Clean Technologies and Environmental Policy

https://doi.org/10.1007/s10098-022-02318-w(0123456789().,-volV)(0123456789().,-volV)

ORIGINAL PAPER

Soft computing techniques to predict the compressive strength

of green self-compacting concrete incorporating recycled plastic
aggregates and industrial waste ashes
Rabar H. Faraj1,2 • Azad A. Mohammed1 • Khalid M. Omer3 • Hemn Unis Ahmed1

Received: 27 November 2021 / Accepted: 26 March 2022

 The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022

Abstract
Rapid urbanization and industrialization with corresponding economic growth have increased concrete production, leading
to resource depletion and environmental pollution. The mentioned problems can be resolved by using recycled aggregates
and industrial waste ashes as natural aggregate and cement replacement in concrete production. Incorporating different by-
product ashes and recycled plastic (RP) aggregates are viable options to produce sustainable self-compacting concrete
(SCC). On the other hand, compressive strength is an essential characteristic among other evaluated properties. As a result,
establishing trustworthy models to forecast the compressive strength of SCC is critical to saving cost, time, and energy.
Furthermore, it provides valuable instruction for planning building projects and determining the best time to remove the
formwork. In this study, four alternative models were suggested to predict the compressive strength of SCC mixes
produced by RP aggregates: the artificial neural network (ANN), nonlinear model, linear relationship model, and multi-
logistic model. To do so, an extensive set of data consisting of 400 mixtures were extracted and analyzed to develop the
models, various mixture proportions and curing times were considered as input variables. To test the effectiveness of the
suggested models, several statistical evaluations, including coefficient of determination (R2), scatter index, root mean
squared error (RMSE), mean absolute error (MAE), and Objective (OBJ) value were utilized. Compared to other models,
the ANN model performed better to forecast the compressive strength of SCC mixes incorporating RP aggregates. The
RMSE, MAE, OBJ, and R2 values for this model were 5.46 MPa, 2.31 MPa, 4.26 MPa, and 0.973, respectively.
Graphical Abstract

& Rabar H. Faraj

rabar.faraj@uoh.edu.iq
1
Civil Engineering Department, College of Engineering,
University of Sulaimani, Sulaimani, Kurdistan Region, Iraq
2
Civil Engineering Department, University of Halabja,
Halabja, Kurdistan Region, Iraq

123
 R. H. Faraj et al.

Keywords Recycled plastic aggregates  Industrial waste ashes  Compressive strength  Curing time  Modeling 
Sensitivity

Introduction The recycling of plastic wastes can be regarded as one

of the optimum solutions for reducing plastic waste’s
Innovative and high-performance materials are necessary influence on the environment in terms of energy con-
to develop sustainable and long-lasting infrastructure. sumption and natural resource, waste disposal, global
Because concrete is the most commonly used construction warming, and environmental pollution. The reuse of plastic
material, researchers and the construction industry have waste in the building sector is an ideal option for disposing
made it as durable as possible by introducing new materials of plastic waste among the various forms of recycling
and methods. Among developed high-performance cement- management approaches (Sadrmomtazi et al. 2016). As
based materials, one can mention self-compacting concrete previously mentioned, to decrease the consumption of
(SCC). The development of SCC can be regarded as a natural aggregates, recycled plastic (RP) aggregates can be
historical achievement in the construction and concrete used effectively to produce cement-based materials such as
industry because of its numerous benefits compared to SCC. In the literature, different types of plastics such as
traditional concrete composites. It is a kind of high-per- Polyethylene Terephthalate (PET) (Mohammed et al.
formance concrete that can flow efficiently without the 2019), expanded polystyrene (EPS) (Aslani & Ma 2018),
need for any compaction or mechanical work (Aslani et al. polypropylene (PP) (Yang et al. 2015), and High Impact
2018). This concrete provides high passing and filling polystyrene (HIPS) (Chunchu and Putta 2019) were used in
ability performance compared to other cement-based the production of SCC. RP aggregates used in the previous
composites without causing segregation and bleeding investigations were replaced by volume to substitute fine or
(Aslani et al. 2018). The passing ability and flowability natural coarse aggregates. Generally, the mechanical
characteristics of SCC can be obtained through the addition properties of SCC, especially its strength was decreased
of superplasticizer (high range water reducer) along with with elevating the percentage of RP aggregate content;
incorporating different supplementary cementitious mate- however, this negative effect of RP aggregates can be
rials (SCMs) such as ground granulated blast furnace slag diminished through incorporating other supplementary
(GGBFS), metakaolin, silica fume, and fly ash to maintain cementitious materials like silica fume, GGBFS, and fly
the stability and proper cohesiveness during the flow of the ash. Moreover, to maintain the stability of SCC containing
mixture (Deilami et al. 2019). The most significant RP aggregates during its flow, a high amount of paste is
advantages of SCC compared to traditional concrete required to thoroughly coat the granular materials (aggre-
include shorter periods for construction, minimizing the gates) (Faraj et al. 2019). Therefore, incorporating other
labor costs, and improved compaction in the structure, cementitious materials to decrease the amount of cement
especially at very congested locations when a high amount consumption is highly recommended.
of rebars are present in the concrete elements. On the other hand, it is well proved that the manufacture
Aggregate comprises about 65–80% of the concrete of cement necessitates a significant amount of energy and,
volume, and it can provide such excellent properties of at the same time, contributes to the direct or indirect
concrete as strength, permeability, volume stability, emission of a large quantity of total carbon dioxide (about
workability, and durability (Faraj et al. 2019). Considerable 7%) into the environment (Yu 2019). Furthermore, one ton
quantities of fine and coarse aggregates are required to of cement requires around 2.8 tons of raw materials; this is
produce large amounts of concrete for global consumption a resource-depleting process that uses various natural raw
(Spiesz et al. 2016). Large amounts of waste can be avoi- materials such as limestone and shale to manufacture
ded by reusing recycled materials to prepare new concrete. clinkers cement (Guo et al. 2010). By-products or other
Environmental issues about aggregate mining and waste waste materials, on the other hand, will increase land
disposal and aggregate shortages on construction sites can occupancy and destroy the environment. Consequently,
be overcome by this strategy (Saikia and Brito 2013). reusing waste materials has a minimal impact on the
Due to the current need for human activities, many environment (Shahbazpanahi et al. 2021).
different types of plastics are produced every year. How- To decrease the negative environmental impacts of
ever, the majority of the plastic types are produced for a cement production, replacing it with other cementing
single-use. Therefore, if not recycled correctly and due to materials that have less impact on the environment is
the very low biodegradability of plastic waste, they cause essential. This can help to produce more sustainable and
environmental pollution. cleaner concrete with improved performance. GGBFS,

