Construction and Building Materials 128 (2016) 108–117

Contents lists available at ScienceDirect

Construction and Building Materials

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

Modelling of the compressive strength development of cement mortar

with furnace slag and desulfurization slag from the early strength
Chien-Chih Wang
Department of Civil Engineering and Geomatics, Cheng Shiu University, Kaohsiung 83347, Taiwan, ROC

h i g h l i g h t s

The compressive strength at other curing ages estimated from the early strength on Day 7 is proposed.

This model can be used for assessing the design strength of concrete during construction phase from early age test results.

The compressive strength decreased as the replacement rate of desulfurization slag increased.

The contribution of the slag powder to the strength is significant at later stage.

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

Article history: Replacing natural resources with renewable materials is a very important research topic for sustainable
Received 10 August 2016 development. Several recycled materials such as furnace slag (S) and desulfurization slag (D) were widely
Received in revised form 11 October 2016 used and applied in concrete or cement mortar. The compressive strength at the 28th day was usually
Accepted 14 October 2016
used as design strength for structure. Thus, an assessment of the compressive strength of cement mortar
Available online 21 October 2016
with furnace slag and desulfurization slag from the early strength is desirable to develop. In this study,
desulfurization slag that could pass through a No. 4 sieve and furnace slag with a fineness of approxi-
Keywords:
mately 4000 cm2/g were used to replace the fine aggregate (0%, 10%, 20% or 40%) and cement (0%, 10%,
Cement mortar
Desulfurization slag
20% or 40%), respectively, through a volumetric method. The cement mortar thus formulated was then
Furnace slag put through a series of compressive strength tests in which the samples were cured at a 23–25 °C room
Compressive strength temperature and a 75 °C high temperature. According to the test results, the compressive strength
Prediction model decreased as the replacement rate of desulfurization slag increased, but increased as that of the furnace
slag increased. Using the test results when the replacement rates of D were 0%, 10% and 20%, a hyperbolic
function was used to perform a multivariate non-linear regression analysis, thus establishing a compres-
sive strength prediction model for cement mortars of other ages based on the early strength at the age of
7 days. From the confirmatory analysis, when the replacement rate of desulfurization slag D = 40%, the
MAPE (mean absolute percentage error) value of the test sample cured at high temperature was found
to be 6.13%, while that for the test sample cured at room temperature was 11.32%. Moreover, when D
was 0%, 10% and 20%, the MAPE value for the test sample cured at high temperature was 3.05%–4.47%,
respectively, while that for the test sample cured at room temperature was 7.83%–9.17%. Generally
speaking, the results delivered by the analytical model established in this study were satisfactory under
different conditions. As it is capable of performing a predictive analysis of the compressive strength of
cement mortar at other ages based on the early strength at the age of 7 days, this model could contribute
to the safety assessment of concrete structures during the construction phase.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Desulfurization waste slag (hereafter abbreviated as desulfur-
ization slag, D) is a solid waste generated from the desulfurization
process of molten iron in a blast furnace with the help of a desul-
furization agent in an integrated steel company that uses iron ore
E-mail address: ccw@gcloud.csu.edu.tw
as the main raw material. Currently, the steel companies in Taiwan
have an annual production capacity of approximately 300,000 tons
of desulfurization slag [1]. Generated in the melting process at high
temperature, desulfurization slag is resistant to high temperatures,
less affected by climatic changes and very safe because of the low
heavy metal mobilization. At present, apart from turning out
metallic iron in a magnetic separation process, which is very
renewable and valuable, desulfurization slag could be reused, after

http://dx.doi.org/10.1016/j.conbuildmat.2016.10.083
0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.
 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117 109

purification, as an agricultural soil amendment and fertilizer

Percentage of pass.weight (%)

100
because it meets the standards for defining corrosive industrial
wastes in terms of its pH value (as high as 12.5). Given its low 80
ASTM C33 upper limit
heavy metal mobilization, it could also be used as an earthwork
backfill material or a partial replacement for limestone in the firing 60
of cement clinker [1]. In recent years, desulfurization slag has been ASTM C33 lower limit
reused to gradually replace the coarse and fine aggregates in some 40
concrete materials and the fine aggregate in cement mortar [1–4].
Kuo and Shu (2014) [2] partly replaced the natural sand in the mix- 20
Fine aggregate
ture of cement mortar with desulfurization slag and suggested that
Desulflurization slag
the speed of the volume increase of the cement mortar containing 0
desulfurization slag would increase with the replacement rate of 10.0 1.0 0.1
desulfurization slag if the mortar was cured at high temperature. Grain size (mm)
The compressive strength tended to decrease as the replacement
rate of desulfurization slag increased [3,4]. Blast furnace slag (S), Fig. 1. Cumulative particles distribution of fine aggregate and desulfurization slag.

