 1.411 21.721 0.000
mated. One dm3 foaming agent is required for generating 70 dm3 Water 126.41 1.411 89.588 0.000
foam. Thus, with the exception of the cost of consumed water tar- Foam 510.15 1.411 361.560 0.000
Cement*Cement 0.59 2.077 0.286 0.786
iff, the cost of materials in every experiment was as follows: Water*Water 0.67 2.077 0.322 0.760
Cost ¼ 0:14mcement þ 0:035msand þ 0:043V foam ð14Þ Foam*Foam 1.39 2.077 0.669 0.533
Cement*Water 2.30 1.995 1.154 0.301
Measured values for 28-day compressive strength and dry den- Cement*Foam 0.14 1.995 0.069 0.948
Water*Foam 0.06 1.995 0.029 0.978
sity for specimens were generalized to consider 15 experiments as
shown in Table 2. The final results are described in Table 4.

4. Results and discussions

Table 6
The regression coefficients for compressive strength based on coded values.
4.1. Determination of the regression models for density and
compressive strength Term Coef SE Coef T P
Constant 8.4933 0.5281 16.083 0.000
MINITAB software, which is an efficient tool to achieve exact Cement 3.7988 0.3234 11.747 0.000
Water 1.4100 0.3234 4.360 0.007
and trustable results, was used to analyze the obtained data. Enter-
Foam 5.0862 0.3234 15.728 0.000
ing the values obtained in Table 4, the first step was to find a mean- Cement*Cement 1.6129 0.4760 3.388 0.019
ingful relationship between input variables and individual outputs. Water*Water 0.9596 0.4760 2.016 0.100
This was made possible by obtaining regression coefficients Foam*Foam 0.2271 0.4760 0.477 0.653
appeared on software’s worksheet. Table 5 shows the regression Cement*Water 0.2450 0.4573 0.536 0.615
Cement*Foam 2.9425 0.4573 6.434 0.001
coefficients related to density response. Based on this Table, we Water*Foam 0.6150 0.4573 1.345 0.236
have:
Y Density ¼ 1166:46 þ 30:65X cement  126:41X water
 510:15X foam  0:59X 2cement  0:67X 2water It is noteworthy that some terms in Eq. (16) have p-values lar-
þ 1:39X 2foam  2:3X cement X water þ 0:14X cement X foam ger than 0.05 (a pre-specified significance level), which can be
easily neglected. The two above-shown equations provide the val-
 0:06X water X foam ð15Þ
ues for the density or compressive strength for a given set of desir-
Actually, because the p-values corresponding to second-order able values of cement, water and sand. Replacing the uncoded
and interaction terms are larger than value 0.05 (a common signif- values (Xcement, Xwater and Xfoam) with coded values of three factors
icance level), these terms can be considered unimportant. Thus, the in Eqs. (15) and (16) provides the expected values corresponding to
equation could be rewritten taking into account only the first- strength and density. Comparing the expected values, obtained
order terms. from fitting the proper model, with measured values proves the
Similarly, regression coefficients for compressive strength accuracy of the fitted model as seen in Table 7. For simplicity,
based on the three variables are shown in Table 6. Figs. 7 and 8 depict the model (expected) and measured density-
Thus, the compressive strength function is written as below: strength values for all experiments. The results clearly reveal that
the two lines belonging to fitted and measured values have negli-
Y Strength ¼ 8:49 þ 3:79X cement  1:41X water  5:08X foam gible distance from each other.
 1:61X 2cement þ 0:95X 2water þ 0:22X 2foam However, as mentioned earlier, the aim of this study is to opti-
mize three responses of foamed concrete including the cost in
þ 0:24X cement X water  2:94X cement X foam
some determined ranges of density. This multi-objective optimiza-
þ 0:61X water X foam ð16Þ tion will be discussed in the next subsection.

Table 4
Experiment results for compressive strength, density tests and calculated cost.

Experiment No. Cement mass (kg) Water volume (dm3) Foam volume (dm3) Sand mass (kg) Dry density (kg/ m3) Strength (MPa) Cost (ICU)
1 300 150 400 905 1256.41 5.39 91
2 450 150 400 784 1324.37 12.74 108
3 300 250 400 650 1010.62 2.45 82
4 450 250 400 529 1069.37 10.78 99
5 300 200 200 1288 1647.91 4.9 96
6 450 200 200 1167 1706.87 18.5 112
7 300 200 600 268 627.36 1.96 77
8 450 200 600 147 686.87 3.43 94
9 375 150 200 1355 1804.89 17.65 108.5
10 375 250 200 1100 1549.78 13.5 99.5
11 375 150 600 335 784.69 4.9 90
12 375 250 600 80 529.35 2.94 81
13 375 200 400 717 1167.25 8.33 95
14 375 200 400 717 1160.65 8.82 95
15 375 200 400 717 1171.47 8.33 95
108 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110

Table 7
Comparison between model values and measured values for density and strength.