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

silica fume, and fly ash are various by-product powders that recycled aggregate concrete, and they found that an artifi-
can be used as cement replacements for SCC production. cial neural network (ANN) is a promising algorithm to
These by-product ashes can be used effectively with the predict the CS of recycled concrete using various mixtures
combination of RP aggregates in the SCC mixtures to proportion and types of recycled aggregates. In the same
improve the mechanical features of SCC mixtures that context, Deshpande et al. (2014) revealed that the ANN
were reduced by incorporating RP aggregates (Sadrmom- model performed better than the model tree and linear
tazi et al. 2016). regression analysis in predicting the CS of concrete with
Compressive strength (CS) is an essential characteristic recycled aggregate concrete.
in constructing engineering structures among the different Mohammed et al. (2020a) established multiscale
properties of SCC. Other durability and mechanical approaches to simulate the CS of concrete containing high
parameters are related to CS and can be estimated from quantities of fly ash. In their study, 450 samples were
indirect relationships with CS (Neville 1995). Multiple utilized for modeling. The qualifications were developed
cubic and cylinder samples are currently made and evalu- using five distinct modeling techniques (Linear regression,
ated at various curing durations to figure out the CS of SCC nonlinear regression, Multi-logistic regression ANN, and
in practice. In general, work on a construction site should M5P-tree). They depicted that the M5P-tree and ANN
not continue until the CS test results are obtained at a models could forecast the CS very well in terms of lower
particular age, such as 28 days. This causes construction RMSE and MAE values and higher R2 values. A similar
projects to be delayed; the testing process is also time- approach was used to predict the effect of large volume fly
consuming and costly (Shariati et al. 2020). ash on the CS of cement-based mortars at various curing
Because changing mix proportions and components can periods and w/c ratios, with findings comparable to the
have a considerable impact on the characteristics of SCC, prior work (Salih et al. 2020).
determining its CS without conducting experimental tests Mohammed et al. (2020b) also employed ANN and
has always been one of the challenges in concrete tech- nonlinear approaches to predict the rheological perfor-
nology (Shariati et al. 2020). This is especially noticeable mance and CS of nano clay-modified cement paste. They
in SCC, where pozzolanic materials like GGBFS, lime- concluded that the nonlinear approach is the best-per-
stone powder (LP), fly ash, and silica fume has been used formed method to estimate the flow behavior and CS of
to replace cement partially, and RP aggregates are incor- cement paste, and it outperforms the ANN model. They
porated to replace natural aggregates. also stated that, among numerous independent factors,
Since CS is sensitive to mixture proportions and nano clay content was the most critical factor in deter-
depends on several parameters, more improved approaches mining CS and cement paste rheological behavior.
should be employed to reduce the necessity for experi- Despite the widespread usage of RP aggregates in the
mental tests in the laboratory as much as possible and SCC, studies regarding the prediction of CS of SCC
afford engineers with more straightforward methods and incorporated RP aggregates are very scarce to be efficiently
mathematical formulas for forecasting experimental out- implemented by the construction industry. Furthermore,
comes. Soft computing techniques might be regarded as an the construction industry’s growing desire for innovative
appropriate solution in this regard. The most significant building materials with unique characteristics to extend the
advantage of these approaches is that they may be used to service life of concrete structures necessitates the devel-
generate alternatives and solutions for linear and nonlinear opment of creative models for forecasting the behavior of
issues when mathematical models cannot easily describe these new materials. As a result, the main goal of this study
the relationship between the problem’s relevant factors is to assess and quantify the impact of a broad range of
(Gao et al. 2019). Artificial intelligence approaches for mixture proportions on the CS of SCC from an early age
evaluating and predicting the mechanical characteristics of (7 days) to late curing (400 days), including RP aggregates
cement-based materials are a hot topic in the cement-based content, binder content (Limestone powder or GGBFS or
composites research field to provide the construction fly ash or silica fume or their combination), natural fine and
industry implementation new methods and techniques. coarse aggregate content, w/b ratio, and superplasticizer
Moreover, some researchers have applied machine learning (SP) content. Using 400 data from past investigations,
methods to evaluate and forecast the compressive strength several model approaches such as multi-logistic, ANN,
of different concrete types. In this regard, multi-linear linear, and nonlinear regression models were used to pre-
regression analysis and ANN models are extensively used dict the CS of SCC, including RP aggregates.
among various dissimilar models to estimate the CS of
concrete (Ghafor et al. 2020).
A previous study conducted by Duan et al. (2013) used
168 datasets from the literature to forecast the CS of

123
 R. H. Faraj et al.

Research significance w/b ratio (%), SP content, natural fine aggregate (FA)
content, natural coarse aggregate (CA) content, curing time
The current study aims to develop, describe, and provide (t) by days, and measured CS (MPa). The given data set,
multiscale models for predicting the CS of sustainable and which included the above-mentioned independent factors,
eco-friendly SCC containing RP aggregates. Extensive was utilized to forecast the CS of SCC produced with RP
experimental data, including 400 tested results with vary- using several approaches compared to the measured
ing RP aggregate concentrations, w/b ratios, and curing reported CS (MPa). Figure 1 depicts the procedure used in
regimes, were utilized in addition to the various modeling this investigation in terms of a flowchart. In addition, the
methods to achieve the following goals: (i) to perform following sections describe and explain the specifics, such
statistical analysis and investigate the impact of mix as data gathering, analysis, modeling, and assessment.
ingredients such as RP aggregates, natural coarse and fine
aggregates, binder, and SP dosage, as well as the curing
time and w/b ratio on the CS of green SCC produced with Statistical evaluation
RP aggregates and different by-product ashes; (ii) to ensure
that the building sector may use the generated models In this part, a statistical study was performed to determine
without the need for any laboratory testing or analytical whether or not there are substantial correlations between
constraints; (iii) to choose and evaluate the accurate model input parameters and CS of SCCs. The statistical analysis
to forecast the CS of sustainable SCC incorporated RP was performed to illustrate that the CS of green SCC is
aggregates among various models (ANN, nonlinear, linear, affected by all mixture proportions, and a single input
and multi-logistic models) using different statistical eval- parameter cannot be used to estimate the CS directly. To do
uating tools. Furthermore, the study’s primary contribution so, all input variables such as (i) RP aggregates content
is that it provides mathematical models to forecast the CS (Fig. 2), (ii) binder content (Fig. 3), (iii) w/b (Fig. 4), (iv)
of a new composite type, such as SCC incorporated RP curing time (Fig. 5), (v) SP content (Fig. 6), (vi) natural FA
aggregates, to be used efficiently by the construction content (Fig. 7), and (vii) natural CA content (Fig. 8) were
industry. plotted and analyzed with actual CS; additionally, the
normal distribution of obtained CS from previous studies is
shown in Fig. 9. All the figures mentioned previously
Methodology demonstrated that strong correlations among CS and vari-
ous input variables did not exist due to the low R2 values
Totally 400 experimental data from earlier studies were for all correlations. In addition, statistical functions such as
gathered and statistically evaluated before being divided minimum, maximum, average, standard deviation, skew-
into three groups. The first and more extensive group ness, kurtosis, and variance were calculated and displayed
comprised 280 datasets utilized to create the models. Each in Table 3 to show the distribution of each variable.
with 60 data points, the second and third groups were Regarding the kurtosis parameter, a high negative value
utilized to test and validate the models (Qadir et al. 2019; represents the shorter tails of the distribution relative to the
Shariati et al. 2020). Table 1 summarizes the SCC mixes normal distribution, and a positive value represents the
database and the measured CS of SCC produced with longer tails. A large negative value for the skewness
various mix proportions and RP contents. During the first parameter indicates a long left tail, while a positive value
step of data extraction, the contents of cement, GGBFS, indicates a right tail.
LP, silica fume, and fly ash were extracted separately, and The rheological and strength behavior of SCC mixtures
then they were combined to form a binder content as one is directly related to the type, shape, and surface texture of
input variable. Moreover, Table 2 shows all 400 sample RP aggregates and the type of natural aggregate it replaced.
data extracted to develop the models. Several databases’ Therefore, the type and properties of RP aggregates for all
search portals, such as Google Scholar, Web of Science, previous studies used for developing the models were
Scopus, and Science Direct, were utilized to conduct a extracted and presented in Table 4. It is evident from the
comprehensive literature search as part of the database table that the RP aggregates are mostly incorporated into
preparation process. The majority of prior studies dis- the SCC to replace natural fine aggregates. Ordinary
cussing the effect of RP aggregates on the characteristics of Portland cement (OPC) Type I was the most common type
SCCs were collected, and their data were retrieved based used in producing SCC mixes, and it met the ASTM C 150
on the authors’ searches. standard. With a specific gravity of 3.05 to 3.2, the fineness
Table 2 shows the input data set, which includes the was in the range of 300 m2/kg to 400 m2/kg. Different by-
content of RP aggregates by weight, binder (B) content, product waste ashes such as LP, GGBFS, silica fume, and
fly ash with different proportions and properties, as shown