as a raw material for Portland cement, on the other hand, is mainly

a byproduct for iron production in an integrated steel company.
According to how it is cooled, it could be divided into water-
cooled and air-cooled furnace slags. Water-cooled furnace slag,
once ground into fine powder, is usually called slag powder and
could be mixed into concrete to replace some cement. This could
not only reduce the cost of the concrete but also improve its qual-
ity [5]. According to Lho et al. (2012) [6], when slag powder
replaces part of the cement, the compressive strength and tensile
strength would first increase with the replacement rate of the fur-
nace slag. However, when the replacement rate of furnace slag sur-
passed 40%, the compressive strength and tensile strength would
decrease instead.
With sustainable development and the reuse of waste materials
as a starting point, applications involving the addition of desulfur-
ization slag and slag powder to replace some of the raw materials
for concrete or cement mortar therefore constitute a research topic
worthy of study. Additionally, the compressive strength at the 28th
day was usually used as design strength for structure. Thus, an
assessment of the compressive strength of cement mortar with fur-
nace slag and desulfurization slag from the early strength is desir-
able developed, it will be very helpful to the safety assessment and
analysis of structures under construction. In this study, a series of
compressive strength tests were carried out on cement mortars
with a fixed water-binder ratio containing desulfurization slag
and furnace slag (when the replacement rates of desulfurization
slag with fine aggregate D are 0%, 10% and 20%), and a compressive
strength prediction model encompassing such influence factors as
the contents of desulfurization slag and slag powder and the age
was proposed. A compressive strength test was then conducted
under a different replacement rate (D = 40%) to validate the com-
pressive strength prediction model established in this study.
The particle size distribution curves of the fine aggregate and
desulfurization slag are shown in Fig. 1.
2.2. Test variables and method
This study used a standard mix design of cement mortar accord-
ing to ASTM C109, with a fixed water-binder-fine aggregate ratio of
0.5:1:2.75. The volumetric method was used for the design. Recy-
cled materials of slag powder and desulfurization slag were used to
replace the cement (0%, 10%, 20% and 40%) and fine aggregate (0%,
10%, 20% and 40%), respectively. A 5 cm * 5 cm * 5 cm cement mor-
tar specimen was cast and solidified. The forms were removed after
24 h, and the specimens were placed and cured at room tempera-
ture (23–25 °C) and high temperature (75 °C) in saturated limewa-
ter, respectively. The flow and setting time were tested in
accordance with ASTM C230 and ASTM C403, respectively. The
compressive strength was tested at the ages of 3, 7, 28, 56 and
91 days according to ASTM C109 and ASTM C597.
The physical properties of the materials are as shown in Table 1,
and the chemical properties of the materials are as shown in
Table 2. The unit weights of the mix design cement mortar with
added recycled materials are as shown in Table 3.
Table 1
The physical properties of materials.
Properties Specific Unit weight Water absorption
gravity (kg/m3) (%)
Fine Aggregate 2.63 1865 1.6
Desulfurization Slag 2.38 1009 34

2. Test program

Table 2
2.1. Test materials
The chemical properties of the materials. Unit:%.

1. Cement: Type I Portland cement produced by the Taiwan Properties Cement Slag Powder Desulfurization Slag

Cement Corporation was used; its properties conformed to SiO2 20.22 34.47 9.93
those of Type I Portland cement specified in ASTM C150. Al2O3 4.96 13.71 4.22
Fe2O3 2.83 0.33 8.23
2. Mixing water: Conforms to ASTM C94 for concrete mixing
CaO 64.51 41.00 69.08
water. MgO 2.33 6.60 2.33
3. Fine aggregate: The aggregate originated from the Ligang Dis- SO3 2.46 – 4.3
trict and conformed to ASTM C33. Alkalis 0.48 – –
4. Furnace Slag (S): GGBFS (Ground-granulated blast-furnace slag) K2O – – 0.06
Na2O – – 0.14
was produced by the CHC Resources Corporation and was ground TiO2 – – –
to 4000 cm2/g, and its properties conformed to ASTM C989. P2O5 – – 0.17
5. Desulfurization slag (D): Desulfurization slag is the surplus MnO – – 0.6
remaining after the purification of the GGBFS produced by an Cr2O3 – – 0.08
LOI 2.4 3.1 –
integrated steel plant using magnetic separation. The slag
f-CaO – – 0.95
passed through a No. 4 sieve as fine aggregate.
110 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117

Table 3
Unit weight of mix design cement mortar adding slag powder and desulfurization slag materials.

No. W/B Cement Slag powder Desulfurization slag Unit: kg/m3

Fine aggregate Water
D0 S0 0.5 541 0 0 1488 271
S10 487 50 0 1476 268
S20 433 99 0 1470 266
S40 330 202 0 1463 264
D10 S0 0.5 541 0 135 1353 271
S10 487 50 134 1342 268
S20 433 99 133 1337 266
S40 330 202 132 1331 264
D20 S0 0.5 541 0 269 1219 271
S10 487 50 267 1209 268
S20 433 99 266 1204 266
S40 330 202 265 1198 264
D40 S0 0.5 541 0 539 950 271
S10 487 50 533 943 268
S20 433 99 531 939 266
S40 330 202 529 934 264

3. Experimental results
The discussions in this study are based only on the results of the
compressive strength tests when the replacement rates of desulfu-
rization slag with fine aggregate were 0%, 10% and 20%. These
results were also used to derive the compressive strength predic-
tion model. The results generated when the replacement rate of
desulfurization slag with fine aggregate was 40% were mainly used
to validate and analyse the prediction model. The details of the
tested results have been described in Chen (2016) [7].
3.1. Flow test
Fig. 2 shows the results of the flow test as desulfurization slag is
added to the cement mortar with slag powder. According to the
figure, when no desulfurization slag or slag powder is added
(D = 0 and S = 0), the flow value is approximately 120%. As slag
powder tends to bind together in water, it will increase the viscos-
ity of the fresh cement mortar. When the replacement rate of
desulfurization slag is fixed, the flow value will decrease as the
replacement rate of slag powder increases. That is to say, the
higher the replacement rate of slag powder is, the poorer its work-
mortar will be better when desulfurization slag is used to replace
fine aggregate.
3.2. Setting time
When no desulfurization slag or furnace slag is added (D = 0 and
S = 0), the initial setting time and final setting time are 170 min
and 273 min, respectively, as is shown in Fig. 3. According to the
test results, when the content of desulfurization slag is fixed, the
setting time will increase as the replacement rate of the slag pow-
der goes up. When the replacement rate of slag powder S = 40%, the
initial setting time and final setting time will be approximately
138%  168% and 124%  152% of the values when S = 0, respec-
tively. When the replacement rate of slag powder is fixed, the set-
ting time will also increase as the replacement rate of
desulfurization slag goes up. When the replacement rate of desul-
furization slag D = 40%, the initial setting time and final setting
time will be approximately 130%  158% and 111%  148% of the
values when D = 0, respectively. A possible reason why the setting
600 (a) S=0% W/B=0.5
Initial setting time (min)