Experiment Model value Measured Model Value Measured

No. for density value for for strength value for
density strength
1 1258.66 1256.41 5.69 5.39
2 1324.56 1324.37 12.79 12.74
3 1010.44 1010.62 2.39 2.45
4 1067.14 1069.37 10.45 10.78
5 1644.47 1647.91 5.45 4.9
6 1705.49 1706.87 18.91 18.14
7 623.19 627.36 1.17 1.96
8 684.77 686.87 2.87 3.43
9 1801.25 1804.89 16.76 17.65
10 1548.55 1549.78 12.72 13.23
11 780.37 784.69 5.38 4.9
12 527.43 529.35 3.78 2.94
13 1166.46 1167.25 8.49 8.33
14 1166.46 1160.65 8.49 8.82
15 1166.46 1171.47 8.49 8.33

Fig. 9. Optimization plots for density from 800 to 1000 kg/m3.

Table 8
The mix-design obtained by optimization plots in
Fig. 9.

Fig. 7. . Comparison between model values (gray line) and measured values (black The best value for dry density 800 kg/m3
line) for density. The best value for strength 6.7 MPa
The best value for cost 89.5 ICU
Cement mass 407 kg
Water volume 250 dm3
(w/c) water to cement ratio 0.55
Foam volume 498 dm3
Sand mass 314 kg

Fig. 8. Comparison between model values (gray line) and measured values (black
line) for strength.

4.2. Responses optimization

One of the most important purposes for the foamed concrete

optimization is to find the maximum compressive strength and
minimum cost in the specified density ranges. Three density ranges
with regard to the variety of foamed concrete were applied:

4.2.1. In density range between 800 kg/m3 and 1000 kg/m3, the
normal compressive strength range was from 3.5 to 10
MPa and the cost interval is from 77ICU to 112ICU according
to Table 4. This range of density was used for producing
blocks and panels. The software gives optimization plots in
Fig. 9. An optimization plot is a graphical representation of
the best value for response or responses based on the fac-
tors. The individual desirability for a single response along
with composite desirability for all responses are displayed
on the plot. The vertical red lines indicate the best values Fig. 10. Optimization plots for density from 1000 to 1200 kg/m3.
 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110 109

Table 9 for the factor affecting the response(s). It should be noted

The mix-design obtained by optimization plots in Fig. 10.
that the desirability close to one is preferred. The composite
The best value for dry density 1035.5 kg/m3 desirability displayed on the plot (0.68) implies a good value
The best value for strength 9.9 MPa in case a three-objective optimization is considered. Con-
The best value for cost 97 ICU
verting the coded values to natural values and using Eq.
Cement mass 434 kg
Water volume 250 dm3 (12) to determinate sand mass, the values for mix-design
(w/c) water to cement ratio 0.57 is reported in Table 8.
Foam volume 410 dm3 4.2.2. The range of density between 1000 kg/m3 and 1200 kg/m3 is
Sand mass 517 kg applied to fill tunnels, voids and canals. The normal com-
pressive strength range is from 8 kg/cm2 to 12 MPa and
the cost varies from 77ICU to 112ICU according to Table 4.
Table 10 The software provides optimization plots as shown in
The mix-design obtained by optimization plots in Fig. 11. Fig. 10. It is remarkable that the obtained composite desir-
ability is equal to 0.55 which is quite satisfactory. Changing
The best value for dry density 1279 kg/m3
The best value for strength 13.58 MPa the coded values to natural values and using Eq. (12) to
The best value for cost 102.5 ICU obtain sand mass, the optimized mix-design is provided in
Cement mass 450 kg Table 9.
Water volume 250 dm3 4.2.3. If foamed concrete is used for sub-base road, the density
(w/c) water to cement ratio 0.55
Foam volume 318 dm3
range of 1200 kg/m3–1400 kg/m3 would be applicable. In
Sand mass 739.5 kg this range, the normal compressive strength is from 12 to
14 MPa and the cost changes from 77ICU to 112ICU. Once
more, acquired composite desirability (0.5) is quite accept-
able. Values shown in Table 10 are obtained from converting
the coded values to natural values (Fig. 11) in line with
determining sand mass using Eq. (12).

4.3. Determination of mix-designs

The most challenging part of the foamed concrete manufactur-

ing is to determinate the precise mix-design based on accurate and
specified density-strength. Nine desirable combinations of density
and strength were considered. Fixing desired strength-density in
MINITAB, defining target values and obtaining optimization plots
give the input values to reach specified outputs. Note that the cor-
responding compressive strength for a density were chosen based
on experts and specialist’s points of view and their experiments.
The obtained optimization plots for nine specified densities and
strength give the mix-design of foamed concrete. Coded values
were changed to uncoded values and the sand mass was deter-
mined based on Eq. (12) for all density-strength combinations.
The results are reported in Table 11. It is noteworthy that the opti-
mization plots have not been provided for the reason of space.
However, they can be easily accessible upon request. The outcomes
satisfy the producer’s needs which reveal the superiority of the
proposed method compared with the trial and error one.