123
 Table 1 Summary of different SCC mixtures containing RP aggregates and different by-product ashes
References Cement LP Fly ash Silica fume GGBFS w/b (%) Curing SP (kg/ RP Fine Coarse Compressive
content (kg/ (kg/ (kg/m3) (kg/m3) (kg/m3) time m3) aggregates aggregate aggregate (kg/ strength (MPa)
m3) m3) (Days) (kg/m3) (kg/m3) m3)

Mohammed 465 30–150 0 0 0 0.285–0.376 7, 28 7–13.2 0–103 861 800 32.76–73.22

et al. (2019)
Aslani and Ma. 160–180 0 135 34 101 0.45 28 3.15 0–3 751–829 759–781 19.1–35.59
(2018)
Faraj et al. 385–440 0 110 55 0 0.32 28, 90 6.6–8.8 0–56 774–784 619.5–784 61.6–87.4
(2019)
Law et al. (2018) 160–180 0 120–135 30–33 90–101 0.45 7, 28 3.25–4.95 0–3.34 751–976 759–808 19.41–54.2
Ranjbar and 400–434 0 0 44–48 0 0.38–0.44 7–365 2.4–4 0–4.76 517–915 387–800 14.8–60.44
Mousavi.
(2015)
Hilal et al. 364 0 156 0 0 0.36 28 3.6 0–123.2 527–832 818–832 25–53
(2018)
Chunchu and 347.9 0 149.1 0 0 0.36 7–90 5.46–6.05 0–138 517–861.69 758.45 20.37–51.26
Putta. (2019)
Sadrmomtazi 315–450 0 135 45 0 0.43 7–90 5.4–8.3 0–108.3 741.7–850 770 9.89–36.62
et al. (2016)
Yang et al. 369 0 154 0 0 0.36 7, 28 5.3 0–78.5 495–703 300 19.64–25.16
(2015)
Mohammed 220–550 0 55–330 0 55–330 0.25 28 17.9–23.4 0 766–866.1 766–866.1 44.96–83.84
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

et al. (2021)
Mohamed 96–480 0 72–192 24–96 48–384 0.36 7–28 7.2 0 896.1 800 28.133–95.3
(2018)
Güneyisi et al. 180–550 0 55–330 22.5–82.5 33.75–330 0.32–0.44 28–90 2.8–10.67 0 685–826 840–935 30.3–100.9
(2010)
Uysal and Sumer 220–550 55–165 83–193 0 110–330 0.33 7–400 8.8 0 860–869 742–778 55.1–105.7
(2011)

123
 R. H. Faraj et al.

Table 2 Comprehensive experimental results of SCC mixtures containing RP aggregates used in the models
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

1 495 0.353 7 7 20 861 800 46.14

2 495 0.353 28 7 20 861 800 57.15
3 585 0.3 7 11.6 83 861 800 39
4 585 0.3 28 11.6 83 861 800 47.02
5 585 0.3 7 7 20 861 800 64.25
6 585 0.3 28 7 20 861 800 73.22
7 495 0.353 7 11.6 20 861 800 39.74
8 495 0.353 28 11.6 20 861 800 45.79
9 495 0.353 7 11.6 83 861 800 35.36
10 495 0.353 28 11.6 83 861 800 44.58
11 585 0.3 7 11.6 20 861 800 47.97
12 585 0.3 28 11.6 20 861 800 57.07
13 585 0.3 7 7 83 861 800 53
14 585 0.3 28 7 83 861 800 60.64
15 495 0.353 7 7 83 861 800 46.77
16 495 0.353 28 7 83 861 800 54.1
17 540 0.32 7 5.4 52 861 800 44.86
18 540 0.32 28 5.4 52 861 800 61.33
19 540 0.32 7 13.2 52 861 800 32.76
20 540 0.32 28 13.2 52 861 800 40
21 540 0.32 7 9.3 0 861 800 54.74
22 540 0.32 28 9.3 0 861 800 67.93
23 540 0.32 7 9.3 103 861 800 37.5
24 540 0.32 28 9.3 103 861 800 50.02
25 615 0.285 7 9.3 52 861 800 53.14
26 615 0.285 28 9.3 52 861 800 59.2
27 465 0.376 7 9.3 52 861 800 40
28 465 0.376 28 9.3 52 861 800 46
29 540 0.32 7 9.3 52 861 800 41.45
30 540 0.32 28 9.3 52 861 800 51.4
31 540 0.32 7 9.3 52 861 800 41.66
32 540 0.32 28 9.3 52 861 800 46.56
33 430 0.45 28 3.15 2 829 781 19.1
34 450 0.45 28 3.15 3 751 759 35.59
35 550 0.32 28 7.4 0 784 784 78.1
36 550 0.32 90 7.4 0 784 784 84.9
37 550 0.32 28 7.4 14 784 744.5 76.4
38 550 0.32 90 7.4 14 784 744.5 79
39 550 0.32 28 7.1 28 784 705.4 69.6
40 550 0.32 90 7.1 28 784 705.4 75.1
41 550 0.32 28 6.9 42 784 667.3 64.3
42 550 0.32 90 6.9 42 784 667.3 71.7
43 550 0.32 28 6.6 56 784 627.1 61.6
44 550 0.32 90 6.6 56 784 627.1 68.2
45 550 0.32 28 8.8 0 774 774 82.7
46 550 0.32 90 8.8 0 774 774 87.4
47 550 0.32 28 8.7 14 774 735.4 79.9
48 550 0.32 90 8.7 14 774 735.4 83.1

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

49 550 0.32 28 8.5 28 774 696.8 73

50 550 0.32 90 8.5 28 774 696.8 78.4
51 550 0.32 28 8.5 42 774 658.2 68.6
52 550 0.32 90 8.5 42 774 658.2 74.8
53 550 0.32 28 8.4 56 774 619.5 64.8
54 550 0.32 90 8.4 56 774 619.5 69.7
55 400 0.45 7 4.95 0 976 808 31.97
56 400 0.45 28 4.95 0 976 808 54.2
57 450 0.45 7 3.75 0 917 759 35.22
58 450 0.45 28 3.75 0 917 759 50.39
59 400 0.45 7 4.45 1.11 917 808 20.79
60 400 0.45 28 4.45 1.11 917 808 35.98
61 400 0.45 7 4.15 2.23 858 808 19.41
62 400 0.45 28 4.15 2.23 858 808 28.38
63 400 0.45 7 4 3.34 799 808 17.62
64 400 0.45 28 4 3.34 799 808 26.29
65 450 0.45 7 3.25 3.14 751 759 22.58
66 450 0.45 28 3.25 3.14 751 759 28.92
67 444 0.44 7 4 0 915 800 35.18
68 444 0.44 28 4 0 915 800 45.18
69 444 0.44 90 4 0 915 800 49.6
70 444 0.44 180 4 0 915 800 51.2
71 444 0.44 270 4 0 915 800 52.1
72 444 0.44 365 4 0 915 800 52.1
73 444 0.44 7 3.5 1.59 783 662 23.4
74 444 0.44 28 3.5 1.59 783 662 26.7
75 444 0.44 90 3.5 1.59 783 662 28.7
76 444 0.44 180 3.5 1.59 783 662 29.11
77 444 0.44 270 3.5 1.59 783 662 30.12
78 444 0.44 365 3.5 1.59 783 662 30.12
79 444 0.44 7 3.2 2.38 716 594 19.87
80 444 0.44 28 3.2 2.38 716 594 22.15
81 444 0.44 90 3.2 2.38 716 594 22.53
82 444 0.44 180 3.2 2.38 716 594 22.78
83 444 0.44 270 3.2 2.38 716 594 23.16
84 444 0.44 365 3.2 2.38 716 594 23.41
85 444 0.44 7 2.8 3.57 617 491 16.07
86 444 0.44 28 2.8 3.57 617 491 18.8
87 444 0.44 90 2.8 3.57 617 491 19.8
88 444 0.44 180 2.8 3.57 617 491 20.3
89 444 0.44 270 2.8 3.57 617 491 20.5
90 444 0.44 365 2.8 3.57 617 491 20.5
91 444 0.44 7 2.5 4.76 517 387 14.8
92 444 0.44 28 2.5 4.76 517 387 17
93 444 0.44 90 2.5 4.76 517 387 18.7
94 444 0.44 180 2.5 4.76 517 387 19.74
95 444 0.44 270 2.5 4.76 517 387 19.87
96 444 0.44 365 2.5 4.76 517 387 20