ability will be. When the replacement rate of slag powder S = 40%, S=10%
the flow value will be approximately 50%  57% of its value when 500 S=20%
S=40%
S = 0. When desulfurization slag is used to replace the fine aggre- 400
gate, the flow value will increase with the replacement rate of
desulfurization slag. When the replacement rate of desulfurization 300
slag D = 40%, the flow value will be approximately 117%  133% of 200
its value when D = 0. In conclusion, the workability of the cement
100
200 0
S=0% W/B=0.5 S=0%
600 (b) W/B=0.5
S=10%
Final setting time (min)

S=10%
160 S=20% S=20%
500
S=40% S=40%
Flow (%)

120 400

300
80
200
40
100

0 0
0 10 20 40 0 10 20 40
Desulfurization slag content (%) Desulfurization slag content (%)

Fig. 2. Relationship between flow and desulfurization slag content. Fig. 3. Relationship between setting time and desulfurization slag content.
 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117 111

time will increase with the addition of desulfurization slag is that (a) W/B=0.5 D=0 T=75 C
0
60
the aggregates are already in the saturated surface-dry (SSD) con-
dition before the mixing happens. Given, moreover, the high water 50
absorption of desulfurization slag, the overall water content in the
40
aggregates will increase as the replacement rate of desulfurization
slag goes up, thus influencing the setting time. With the addition of 30
slag powder, the setting time of the cement mortar may be longer
because of the lower CH crystal growth rate after the addition. 20

10
3.3. Compressive strength
0

Compressive Strength (MPa)

60 (b) S=0% W/B=0.5 D=0.1 T=75 C
0

As it was suggested in Kuo and Shu (2014) [2] that the volume S=10%
of the cement mortar containing desulfurization slag would go up 50 S=20%
S=40%
significantly if left at high temperature and the compressive
40
strength and durability would also be significantly affected, during
the preparation of the test sample in this study, two curing condi- 30
tions were established: room temperature (23–25 °C) and high
temperature (75 °C). This section will only consider the results of 20
the three compressive strength tests when the replacement rates 10
of desulfurization slag D were 0%, 10% and 20%, while the results
generated when D was 40% were mainly used to validate and anal- 0 0
60 (c) W/B=0.5 D=0.2 T=75 C
yse the model. Figs. 4 and 5 show the results of compressive
strength tests conducted at room temperature and at high temper- 50
ature, respectively, as the replacement rate of the desulfurization
slag in the cement mortar changes. In general, for the test samples 40
cured both at room temperature and at high temperature, the com- 30
pressive strength of the cement mortar containing desulfurization
slag will increase as it ages and decrease as the replacement rate of 20
desulfurization slag increases. A possible cause for these changes 10

0
3 7 28 56 91
(a) W/B=0.5 D=0 Room Temp.
60 Curing Age (day)
50
Fig. 5. Compressive strength of desulfurization slag cement mortar (cured at high
temperature, 75 °C).
40

30 might be the dilation effect generated by the reaction between

the water and the f-CaO in the desulfurization slag.
20
When cured at room temperature, the early strength of the
10 cement mortar containing desulfurization slag at the age of 7 days
will decrease as the replacement rate of slag powder S increases.
0
Compressive Strength (MPa)

60 (b) S=0% W/B=0.5 D=0.1 Room Temp. However, the compressive strength will increase with the replace-
S=10% ment rate S after the age of 28 days, as is shown in Fig. 4. Therefore,
S=20%
50 the addition of slag powder will lead to a significant increase in the
S=40%
40
late compressive strength. Taking the compressive strength on the
28th day as an example, when no slag powder is added and the
30 replacement rate of desulfurization slag D = 0–20%, the compres-
sive strength is 37.7–46.9 MPa. When the content of slag powder
20
is fixed and the replacement rate of desulfurization slag D = 20%,
10 the compressive strength is approximately 80–85% of its value
when D = 0%. On the 28th day, when the content of desulfurization
0
60 (c) W/B=0.5 D=0.2 Room Temp. slag is fixed and the replacement rate of slag powder S = 40%, the
compressive strength is approximately 105–112% of its value when
50 S = 0%. In the test samples cured at high temperature, the compres-
sive strength of the cement mortar containing desulfurization slag
40
increases with the replacement rate of slag powder S, regardless of
30 the age, as is shown in Fig. 5. The reason for this might be that the
hydration reaction will be accelerated at the beginning of the high-
20
temperature curing, thus filling the gaps in the test sample and
10 resulting in a different trend from that at room temperature. Tak-
ing the compressive strength at the age of 28 days as an example,
0
3 7 28 56 91 when no slag powder is added and the replacement rate of desul-
furization slag D = 0–20%, the compressive strength is 27.7–
Curing Age (day)
45.8 MPa. When the content of slag powder is fixed and the
Fig. 4. Compressive strength of desulfurization slag cement mortar (cured at room replacement rate of desulfurization slag D = 20%, the compressive
temperature, 23–25 °C). strength is approximately 105–117% of its value when D = 0%. On
112 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117