5. Conclusions

In this paper, Box-Behnken design (BBD) was applied as a

useful statistical method for optimization of mix-design foamed
Fig. 11. Optimization plots for density from 1200 to 1400 kg/m3.
concrete which is produced from cement, water, foam and sand.

Table 11
The determined mix-design for nine density-strength obtained by optimization plots.

Density 28-Day compressive Cement Cement volume Foam volume Water to cement Water volume Sand volume Sand mass
(kg/m3) strength (MPa) mass (kg) (dm3) (dm3) ratio (w/c) (dm3) (dm3) (kg)
650 2.5 325 103 600 0.60 195 102 260
800 4 311 98 584 0.48 150 168 428
900 6 344 109 550 0.43 150 191 487
1050 9 450 142 441 0.48 217 200 511
1200 12 446.5 141 361 0.53 237 261 665.5
1300 13 416 132 304 0.60 250 314 800
1400 14.5 426 135 274 0.57 243 348 887
1500 16 424 134 227 0.59 250 389 992
1600 16.5 416.5 132 200 0.56 237 431 1100
110 S. Asadzadeh, S. Khoshbayan / Construction and Building Materials 170 (2018) 101–110
Single and multi-objective optimizations of three responses
including cost, compressive strength and dry density were
addressed. The below accomplishments have been obtained from
the present study:
1. Fully quadratic models including first-order, second-order and
interaction terms were obtained, all of which describe the rela-
tionship among responses (density and strength) and three
influential factors called cement, water and foam. The coeffi-
cients of the regression models were thoroughly calculated
with the aid of MINITAB software.
2. The comparison between the model values (fitted values) and
measured values revealed that the fitted model is accurate
and reliable to be applied in real practice.
3. Manufacturers can effectively use the obtained model to
find the expected values corresponding to the three
responses called density, compressive strength and cost.
This could be easily obtained for different application of
foamed concrete.
4. The mix-design of foamed concrete was determined based on
different target values of density and strength. The results
ensure the producers to use optimal values of influential factors
leading to lowest cost and the best physical-mechanical proper-
ties of foamed concrete. The obtained mix-design could be
effectively used with determined density and compressive-
strength to produce foamed concrete.
References
[1] A.M. Neville, Properties of Concrete, fifth ed., Pearson, San Francisco, 2014.
[2] N.U. Kockal, T. Ozturan, Optimization of properties of ash aggregates for high-
strength lightweight concrete production, Mater. Des. 32 (2011) 3586–3593.
[3] B. Simsek, Y. Tansel, E.H. Simsek, A TOPSIS-based Taguchi optimization to
determine optimal mixture properties of the high strength self – compacting
concrete, Chemo Metrics and Intell. Lab. Syst. 125 (2013) 18–32.
[4] M. Joshaghani, A.A. Ramezanipour, O. Ataei, A. Golroo, Optimizing pervious
concrete pavement mixture design by using the Taguchi method, Constr. Build.
Mater. 101 (2015) 317–325.
[5] Q. Li, L. Cai, Y. Fu, H. Wang, Y. Zou, Fracture properties and response surface
methodology model of alkali-slag concrete under freeze-thaw cycles, Constr.
Build. Mater. 93 (2015) 620–626.
[6] R. Mohebi, K. Behfarnia, M. Shojaei, Abrasion resistance of alkali-activated slag
concrete designed by Taguchi method, Constr. Build. Mater. 98 (2015) 792–
798.
[7] H. Tanyildizi, M. Sahin, Application of Taguchi method for optimization of
concrete strengthened with polymer after high temperature, Constr. Build.
Mater. 79 (2015) 97–103.
[8] D.C. Montgomery, Design and Analysis of Experiments, eighth ed., John Wiley
& Sons, New York, 2012.
[9] M.J. Anderson, P.J. Withcomb, DOE Simplified, third ed., CRC, Boca Raton, 2015.
[10] R. Myers, D.C. Mongomery, C.M. Anderson, Response Surface Methodology,
fourth ed., John Wiley & Sons, New York, 2016.
[11] ASTM C150, (2017), Standard specification for Portland cement. Annual Book
of ASTM Standards.
[12] A.M. Neville, J.J. Brooks, Concrete Technology, second ed., Pearson, San
Francisco, 2010.
[13] ISO 1920, (2004), Testing of concrete, Part 3: Making and curing specimens.
[14] ISO 1920, (2004), Testing of concrete. Part 5: Properties of hardened concrete
other than strength.
[15] ISO 1920, (2005), Testing concrete. Part 4: Strength of hardened concrete.