123
 R. H. Faraj et al.

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

97 482 0.38 7 4 0 915 800 45.1

98 482 0.38 28 4 0 915 800 57.4
99 482 0.38 90 4 0 915 800 59
100 482 0.38 180 4 0 915 800 60.29
101 482 0.38 270 4 0 915 800 60.44
102 482 0.38 365 4 0 915 800 60.29
103 482 0.38 7 3.5 1.59 783 662 24.89
104 482 0.38 28 3.5 1.59 783 662 31
105 482 0.38 90 3.5 1.59 783 662 32.1
106 482 0.38 180 3.5 1.59 783 662 33
107 482 0.38 365 3.5 1.59 783 662 33.82
108 482 0.38 7 3.2 2.38 716 594 22.39
109 482 0.38 28 3.2 2.38 716 594 26.15
110 482 0.38 90 3.2 2.38 716 594 26.62
111 482 0.38 180 3.2 2.38 716 594 26.77
112 482 0.38 270 3.2 2.38 716 594 27.24
113 482 0.38 365 3.2 2.38 716 594 27.71
114 482 0.38 7 2.8 3.57 617 491 19.4
115 482 0.38 28 2.8 3.57 617 491 22.39
116 482 0.38 90 2.8 3.57 617 491 23.33
117 482 0.38 180 2.8 3.57 617 491 23.95
118 482 0.38 270 2.8 3.57 617 491 24.74
119 482 0.38 365 2.8 3.57 617 491 24.74
120 482 0.38 7 2.5 4.76 517 387 17
121 482 0.38 28 2.5 4.76 517 387 18.9
122 482 0.38 90 2.5 4.76 517 387 20
123 482 0.38 180 2.5 4.76 517 387 20.98
124 482 0.38 270 2.5 4.76 517 387 21.92
125 482 0.38 365 2.5 4.76 517 387 21.92
126 520 0.36 28 3.6 0 832 832 53
127 520 0.36 28 3.6 15.4 798 818 50
128 520 0.36 28 3.6 30.8 692 818 47
129 520 0.36 28 3.6 46.2 675 818 44
130 520 0.36 28 3.6 61.6 655 818 41
131 520 0.36 28 3.6 77 614 818 36
132 520 0.36 28 3.6 92.4 584 818 33
133 520 0.36 28 3.6 107.8 554 818 30
134 520 0.36 28 3.6 123.2 527 818 25
135 497 0.36 7 5.46 0 861.69 758.45 33.8
136 497 0.36 28 5.46 0 861.69 758.45 41.78
137 497 0.36 90 5.46 0 861.69 758.45 51.26
138 497 0.36 7 5.65 34.48 776.18 758.45 29.76
139 497 0.36 28 5.65 34.48 776.18 758.45 40.46
140 497 0.36 90 5.65 34.48 776.18 758.45 49.67
141 497 0.36 7 5.9 69.01 690.13 758.45 28
142 497 0.36 28 5.9 69.01 690.13 758.45 39.6
143 497 0.36 90 5.9 69.01 690.13 758.45 47.4
144 497 0.36 7 5.95 103.51 603.85 758.45 40.75

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

145 497 0.36 28 5.95 103.51 603.85 758.45 38.59

146 497 0.36 90 5.95 103.51 603.85 758.45 27.32
147 497 0.36 7 6.05 138 517.53 758.45 34.46
148 497 0.36 28 6.05 138 517.53 758.45 32.67
149 497 0.36 90 6.05 138 517.53 758.45 20.37
150 450 0.43 7 6.8 0 850 770 29.84
151 450 0.43 28 6.8 0 850 770 36.19
152 450 0.43 60 6.8 0 850 770 38.87
153 450 0.43 90 6.8 0 850 770 39.9
154 450 0.43 7 7 36.1 813.9 770 16.49
155 450 0.43 28 7 36.1 813.9 770 22.03
156 450 0.43 60 7 36.1 813.9 770 23.69
157 450 0.43 90 7 36.1 813.9 770 24.8
158 450 0.43 7 7.4 72.2 777.8 770 15.16
159 450 0.43 28 7.4 72.2 777.8 770 20.25
160 450 0.43 60 7.4 72.2 777.8 770 22.58
161 450 0.43 90 7.4 72.2 777.8 770 23.25
162 450 0.43 7 8.1 108.3 741.7 770 14.27
163 450 0.43 28 8.1 108.3 741.7 770 18.7
164 450 0.43 60 8.1 108.3 741.7 770 21.47
165 450 0.43 90 8.1 108.3 741.7 770 21.92
166 450 0.43 7 6.8 36.1 813.9 770 27.9
167 450 0.43 28 6.8 36.1 813.9 770 33.77
168 450 0.43 60 6.8 36.1 813.9 770 35.8
169 450 0.43 90 6.8 36.1 813.9 770 36.62
170 450 0.43 7 7.7 72.2 777.8 770 19.96
171 450 0.43 26 7.7 72.2 777.8 770 28.82
172 450 0.43 60 7.7 72.2 777.8 770 34.25
173 450 0.43 90 7.7 72.2 777.8 770 35.49
174 450 0.43 7 8.3 108.3 741.7 770 12.31
175 450 0.43 28 8.3 108.3 741.7 770 21.44
176 450 0.43 60 8.3 108.3 741.7 770 26.64
177 450 0.43 90 8.3 108.3 741.7 770 29.56
178 450 0.43 7 5.4 36.1 813.9 770 15.71
179 450 0.43 26 5.4 36.1 813.9 770 26.79
180 450 0.43 28 5.4 36.1 813.9 770 34.55
181 450 0.43 90 5.4 36.1 813.9 770 35.16
182 450 0.43 7 6.3 72.2 777.8 770 10.86
183 450 0.43 26 6.3 72.2 777.8 770 22.3
184 450 0.43 28 6.3 72.2 777.8 770 25.41
185 450 0.43 90 6.3 72.2 777.8 770 29.06
186 450 0.43 7 6.8 108.3 741.7 770 9.89
187 450 0.43 28 6.8 108.3 741.7 770 19.04
188 450 0.43 60 6.8 108.3 741.7 770 23.19
189 450 0.43 90 6.8 108.3 741.7 770 25.74
190 523 0.36 7 5.3 0 703 300 20.27
191 523 0.36 28 5.3 0 703 300 25.16
192 523 0.36 7 5.3 26.2 633 300 20.33