the 28th day, when the content of desulfurization slag is fixed and
the replacement rate of slag powder S = 40%, the compressive
strength is approximately 111–117% of its value when S = 0%. It
should be noted, however, that when cured at high temperature,
the compressive strengths of the test samples at all mixture ratios,
especially those with a higher replacement rate of desulfurization
slag, are lower than at room temperature, except at the age of
3 days, when the early strength at high temperature is slightly
higher than that at room temperature. The dilation effect caused
by the high-temperature curing of the desulfurization slag at the
later stages, will therefore lead to a significant downward trend
in compressive strength. This coincides with the circumstances
for ordinary concrete, whose compressive strength is higher at
the beginning of high-temperature curing but decreased to lower
than that of room-temperature curing after a certain age [8,9].
4. Development of the compressive strength prediction model
In the design of a concrete structure, the compressive strength
of the concrete is an essential element. The 28-day compressive
strength of concrete is usually considered the design strength. To
60 (a) W/B=0.5 D=0.1
Room Temp. (23-250C)
50
40
30
Compressive strength (MPa)
ensure this strength, it is necessary to wait a considerable time,
i.e., 28 days [10]. Therefore, if a prediction model could be estab-
lished to analyse the compressive strength at the later stages based
on the early strength, it will be very helpful for the safety assess-
ment and analysis of structures. Although there many studies
[10–15] have been conducted to predict and analyse the compres-
sive strength of concrete or cement mortar on the 28th day or
other days based on the early strength, rarely has any compressive
strength prediction model for cement mortar containing desulfur-
ization slag been proposed. Based on a series of test results on
cement mortar, Kheder et al. (2003) [11] suggested a model for
the prediction of the cement compressive strength at the ages of
7 and 28 days within 24 h. This model could consider as few as 6
variables and as many as 17 variables, and the more the variables
that are considered, the more precise the prediction will be. Kabir
et al. (2012) [10] used 7-day test results as the basis, while Alilou
and Teshnehlab (2010) [12] started from the strength on the 3rd
day to establish their own prediction models of the compressive
strength on the 28th day. The influence of partial aggregate
replacements, including with fly ash, furnace slag, and waste glass,
on the compressive strength increased with age under identical
conditions, but the increasing tendency became smooth over time.
Our previous study used a hyperbolic function to calculate the rela-
tionship between the compressive strength and curing time for
waste LCD glass concrete [16,17].
In this study, when a cement mortar whose replacement rate of
desulfurization slag D = 10% is left to cure at high temperature or
room temperature, the relationship between the compressive

S=0%
S=10% strength and curing age within 28 days could also be represented
20 hyperbolic function S=20%
as a hyperbolic function, as shown in Fig. 6. The test results gener-
S=40%
10 envelope ated when the replacement rates of desulfurization slag are D = 0%
and D = 20% demonstrate the same trend. In combination with this
0
60 (b) W/B=0.5 D=0.1 T=750C hyperbolic function, the compressive strength test results when
the replacement rates of desulfurization slag D are 0%, 10% and
50 20% and the results generated within 28 days are considered to
40 derive a compressive strength prediction model. The compressive
strength test results generated when the replacement rate of
30 S=0% desulfurization slag D is 40% or when the curing age is older than
20
S=10% 28 days were mainly used to validate and analyse the prediction
hyperbolic function S=20%
S=40%
model.
10 In deriving the older-age compressive strength prediction
envelope

0 model on the basis of the early strength, the compressive strength

0 10 20 30 at the age of 7 days will be used as the normalized basis. In combi-
Curing Age (day) nation with the hyperbolic function, an equation for the prediction
of the compressive strength of desulfurization slag cement mortar
Fig. 6. Characteristic of compressive strength versus curing age. is obtained, as is shown by Eq. (1), wherein a and b refer to the

Table 4
The coefficient values of hyperbolic function.

Desulfurization slag content Slag powder content Temperature of curing condition

23–25 °C 75 °C
Coefficients of hyperbolic function
a b a b
D=0 S=0 1.067 0.869 0.888 0.870
S = 0.1 1.449 0.756 1.192 0.818
S = 0.2 1.204 0.826 1.207 0.843
S = 0.4 1.472 0.786 0.955 0.843
D = 0.1 S=0 1.118 0.845 1.024 0.832
S = 0.1 1.921 0.774 1.141 0.831
S = 0.2 1.735 0.766 0.893 0.871
S = 0.4 1.084 0.846 0.792 0.887
D = 0.2 S=0 1.109 0.851 0.906 0.846
S = 0.1 1.601 0.809 0.846 0.883
S = 0.2 1.523 0.812 0.881 0.843
S = 0.4 1.184 0.809 0.797 0.868

Note: the test results of compressive strength at 3, 7 and 28 days were used and regressed by hyperbolic function.
 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117 113

coefficients of the hyperbolic function and t refers to the curing
age. Table 4 shows the coefficient values of a and b obtained when
the replacement rates of desulfurization slag D are 0%, 10% and 20%
and as the replacement rate of slag powder varies on the 3, 7th and
28th days, normalized by the compressive strength on the 7th day.
According to the results, the values of coefficients a and b are
2.0
(a)
W/B=0.5 T=75 C
Coefficient a
related to the replacement rates of the desulfurization slag D and
the slag powder S, apart from the curing temperature. Therefore,
when the curing temperature is fixed, the values of coefficients a
and b could be expressed as functions of D and S, as shown by
Eqs. (2) and (3). Fig. 7 shows the results of the test samples cured
at high temperature, namely, the relationship between the values
of the hyperbolic function coefficients a and b and the replacement
rate of slag powder S. The values of coefficient a (replacement rate
0
D=0 of desulfurization slag) are distributed in the interval 0.8–1.2 and
D=0.1 do not show an upward or downward trend if put alongside the
D=0.2
replacement rate of slag powder S. Roughly speaking, their rela-

1.5
tionship could be expressed as a non-linear quadratic polynomial
function, as is shown by Fig. 7(a) and Eq. (4). The values of coeffi-
cient b, on the other hand, are distributed in the interval 0.82–
1.0 0.88. As they also show no upward or downward trend if put
alongside the replacement rate of slag powder S, their relationship
could also be expressed as a non-linear quadratic polynomial func-
0.5
0
tion, as is shown by Fig. 7(b) and Eq. (5).
(b) W/B=0.5 T=75 C D=0
1.0 0
D=0.1 f c;t t
D=0.2 0 ¼ f ðD; S; tÞ ¼ ð1Þ
f c;7 a þ bt
Coefficient b

0.9

0.8 a ¼ aðD; SÞ ð2Þ

0.7 b ¼ bðD; SÞ ð3Þ

0.6 a ¼ a1 þ a2 S þ a3 S2 ð4Þ
0.0 0.1 0.2 0.3 0.4
Replace of slag (S) b ¼ b 1 þ b 2 S þ b 3 S2 ð5Þ
0 0
Fig. 7. Relationship between coefficients of hyperbolic function and slag powder where f c;t
refers to the compressive strength at any curing age, f c;7
content. refers to the compressive strength at the age of 7 days, D refers to

1.2 1.0
(a) W/B=0.5 T=750C a1 = a11 + a12*D (d) W/B=0.5 T=750C b1 = b11 + b12*D
a11 =0.9589 b11 =0.8518
1.1 a12=-0.048 b12=-0.037
Coefficient b 1
Coefficient a 1

0.9

1.0

0.8
0.9

0.8 0.7
4.0 (b) W/B=0.5 T=750C a2 = a21 + a22*D 0.4 (e) W/B=0.5 T=750C b2 = b21 + b22*D
a21=2.6594 b21=-0.2017
a22=-16.8645 b22=1.736
Coefficient a 2

Coefficient b 2

3.0

2.0
0.0
1.0

0.0

-1.0 -0.4
2.0 (c) W/B=0.5 T=750C 2.0 (f) W/B=0.5 T=750C b3 = b31 + b32*D
b31=0.5342
b32=-3.503
Coefficient a 3

Coefficient b 3

0.0
1.0
-2.0

-4.0
a3 = a31 + a32*D 0.0
-6.0 a31=-6.9105
a32=37.817
-8.0 -1.0
0.0 0.1 0.2 0.0 0.1 0.2

Replace of Desulfurization Slag (D) Replace of Desulfurization Slag (D)

Fig. 8. Relationship between of coefficients of a and b versus desulfurization slag content.

114 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117

the replacement rate of desulfurization slag, S refers to the replace- 70

Compressive Strength (MPa)

D=0 W/B=0.5 T=75 °C
ment rate of slag powder and a1, a2, a3 and b1, b2, b3 are model coef-
60 D=0.1 Age=7 day
ficients related to the replacement rate of slag powder S.
D=0.2 f 'c,7= x + θ*S
Fig. 8(a) to (c) shows the relationship between the model coef- 50 D=0.4
ficients a1, a2, a3 and the replacement rate of desulfurization slag D.
The relationship between coefficients a1, a2 and D could roughly be 40
represented as a decreasing function, while that between coeffi-
30
cient a3 and D could roughly be represented as an increasing func-
tion. To simplify the analysis, coefficientsa1,a2,a3 and D are 20
assumed to have an approximately linear relationship during the
derivation of the model. a1, a2, a3 could therefore be expressed as 10
Eq. (6). Coefficients b1 , b3 and D and coefficient b2 and D have rela- 0.0 0.1 0.2 0.3 0.4
tionships that are assumed to be roughly similar to decreasing and Replace of slag (S)
increasing functions, respectively. Likewise, when coefficients b1,
Fig. 9. Compressive strength on Day 7 versus slag powder content.
b2, b3 and D are assumed to have a linear relationship, b1, b2, and
b3 could be represented as Eq. (7). After Eq. (4) and Eq. (7) are com-
bined, Eq. (1), the equation for the prediction of the compressive
60
strength of the desulfurization slag cement mortar, could be (a) W/B=0.5 T=750C
2
rewritten as Eq. (8). Therefore, as long as the compressive strength 50 x = x1 + x2*D + x3*D
0
on the 7th day f c;7 is obtained, it is possible to predict the compres- x1=41.3324

Coefficient x
sive strength at any other age and under other mixture ratios using 40
x2=-100.1018
Eq. (8). Table 5 shows the values of the model coefficients a11, a12, x3=103.8182
a21, a22, a31, a32, b11, b12, b21, b22, b31, b32, etc. that are obtained by 30
regressing the results of the test samples cured at high tempera-
ture in the function. 20

a1 ¼ a11 þ a12 D
10
a2 ¼ a21 þ a22 D ðfor coefficient aÞ ð6Þ 20
(b) W/B=0.5 T=750C
θ = θ1 + θ2*D + θ3*D
2
a3 ¼ a31 þ a32 D
θ1=8.5332
Coefficient θ