123
 R. H. Faraj et al.

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

193 523 0.36 28 5.3 26.2 633 300 24.72

194 523 0.36 7 5.3 41.8 602 300 22.78
195 523 0.36 28 5.3 41.8 602 300 27
196 523 0.36 7 5.3 52.3 570 300 20.14
197 523 0.36 28 5.3 52.3 570 300 26.86
198 550 0.25 28 23.4 0 866.1 866.1 79.52
199 550 0.25 28 23.4 0 856.9 856.9 71.42
200 550 0.25 28 22.3 0 848.8 848.8 65
201 550 0.25 28 21.7 0 840.2 840.2 60.19
202 550 0.25 28 20.9 0 831.8 831.8 55
203 550 0.25 28 19.5 0 824.1 824.1 51.12
204 550 0.25 28 18.7 0 815.7 815.7 44.96
205 550 0.25 28 23.1 0 840.3 840.3 79.63
206 550 0.25 28 22 0 815.4 815.4 81.68
207 550 0.25 28 21.2 0 790.2 790.2 82.76
208 550 0.25 28 20.1 0 765.3 765.3 83.84
209 550 0.25 28 19 0 740.3 740.3 76.93
210 550 0.25 28 18.2 0 715.1 715.1 70.34
211 550 0.25 28 22.8 0 849.1 849.1 80.71
212 550 0.25 28 21.5 0 832.9 832.9 83.52
213 550 0.25 28 20.6 0 816.1 816.1 86
214 550 0.25 28 19.5 0 799.6 799.6 80.38
215 550 0.25 28 18.4 0 783.1 783.1 73.47
216 550 0.25 28 17.9 0 766 766 65.16
217 480 0.36 7 7.2 0 896.1 800 51
218 480 0.36 28 7.2 0 896.1 800 51.48
219 480 0.36 90 7.2 0 896.1 800 66.08
220 480 0.36 7 7.2 0 896.1 800 46.3
221 480 0.36 28 7.2 0 896.1 800 54.3
222 480 0.36 90 7.2 0 896.1 800 61.325
223 480 0.36 7 7.2 0 896.1 800 42
224 480 0.36 28 7.2 0 896.1 800 50.3
225 480 0.36 90 7.2 0 896.1 800 62.5
226 480 0.36 7 7.2 0 896.1 800 46.225
227 480 0.36 28 7.2 0 896.1 800 50.025
228 480 0.36 90 7.2 0 896.1 800 69.96
229 480 0.36 7 7.2 0 896.1 800 38.6
230 480 0.36 28 7.2 0 896.1 800 43
231 480 0.36 90 7.2 0 896.1 800 61.7
232 480 0.36 7 7.2 0 896.1 800 38.6
233 480 0.36 28 7.2 0 896.1 800 46.4
234 480 0.36 90 7.2 0 896.1 800 56.5
235 480 0.36 7 7.2 0 896.1 800 31.43
236 480 0.36 28 7.2 0 896.1 800 36.7
237 480 0.36 90 7.2 0 896.1 800 55.75
238 480 0.36 7 7.2 0 896.1 800 28.11
239 480 0.36 28 7.2 0 896.1 800 56.7
240 480 0.36 90 7.2 0 896.1 800 72.3

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

241 480 0.36 7 7.2 0 896.1 800 49

242 480 0.36 28 7.2 0 896.1 800 58.5
243 480 0.36 90 7.2 0 896.1 800 81.11
244 480 0.36 7 7.2 0 896.1 800 55.65
245 480 0.36 28 7.2 0 896.1 800 71.45
246 480 0.36 90 7.2 0 896.1 800 95.3
247 480 0.36 7 7.2 0 896.1 800 37.3
248 480 0.36 28 7.2 0 896.1 800 53.41
249 480 0.36 90 7.2 0 896.1 800 75.83
250 480 0.36 7 7.2 0 896.1 800 51
251 480 0.36 28 7.2 0 896.1 800 58.5
252 480 0.36 90 7.2 0 896.1 800 66.75
253 480 0.36 7 7.2 0 896.1 800 52.02
254 480 0.36 28 7.2 0 896.1 800 65.3
255 480 0.36 90 7.2 0 896.1 800 77.53
256 480 0.36 7 7.2 0 896.1 800 45.93
257 480 0.36 28 7.2 0 896.1 800 66.33
258 480 0.36 90 7.2 0 896.1 800 81
259 480 0.36 7 7.2 0 896.1 800 51.55
260 480 0.36 28 7.2 0 896.1 800 65.54
261 480 0.36 90 7.2 0 896.1 800 78.4
262 480 0.36 7 7.2 0 896.1 800 43.5
263 480 0.36 28 7.2 0 896.1 800 56.11
264 480 0.36 90 7.2 0 896.1 800 74
265 480 0.36 7 7.2 0 896.1 800 59.66
266 480 0.36 28 7.2 0 896.1 800 63.95
267 480 0.36 90 7.2 0 896.1 800 75.655
268 480 0.36 7 7.2 0 896.1 800 46.7
269 480 0.36 28 7.2 0 896.1 800 60.42
270 480 0.36 90 7.2 0 896.1 800 62
271 480 0.36 7 7.2 0 896.1 800 41.9
272 480 0.36 28 7.2 0 896.1 800 43
273 480 0.36 90 7.2 0 896.1 800 50.45
274 550 0.32 28 8.43 0 728 935 80.9
275 550 0.32 90 8.43 0 728 935 91.1
276 550 0.32 28 7.43 0 714 917 69.8
277 550 0.32 90 7.43 0 714 917 84.4
278 550 0.32 28 7.43 0 700 899 60.9
279 550 0.32 90 7.43 0 700 899 77.9
280 550 0.32 28 6.67 0 686 881 47.5
281 550 0.32 90 6.67 0 686 881 64.8
282 550 0.32 28 10.43 0 725 931 75.1
283 550 0.32 90 10.43 0 725 931 86.8
284 550 0.32 28 10 0 722 928 80.1
285 550 0.32 90 10 0 722 928 92.6
286 550 0.32 28 8.89 0 720 924 78.1
287 550 0.32 90 8.89 0 720 924 89.2
288 550 0.32 28 9.56 0 724 930 80.4

123
 R. H. Faraj et al.

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

289 550 0.32 90 9.56 0 724 930 91.8

290 550 0.32 28 10.67 0 720 925 85.7
291 550 0.32 90 10.67 0 720 925 99.2
292 550 0.32 28 12 0 716 920 84.4
293 550 0.32 90 12 0 716 920 96.7
294 550 0.32 28 8.22 0 713 916 79.2
295 550 0.32 90 8.22 0 713 916 86.3
296 550 0.32 28 9.11 0 699 898 67.2
297 550 0.32 90 9.11 0 699 898 80.1
298 550 0.32 28 8.89 0 685 880 60
299 550 0.32 90 8.89 0 685 880 68.3
300 550 0.32 28 9.78 0 721 927 79.6
301 550 0.32 90 9.78 0 721 927 96.1
302 550 0.32 28 10.78 0 716 920 87.6
303 550 0.32 90 10.78 0 716 920 95
304 550 0.32 28 10.22 0 710 912 84.5
305 550 0.32 90 10.22 0 710 912 100.9
306 550 0.32 28 8 0 720 924 77
307 550 0.32 90 8 0 720 924 94.9
308 550 0.32 28 7.5 0 711 913 62.3
309 550 0.32 90 7.5 0 711 913 82.8
310 550 0.32 28 4.44 0 703 903 69.4
311 550 0.32 90 4.44 0 703 903 79.2
312 550 0.32 28 6.4 0 717 922 78.6
313 550 0.32 90 6.4 0 717 922 88.4
314 550 0.32 28 6.48 0 707 909 72.7
315 550 0.32 90 6.48 0 707 909 88.9
316 550 0.32 28 8 0 697 896 64.3
317 550 0.32 90 8 0 697 896 79
318 450 0.44 28 3.5 0 826 868 61.5
319 450 0.44 90 3.5 0 826 868 73.6
320 450 0.44 28 3.2 0 813 855 52.1
321 450 0.44 90 3.2 0 813 855 68
322 450 0.44 28 2.96 0 801 842 44.7
323 450 0.44 90 2.96 0 801 842 60.3
324 450 0.44 28 3 0 788 829 30.3
325 450 0.44 90 3 0 788 829 42.5
326 450 0.44 28 3.7 0 823 866 59
327 450 0.44 90 3.7 0 823 866 72.6
328 450 0.44 28 3.44 0 821 863 58
329 450 0.44 90 3.44 0 821 863 74.9
330 450 0.44 28 2.8 0 819 860 56.2
331 450 0.44 90 2.8 0 819 860 65.7
332 450 0.44 28 4.88 0 823 865 60.7
333 450 0.44 90 4.88 0 823 865 71.2
334 450 0.44 28 5.2 0 819 861 58.5
335 450 0.44 90 5.2 0 819 861 76.1
336 450 0.44 28 7.76 0 816 858 71.1