15
b1 ¼ b11 þ b12 D θ2=28.8485
θ3=-91.643
b2 ¼ b21 þ b22 D ðfor coefficient bÞ ð7Þ 10
b3 ¼ b31 þ b32 D
5
0 0
f c;t ¼ ða 2
t
2 f c;7
1 þa2 Sþa3 S Þþðb1 þb2 Sþb3 S Þt
0
¼ ½ða 2
t
þ½ðb11 þb12 DÞþðb21 þb22 DÞSþðb31 þb32 DÞS2 t f c;7
0
11 þa12 DÞþða21 þa22 DÞSþða31 þa32 DÞS
0.0 0.1 0.2 0.3 0.4
ð8Þ
Replace of Desulfurization Slag (D)
According to the results of the test samples cured at high tem-
perature, the relationship between the compressive strength of the Fig. 10. Relationship between of coefficients of x and h versus desulfurization slag
content.
desulfurization slag cement mortar on Day 7 and the replacement
rate of slag powder S could be roughly expressed as a linearly
increasing function. As the replacement rate of desulfurization slag with further increases in the value of D. This trend could also be
D increases, the compressive strength will decrease, as is shown in simulated as a non-linear quadratic polynomial function, as is
0
Fig. 9. It is therefore possible to assess f c;7 , the compressive shown in Eq. (11). Table 5 shows the values of the model coeffi-
strength on Day 7, through Eq. (9). Fig. 10 shows the relationship cients x1, x2, x3, h 1, h 2, h 3, etc.
between the model coefficients x and h and the replacement rate 0
f c;7 ¼ x þ h  S ð9Þ
of desulfurization slag D. As shown in the figure, the relationship
between coefficient x and D could be roughly represented as a
x ¼ x1 þ x2 D þ x3 D2 ð10Þ
non-linear quadratic polynomial decreasing function, as is shown
in Eq. (10). The value of coefficient h first increases with the value
of D, but after the value of D exceeds a certain point, h decreases h ¼ h1 þ h2 D þ h3 D2 ð11Þ

Table 5
The coefficient values for compressive strength prediction model.

a b x&h
T = 75 °C T = 23–25 °C T = 75 °C T = 23–25 °C T = 75 °C T = 23–25 °C
a11 0.959 1.155 b11 0.852 0.847 x1 41.332 38.394
a12 -0.048 0.057 b12 -0.037 -0.011 x2 100.102 43.211
a21 2.659 2.755 b21 0.202 0.627 x3 103.818 32.909
a22 16.865 15.975 b22 1.736 0.760 h1 8.533 3.823
a31 6.911 5.575 b31 0.534 1.318 h2 28.849 29.828
a32 31.817 48.956 b32 3.503 1.272 h3 91.643 25.714

Note: T is stand for temperature.

 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117 115

When predicting the compressive strength at other ages and strengths, besides conducting a regression analysis on the relevant
under different mixture ratios, the necessary value of the compres- influencing parameters. To validate the forecasting capacities of
0
sive strength on Day 7 f c;7 could therefore be calculated with Eq. the compressive strength prediction model established in this
(9). The results of the room temperature curing tests (23–25 °C) study, the test results generated when the replacement rate of
demonstrate similar trends to those on the samples cured at high desulfurization slag D = 40% are predicted and compared. Fig. 11
temperature. The compressive strength prediction model estab- compares the test results with the compressive strength values
lished in this study is therefore also applicable to samples cured at different curing ages obtained with the prediction model estab-
at room temperature. Table 5 shows the relevant model coeffi- lished in this study when the replacement rate of desulfurization
cients when the samples are cured at room temperature. slag D = 40% while the slag powder content changes, cured at
either high temperature or room temperature. According to the
5. Comparison between prediction and test result results, when the samples are cured at high temperature, the com-
pressive strength values at different curing ages obtained in the
In the course of deriving the model, this study only considers prediction based on the Day 7 early strength are quite close to
the test results generated when the replacement rates of desulfur- the rest results, with an MAPE (mean absolute percentage error)
ization slag D are 0%, 10% or 20% and examines their compressive value of 6.13%, as is shown in Fig. 11(a). For the room temperature

50
0 (b) W/B=0.5 D=0.4
(a) W/B=0.5 D=0.4 T=75 C
S=0% S=0% Room Temp.(23-250C)
40 S=10%
S=10%
S=20% S=20%
f 'c (MPa)

30 S=40% S=40%
calculated measured
20

10

0
3 7 28 56 91 3 7 28 56 91
Curing Age (day) Curing Age (day)
f 'c,calculated (MPa)

40

30

20

MAPE 6.13% MAPE 11.32%

10
10 20 30 40 10 20 30 40
f 'c,measured (MPa) f 'c,measured (MPa)

Fig. 11. Comparison of compressive strength and MAPE value for calculated by proposed model and tested results (D = 40%).