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

337 450 0.44 90 7.76 0 816 858 74.8

338 450 0.44 28 4.24 0 813 855 61.5
339 450 0.44 90 4.24 0 813 855 67.2
340 450 0.44 28 4.52 0 801 841 46.9
341 450 0.44 90 4.52 0 801 841 57.6
342 450 0.44 28 4.82 0 788 828 37.4
343 450 0.44 90 4.82 0 788 828 44.8
344 450 0.44 28 4 0 820 863 60.1
345 450 0.44 90 4 0 820 863 68
346 450 0.44 28 4.56 0 816 857 58.3
347 450 0.44 90 4.56 0 816 857 68.2
348 450 0.44 28 5.76 0 810 852 57.6
349 450 0.44 90 5.76 0 810 852 70.7
350 450 0.44 28 3.2 0 819 860 62.4
351 450 0.44 90 3.2 0 819 860 78
352 450 0.44 28 3.2 0 811 852 53.6
353 450 0.44 90 3.2 0 811 852 69.2
354 450 0.44 28 2.8 0 803 845 45.9
355 450 0.44 90 2.8 0 803 845 60.6
356 450 0.44 28 4.2 0 817 859 60.6
357 450 0.44 90 4.2 0 817 859 76
358 450 0.44 28 4.24 0 808 849 54.7
359 450 0.44 90 4.24 0 808 849 66.8
360 450 0.44 28 5.04 0 799 840 44.2
361 450 0.44 90 5.04 0 799 840 55.2
362 550 0.33 7 8.8 0 869 778 64.9
363 550 0.33 28 8.8 0 869 778 75.9
364 550 0.33 90 8.8 0 869 778 85
365 550 0.33 400 8.8 0 869 778 100.5
366 550 0.33 7 8.8 0 865 762 62
367 550 0.33 28 8.8 0 865 762 74.2
368 550 0.33 90 8.8 0 865 762 87
369 550 0.33 400 8.8 0 865 762 102
370 550 0.33 7 8.8 0 887 752 61.4
371 550 0.33 28 8.8 0 887 752 73.4
372 550 0.33 90 8.8 0 887 752 86.3
373 550 0.33 400 8.8 0 887 752 105.7
374 550 0.33 7 8.8 0 878 742 55.1
375 550 0.33 28 8.8 0 878 742 67.5
376 550 0.33 90 8.8 0 878 742 79.4
377 550 0.33 400 8.8 0 878 742 92.2
378 550 0.33 7 8.8 0 866 775 62.3
379 550 0.33 28 8.8 0 866 775 77.9
380 550 0.33 90 8.8 0 866 775 87.9
381 550 0.33 400 8.8 0 866 775 101.8
382 550 0.33 7 8.8 0 863 772 63.4
383 550 0.33 28 8.8 0 863 772 74.8
384 550 0.33 90 8.8 0 863 772 89.1

123
 R. H. Faraj et al.

Table 2 (continued)
Sample Binder content w/b Curing time SP (kg/ RP aggregate Fine aggregate Coarse aggregate Compressive
(kg/m3) (%) (Days) m3) (kg/m3) (kg/m3) (kg/m3) strength (MPa)

385 550 0.33 400 8.8 0 863 772 96.8

386 550 0.33 28 8.8 0 861 769 71.6
387 550 0.33 90 8.8 0 861 769 80.7
388 550 0.33 400 8.8 0 861 769 85.5
389 550 0.33 7 8.8 0 866 775 60.4
390 550 0.33 28 8.8 0 866 775 69.3
391 550 0.33 90 8.8 0 866 775 80.2
392 550 0.33 400 8.8 0 866 775 86
393 550 0.33 7 8.8 0 863 771 57.4
394 550 0.33 28 8.8 0 863 771 65.2
395 550 0.33 90 8.8 0 863 771 74.2
396 550 0.33 400 8.8 0 863 771 78.6
397 550 0.33 7 8.8 0 860 768 54.8
398 550 0.33 28 8.8 0 860 768 60.2
399 550 0.33 90 8.8 0 860 768 67.2
400 550 0.33 400 8.8 0 860 768 72.4

in Table 1, were also incorporated in SCC production. have a higher R2 value with lower RMSE, Objective (OBJ),
Furthermore, in previous investigations, the natural FA MAE, and SI values.
used was river sand with a maximum size of 4.75 mm and
a specific gravity of 2.65–2.75. Its grade also met the LR model
requirements of ASTM C 33. Based on the past investi-
gations, natural CA for SCC manufacturing was gravel The linear regression model (LR) (Zain and Abd 2009), as
type or crushed stone with a maximum size of 20 mm. illustrated in Eq. (1), is the most general technique for
predicting the CS of concrete:
 
rc ¼ a þ b w=c ð1Þ
Modeling
where a and b rc and w=c, respectively, denote equation
Due to the low R2 value, direct relationships cannot be parameters, CS, and water/cement ratio. Other components
detected among CS, and other ingredients of RP based SCC and variables of SCC mixes incorporated RP aggregates,
mixes such as RP aggregate proportion, binder content, including curing time and other mix contents, are not
natural FA content, w/b ratio, curing time up to 400 days, included in the previous formula, although they impact the
natural CA content, and SP dosage, according to the sta- CS. Equation (2) is presented to incorporate all different
tistical analysis and Figures reported in Sect. 4. As a result, mix proportions and factors that may affect CS to obtain
four distinct models are developed to analyze the influence more reliable scientific findings.
of the various mixture contents indicated previously on the  
rc ¼ a þ bðRPÞ þ cðBÞ þ d w=b þ eðtÞ þ f ðSPÞ þ gðFAÞ
CS of SCC incorporated with RP aggregates, as shown
þ hðCAÞ
below.
The models provided in this work are utilized to forecast ð2Þ
the CS of SCC and choose the best one that provides a where RP stands for recycled plastic aggregate content (kg/
superior estimate of CS compared to the reported CS from m3), B stands for binder content (kg/m3), w/b stands for
the original data. The following evaluation criteria were water to binder ratio, t stands for curing time, SP stands for
used to compare the forecasts of different models: The superplasticizer dosage, FA stands for natural fine aggre-
model had to be scientifically accurate, have a minor per- gate, and CA stands for natural coarse aggregate.
centage error between observed and forecasted data, and

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Fig. 1 The flowchart diagram

Data Collection and Analysis
process followed in this study

Dependent Variable (CS) Independent Variables Statistical Analysis

Recycled Binder W/b Curing SP FA CA

plastic (kg/m3) (kg/m3) (%) time (days) (kg/m3) (kg/m3) (kg/m3)

Relationship between independent and

dependent variables

Data Splitting Randomly

Training Data (70%) Testing Data (15%) Validating Data (15%)

Developing Models

LR NLR MLR ANN

Performance evaluation criteria

R2, RMSE, MAE, SI, OBJ

Sensitivity Analysis for

Independent Variables Using ANN

Furthermore, the model parameters are a, b, c, d, e, f, g, and  d

rc ¼aðBÞC w=b ðtÞe ðSPÞf ðFAÞg ðCAÞh
h. Because all variables may be modified linearly, Eq. (2)  j ð3Þ
can be considered as an expansion of Eq. (1). Although all þ bðBÞi w=b ðtÞk ðSPÞl ðFAÞm ðCAÞn ðRPÞo
of the numerous variables may impact CS and interact with
where B stands for the binder content, w/b stands for the
one another, this may not always be the case. Thus, to
water to binder ratio, t stands for curing time, SP stands for
accurately estimate the CS, the model should be updated
the superplasticizer dosage, FA stands for natural fine
regularly (Zain and Abd 2009; Demircan et al. 2011).
aggregate, CA stands for natural coarse aggregate, and RP
stands for the recycled plastic content. Moreover, the
NLR model model parameters are a, b, c, d, e, f, g, h, I, j, k, l, m, n, and
o, calculated based on the least square method.
The following Eq. (3) can be implemented to build a
nonlinear model in general (Ahmed et al. 2021; Sarwar
MLR model
et al. 2019) To forecast the CS of normal SCC mixes and
SCCs incorporated RP aggregates, the connection among
The MLR, also a regression procedure, can be employed
various variables in Eqs. (1) and (2) can be expressed in
when the expectable variable has a parameter greater than
Eq. (3) to forecast the CS.

123
 R. H. Faraj et al.

300

Mean 17.03
250 StDev 31.85
No. of data 400

200

Frequency
150

100

50

0
0 10 20 30 40 50 60 70 80 90 00 10 20 30 40
1 1 1 1 1
Recycled plastic (kg/m3)

Fig. 2 Variation between compressive strength and RP content with Histogram of SCC mixtures containing RP aggregates

120 Mean 497.6

StDev 46.17
No. of data 400
100

80
Frequency

60

40

20

0
400 425 450 475 500 525 550 575 600 625 650
Binder content (kg/m3)

Fig. 3 Variation between compressive strength and binder content with normal distribution and Histogram of SCC mixtures containing RP
aggregates

two stages. MLR is a statistical approach that is compa- ANN model

rable to multiple linear regressions. Equation (4) can be
utilized to find the variance among predictable and inde- ANN is a powerful simulation software designed for data
pendent variables. analysis and computation to think like a human brain in
 d processing and analyses. This machine learning tool is
rc ¼ aðRPÞb ðBÞC w=b ðtÞe ðSPÞf ðFAÞg ðCAÞh ð4Þ
widely used in construction engineering to predict several
Equation (4), on the other hand, has a drawback in that numerical problems’ future behavior (Sihag et al. 2018;
it cannot be used to forecast the CS of SCC without RP. Mohammed 2018). The ANN model is generally divided
Thus, the RP content in this model should be larger than into three main layers: input, hidden, and output. Each
zero (RP content [ 0%). The least-square method was input and output layer can be one or more layers depending
implemented to find the parameters (a, b, c, d, e, f, g, and h) on the proposed problem. However, the hidden layer is
and model variables. usually ranged for two or more layers. Although the input
and output layers generally depend on the collected data
and the designed model purpose, the hidden layer is
determined by rated weight, transfer function, and the bias
of each layer to other layers. A multi-layer feed-forward

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

150
Mean 0.3690
StDev 0.05460
125 No. of data 400

100

Frequency
75

50

25

0
0.20 0.25 0.30 0.35 0.40 0.45 0.50
W/B ratio (%)

Fig. 4 Variation between compressive strength and w/b ratio with normal distribution and Histogram of SCC mixtures incorporating RP
aggregates

300
Mean 68.19
StDev 88.00
250 No. of data 400

200
Frequency

150

100

50

0
7 28 90 120 150 200 250 300 0 0 0 0
35 40 45 50
Curing time (days)

Fig. 5 Variation between compressive strength and curing time with Histogram of SCC mixtures containing RP aggregates

network is built based on a mixture of proportions, weight/ set, and the rest of the remaining data was used to validate
bias, several parameters, including (RP, B, w/b, t, SP, FA, the trained network (Faraj et al. 2021). The designed ANN
and CA) as inputs, and output ANN here is the CS of SCC. was trained and tested for various hidden layers to deter-
There is no standard approach to designing the network mine optimal network structure based on the fitness of the
architecture. Therefore, the number of hidden layers and predicted CS of SCC containing RP aggregates with the CS
neurons is determined based on a trial and error test. One of of the actual collected data. It was observed that the ANN
the main objectives of the training process of the network is structure with three hidden layers, ten neurons, and a
to determine the optimum number of iterations (epochs) hyperbolic tangent transfer function (as shown in Fig. 10)
that provide the minimum mean absolute error (MAE), root was a best-trained network that provides a maximum R2
means square error (RMSE), and best R2-value that is close and minimum both MAE and RMSE (shown in Table 5).
to one. The effect of several iterations on reducing the The General Equation of the ANN model is shown in
MAE and RMSE has been studied. The collected data set Eqs. 5, 6, and 7.
(a total of 400 data) has been divided into three parts for From linear node 0:
the training purpose of the designed ANN. About 70% of    
Node1 Node 2
the collected data was used as trained data for training the rc ¼ Threshold þ þ þ  ð5Þ
1 þ eB1 1 þ eB2
network. The 15% of overall data was used to test the data

123
 R. H. Faraj et al.

175
Mean 7.127
150 StDev 3.865
No. of data 400

125

100

Frequency
75

50

25

0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
Superplasticizer content (kg/m3)

Fig. 6 Variation between compressive strength and superplasticizer content with Histogram of SCC mixtures incorporating RP aggregates

150
Mean 792.4
StDev 99.33
125 No. of data 400

100
Frequency

75

50

25

0
500 550 600 650 700 750 800 850 900 950 1000
Fine aggregate content (kg/m3)

Fig. 7 Variation between compressive strength and fine aggregate content with Histogram of SCC mixtures containing RP aggregates

0 12
From sigmoid node 1: Pp  0
 0

B p¼1 tp  t yp  y C
B1 ¼ Threshold þ RðAttribute  VariableÞ ð6Þ R2 ¼ B
@rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i h i
C ð8Þ
Pp  2 Pp  2 A
0 0
From sigmoid node 2: p¼1 tp  t p¼1 yp  y

B2 ¼ Threshold þ RðAttribute  VariableÞ ð7Þ sPffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

p  2
p¼1 yp  tp
RMSE ¼ ð9Þ
p
Model assessment tools Pp  

p¼1 yp  tp
MAE ¼ ð10Þ
Different evaluating metrics such as coefficient of deter- p
mination (R2), scatter index (SI), OBJ, root mean squared RMSE
error (RMSE), and mean absolute error (MAE), were SI ¼ ð11Þ
t0
implemented to analyze and assess the effectiveness of the
suggested models, which can be computed using the for-
mulae below:

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

250
225 Mean 764.5
StDev 127.9
200 No. of data 400

175
150

Frequency
125
100
75
50
25
0
300 400 500 600 700 800 900 1000 1100
Coarse aggregate content (kg/m3)

Fig. 8 Variation between compressive strength and coarse aggregate content with Histogram of SCC mixtures incorporating RP aggregates

 
100 ntr RMSEtr þ MAEtr
Mean 52.82 OBJ ¼ 
90 StDev 22.93 nall R2tr þ 1 
80
No. of data 400 ntst RMSEtst þ MAEtst
þ 
70 nall R2tst þ 1 
nval RMSEval þ MAEval
60 þ  ð12Þ
nall R2val þ 1
Frequency

50

40
From the formulas above, yp and tp are the expected and
actual values of the path pattern, and t0 and y0 are the
30
averages of the actual and forecasted values. Training,
20 testing, and validating datasets are denoted as tr, tst, and
10 val, respectively; also the number of patterns (collected
0 data) in the associated dataset is denoted as n.
0 10 20 30 40 50 60 70 80 90 100 110 120
Compressive strength (MPa) Except for R2, the optimum value for all evaluating
factors is zero; nevertheless, R2 has the optimum value of
Fig. 9 Normal distribution of compressive strength in (MPa) for one. When it comes to the SI parameter, a model has (bad
sustainable SCC mixtures containing RP aggregates