0 0
60 (a) W/B=0.5 D=0 T=75 C MAPE=3.05% (b) W/B=0.5 D=0.1 T=75 C MAPE=4.45%

50

40
Compressice Strength (MPa)

S=0%
30 S=0%
S=10%
S=10%
S=20%
20 S=20%
S=40%
S=40%
calculated by this study
10 calculated by this study

0 0 0
60 (c) W/B=0.5 D=0.2 T=75 C MAPE=4.47% (d) W/B=0.5 D=0.4 T=75 C MAPE=6.13%
S=0%
50
S=10%
S=20%
40
S=40%
calculated by this study
30
S=0%
20 S=10%
S=20%
10 S=40%
calculated by this study
0
0 30 60 90 0 30 60 90
Curing Age (day) Curing Age (day)

Fig. 12. Comparison of predicted model and tested results for compressive strength (cured at high temperature).
116 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117

curing, the analytical results within 28 days are also very reason- (a) 0
60 W/B=0.5 T=75 C
able, with an MAPE value of 6.24%. However, when the curing
age is greater than 56 days, the values generated in the analysis 50
tend to be significantly underestimated, with the MAPE value
being as high as 18.94%. The main reason for this phenomenon is 40
that room temperature curing will not impose much influence on
MAPE
the dilation features of the desulfurization slag or bring significant 30 D=0 3.05%
changes to the strength. However, at this point in time, the influ-

(MPa)
D=0.1 4.45%
ence of the slag powder on the late strength is still there and is 20 D=0.2 4.47%
gradually increasing. As a result, the compressive strength will D=0.4 6.13%

f c,calculated
10
continue to increase after 28 days of curing age. The compressive (b)
60 W/B=0.5 Room Temp.(23-25 0C)
strength at the age of 91 days, for example, has increased by

'
approximately 25%–40% compared to that on the 28th day. 50
Another reason is that as the regression analysis in this study uses
only the test results within 28 days of curing age to obtain the rel- 40
evant model parameters and the early strength on Day 7 is used as
MAPE
the basis of the prediction, the inaccuracy of using the model to 30
D=0 9.10%
predict the compressive strength of samples that continue to D=0.1 7.83%
20
develop at later stages will be relatively high. For the room temper- D=0.2 9.17%
ature curing, though, the overall MAPE value for the analytical D=0.4 11.32%
10
results is 11.32%, which is quite satisfactory. 10 20 30 40 50 60
As suggested in Lewis (1982) [18], when the MAPE value is less '
f c,measured (MPa)
than 10%, it indicates an excellent predictive ability; when the
MAPE value is 10%–20%, it indicates a good predictive ability; when Fig. 14. Comparison of predicted model and measured compressive strength.
the MAPE value is 20%–50%, it indicates a reasonable predictive
ability, and when the MAPE value is greater than 50%, it indicates
a poor predictive ability. Although the analysis of the compressive
strength of the desulfurization slag cement mortar cured at room Fig. 14(a). As they are all less than 10%, they indicate an excellent
temperature at older ages is relatively inaccurate, the overall MAPE predictive ability. For the results of the test samples cured at room
value still falls in the range of good predictive ability. temperature, when the replacement rates of desulfurization slag D
Figs. 12 and 13 show and compare the results of the compres- are 0, 10%, 20% and 40%, the MAPE value is 9.10%, 7.83%, 9.17% and
sive strength tests conducted at high temperature and room tem- 11.32%, respectively, as shown in Fig. 14(b). Although the MAPE
perature at different curing ages based on the early strength on value is greater than 10% when D = 40%, the overall MAPE value
Day 7 when the replacement rate of the desulfurization slag is still approximately 10%, also indicating an excellent predictive
changes. According to the figures, the compressive strength predic- ability. As shown in Fig. 13, the predicted values within 28 days
tion model established in this study could evaluate the relationship when the test samples are cured at room temperature are also
between the compressive strength and the curing age when the quite close to the test results. Beyond 28 days, however, the pre-
replacement rates of the desulfurization slag and slag powder dicted values become significantly inaccurate. The causes of this
change. According to the analytical results, when the replacement phenomenon are the same as those leading to the trend when
rates of desulfurization slag D are 0, 10%, 20% and 40%, the MAPE D = 40% at room temperature. In conclusion, the prediction model
value is 3.05%, 4.45%, 4.47% and 6.13%, respectively, as shown in established in this study to predict the compressive strength of the

60 (a) W/B=0.5 D=0 T=23-250C (b) W/B=0.5 D=0.1 T=23-250C

MAPE=7.83%
50

40
Compressice Strength (MPa)

MAPE=9.10%
30 S=0% S=0%
S=10% S=10%
20 S=20% S=20%
S=40% S=40%
10 calculated by this study calculated by this study

0
60 (c) W/B=0.5 D=0.2 T=23-250C (d) W/B=0.5 D=0.4 T=23-250C
MAPE=9.17% MAPE=11.32%
50

40

30
S=0% S=0%
20 S=10% S=10%
S=20% S=20%
10 S=40% S=40%
calculated by this study calculated by this study
0
0 30 60 90 0 30 60 90
Curing Age (day) Curing Age (day)

Fig. 13. Comparison of predicted model and tested results for compressive strength (cured at room temperature).
 C.-C. Wang / Construction and Building Materials 128 (2016) 108–117
desulfurization slag cement mortar from the early strength on Day
7 is proved to be able to deliver satisfactory results. However, if the
strength continues to increase significantly at later stages, the inac-
curacies will be significant. Given the mixture ratios in this study,
however, these inaccuracies could still be called reasonable.
6. Conclusions
1. On the basis of a hyperbolic function and combining the prop-
erties of desulfurization slag and slag powder, when the
water-binder ratio is fixed, a model is derived in this study that
could predict the compressive strength at other curing ages
from the early strength on Day 7. This model will contribute
greatly to the safety assessment of concrete structures during
the construction phase, when it could thus serve as a reference
for the mixture ratio design of desulfurization slag and slag
powder in future engineering practice. It still awaits further
tests and confirmatory analyses, however, to see whether the
compressive strength prediction model established in this
paper is applicable to other mixture ratios and materials.
2. From the comparison of the compressive strength values gener-
ated in the tests and the prediction model, it is noted that the
desulfurization slag cement mortar test samples cured at high
temperature has an MAPE value between 3.05% and 6.13%,
while the MAPE value of the test sample cured at room temper-
ature falls in 7.83%-11.32%. Most of the MAPE values are less
than 10%. Therefore, the prediction model established in this
study has a good predictive ability under both circumstances.
3. According to the test results, both for the test samples cured at
high temperature and room temperature, the compressive
strength of the cement mortar will go down as the replacement
rate of desulfurization slag increases. However, this strength
will go up as the replacement rate of slag powder increases.
At later stages, in particular, the contribution of the slag powder
to the strength is quite significant. Therefore, at room tempera-
ture, the influence on the dilation features of desulfurization
slag is quite limited, leading to the insignificant changes to
the strength. Under the same circumstances, however, the
influence of the slag powder on the late strength is still there
and gradually increasing, leading to the significant upward
trend in the compressive strength of the test samples cured at
room temperature after 28 days (shown as Fig. 12).
4. Test results shown that the workability of the cement mortar
will be better when desulfurization slag is used to replace fine
aggregate. However, the workability will decrease as the
117
replacement rate of slag powder increases. Moreover, the set-
ting time has an increasing tendency when the replacement
rate of the slag powder or desulfurization slag goes up.
References
[1] W.T. Kuo, M.W. Weng, Utilization of desulfurization/granulated blast furnace
slag as controlled low strength material without Portland cement, J. C.C.I.T 38
(1) (2009) 157–166.
[2] W.T. Kuo, C.Y. Shu, Application of high-temperature rapid catalytic technology
to forecast the volumetric stability behavior of containing steel slag mixtures,
Constr. Build. Mater. 50 (2014) 463–470.
[3] W.T. Kuo, Properties of compressed concrete paving units made produced
using desulfurization slag, Environ. Prog. Sustainable Energy 34 (5) (2015)
1365–1371.
[4] W.T. Kuo, H.Y. Wang, C.Y. Shu, Engineering properties of cementless concrete
produced from GGBFS and recycled desulfurization slag, Constr. Build. Mater.
63 (2014) 189–196.
[5] T. Yen, C.W. Tang, K.H. Chen, J.C. Chen, Effects of partial replacement cement
using blast furnace slag on the rheological property of flowing concrete, J. Chin.
Inst. Civil Hydraul. Eng. 17 (1) (2005) 1–9.
[6] B.C. Lho, M.K. Joo, K.H. Choi, J.Y. Choi, Effects of polymer-binder ratio and slag
content on strength properties of autoclaved polymer-modified concrete, KSCE
J. Civil Eng. 16 (5) (2012) 803–808.
[7] C.T. Chen, Effect of engineering properties on cement mortar with slag and
desulfurization slag (Master dissertation), National Kaohsiung University of
Applied Sciences, Kaohsiung, 2016.
[8] C. Lee, S. Lee, N. Nguyen, Modeling of compressive strength development of
high-early-strength-concrete at different curing temperatures, Int. J. Concr.
Struct. Mater. 10 (2) (2016) 205–219.
[9] K.H. Yang, J.S. Mun, M.S. Cho, Effect of curing temperature histories on the
compressive strength development of high-strength concrete, Adv. Mater. Sci.
Eng. (2015) 1–12.
[10] A. Kabir, M. Hasan, M.K. Miah, Predicting 28 days compressive strength of
concrete from 7 days test result, Proc. of Int. Conf. Adv. Des. Constr. Struct.
(2012) 18–22.
[11] G.F. Kheder, A.M.A. Gabban, S.M. Abid, Mathematical model for the prediction
of cement compressive strength at the ages of 7 and 28 days within 24 hours,
Mater. Struct. 36 (2003) 693–701.
[12] M. Benaicha, Y. Burtschell, A.H. Alaoui, Prediction of compressive strength at
early age of concrete–application of maturity, J. Build. Eng. 6 (2016) 119–125.
[13] V.K. Alilou, M. Teshnehlab, Prediction of 28-day compressive strength of
concrete on the third day using artificial neural networks, Int. J. Eng. 3 (6)
(2010) 565–576.
[14] T.R. Neelakantan, S. Ramasundaram, R. Shanmugavel, R. Vinoth, Prediction of
28-day compressive strength of concrete from early strength and accelerated
curing parameters, Int. J. Eng. Technol. 5 (2) (2013) 1197–1201.
[15] K. Hwang, T. Noguchi, F. Tomosawa, Prediction model of compressive strength
development of fly-ash concrete, Cem. Concr. Res. 34 (2004) 2269–2276.
[16] C.C. Wang, T.T. Chen, H.Y. Wang, C. Huang, A predictive model for compressive
strength of waste LCD glass concrete by nonlinear-multivariate regression,
Comput. Concr. 13 (4) (2014) 531–545.
[17] C.C. Wang, H.Y. Wang, C. Huang, Predictive models of hardened mechanical
properties of waste LCD glass concrete, Comput. Concr. 14 (5) (2014) 577–597.
[18] C.D. Lewis, Industrial and Business Forecasting Method, Butterworth Scientific
Publishers, London, 1982.