Table 3 Statistical parameters for the input and output variables

Code Variable Min Max Average Variance Standard deviation Skewness Kurtosis

1 Recycled plastic (kg/m3) 0 138 17.035 1012.003 31.852 1.932 2.755

2 Binder (kg/m3) 400 615 498 2126.06 46.167 0.1239 - 1.190
3 W/b 0.25 0.45 0.369 0.002974 0.055 - 0.06488 - 0.8309
4 Curing time (days) 7 400 68.18 7724.567 88 2.5392 6.1791
5 Superplasticizer (kg/m3) 2.5 23.4 7.127 14.9025 3.865 2.0176 5.5875
6 Fine aggregate (kg/m3) 517 976 792.44 9842.4 99.333 - 1.0117 0.6343
3
7 Coarse aggregate (kg/m ) 300 935 764.5 16,329.547 127.947 - 1.9066 3.8482
8 Compressive strength (MPa) 9.89 105.7 52.819 524.6398 22.934 - 1.0542 22.9050

123
 R. H. Faraj et al.

Table 4 Properties of RP aggregates used in the literature

Refs Type of plastic Material replaced Shape of particles Specific Bulk density (kg/ Particle size
gravity m3)

Mohammed et al. PET Fine aggregate Flaky and flat 1.375 417.87 Less than
(2019) 4.75 mm
Aslani and Ma (2018) polystyrene Fine aggregate Granular – – Less than
4.75 mm
Faraj et al. (2019) PP Medium aggregate Granular and 0.95 – 4–8 mm
smooth
Law et al. (2018) polystyrene Fine aggregate Granular and – – Less than
smooth 4.75 mm
Ranjbar and Mousavi EPS Fine and coarse Granular and 0.025 13.6 2.36–9.5
(2015) aggregates smooth
Hilal et al. (2018) Polyethylene Fine aggregate Granular and 0.94 – Less than
Waste smooth 4.75 mm
Chunchu and Putta HIPS Fine aggregate Granular and 1.04 – 1.18–3
(2019) smooth
Sadrmomtazi et al. PET Fine aggregate Flaky and flat 1.2 – Less than
(2016) 4.75 mm
Yang et al. (2015) PP Fine aggregate Granular and 0.95 515 1.5–4 mm
smooth

Fig. 10 The best ANN Model

with three hidden layers and ten
neurons

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Table 5 The tested ANN architectures

No. of hidden No. of neurons in the first No. of neurons in the second No. of neurons in the third R2 MAE RMSE
layers hidden layer hidden layer hidden layer (MPa) (MPa)

1 5 0 0 0.9622 5.2909 6.7918

1 6 0 0 0.9697 4.462 5.9059
1 7 0 0 0.9702 4.385 5.8232
1 8 0 0 0.9689 4.7468 6.171
1 9 0 0 0.9704 4.516 5.9046
1 10 0 0 0.9691 4.5205 6.0367
1 11 0 0 0.9688 4.3943 5.8706
1 12 0 0 0.9701 4.4231 5.8527
2 2 0 2 0.9311 6.7535 8.719
2 2 0 3 0.9312 6.766 8.7545
2 3 0 3 0.959 5.5959 7.1188
2 4 0 4 0.9701 5.1845 6.6005
2 5 0 5 0.970 4.6151 6.0873
2 6 0 6 0.9713 4.7771 6.2045
2 7 0 7 0.9697 4.3264 5.7472
2 8 0 8 0.9713 4.2101 5.5885
2 9 0 9 0.9716 4.5431 5.9605
2 8 0 9 0.9707 4.4307 5.7947
2 9 0 10 0.9709 4.2577 5.6709
2 12 0 12 0.9726 4.201 5.538
2 12 0 13 0.9731 4.0297 5.4648
2 13 0 13 0.9729 4.0236 5.4849
3 5 5 5 0.9694 4.5201 5.8917
3 8 8 8 0.9715 4.2713 5.6633
3 10 10 10 0.9733 3.9828 5.4617
3 11 11 11 0.9717 4.1707 5.6008
3 12 12 12 0.9728 4.0495 5.4865

performance) when it is [ 0.3, (fair performance) when it Results and analysis

is between 0.2 and 0.3, (good performance) when it is
between 0.1 and 0.2, and (great performance) when it is Relationships among forecasted and measured
less than 0.1 (Li et al. 2013; Golafshani et al. 2020). In compressive strength
addition, the OBJ parameter was employed as an integrated
performance parameter in Eq. (12) to measure the effi- The LR model
ciency of the suggested models. Finally, positive and
negative error margin lines were included in the model Figure 11a,b, and c demonstrate the connection between
results to graphically illustrate how each model overesti- estimated and actual CS of SCC mixes, including RP
mates and underestimates the predicted results of CS aggregates for all phases, including training, testing, and
compared to the actual values from the experiments. A validating datasets. Based on the model parameters, the w/
positive value means the overestimated percentage of CS, b ratio and RP aggregate content substantially impact the
while the negative value means the underestimated per- compression strength of SCC incorporated RP aggregates.
centage of CS. By optimizing the sum of error squares and the least square
approach, which was performed in Excel using Solver to
obtain the ideal value (a given value, minimum or maxi-
mum) for the Equation, the present model’s weight of each
parameter on the compression strength of SCC

123
 R. H. Faraj et al.

incorporated RP aggregates was found. The values of other

equation cells in the worksheet were used to set limitations
or restrictions on this object cell (Burhan et al. 2020). The
following is the Equation for the LR model with various
weight parameters (Eq. 13):
rc ¼ 217:74  0:2527RP þ 0:334B þ 86:98w=b þ 0:056t
þ 0:779SP þ 0:0057FA þ 0:083CA
ð13Þ
The w/b ratio has the most impact on lowering the CS
among all factors, as seen in the Equation above. This may
be consistent with the experimental findings reported in the
literature. This model’s R2, RMSE, and MAE assessment
parameters are 0.79, 10.73 MPa, and 8.97 MPa, respec-
tively. Furthermore, as shown in Figs. 20, 21, the current
model’s OBJ and SI values for the training dataset are
11.52 and 0.196, respectively.

NLR model

Figures 12a, b, and c show the forecasted vs. real CS

derived from past investigations of SCC mixes incorpo-
rated RP aggregates for training, testing, and validating
datasets, respectively. Due to this model, the w/b ratio, FA,
and CA concentration are the most significant factors
influencing the CS of SCC mixes. Many experimental
programs from previous research supported this, as shown
in Table 1, in which lowering the w/b ratio and altering the
quantity of aggregates had a substantial impact on the CS
of SCCs containing RP aggregates. The following is the
suggested formula for the NLR model with various vari-
able parameters (Eq. 14):
 0:83 0:16
rc ¼ 6:87107 ðBÞ0:781 w=b ðtÞ ðSPÞ0:021 ðFAÞ0:43 ðCAÞ1:34
 36:43 0:069
þ266627:89ðBÞ1:92 w=b ðt Þ ðSPÞ12:74 ðFAÞ100:02
72:96 0:014
ðCAÞ ðRPÞ
ð14Þ
The R2, RMSE, and MAE assessment parameters for
this model are 0.77, 11.05 MPa, and 9.28 MPa, respec-
tively. Furthermore, the current model’s OBJ and SI values
for the training dataset are 11.44 and 0.2, respectively.

MLR model

Figures 13a, b, and c demonstrate the comparison of pre-

dicted CS against actual CS extracted from previous
studies of SCC mixes made with RP aggregates for all
Fig. 11 Comparison between measured and predicted compressive phases (training, testing, and validating datasets). As
strength of SCC mixtures containing RP aggregates using Linear
Regression model (LR), a training data, b testing data, and
shown in previous studies reported in Table 1, the most
c validating data influential parameter that impacts the CS of SCC mixes,
including RP aggregates, is the w/b ratio, similar to other
models. Equation 15 may be used to forecast the CS of

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Fig. 12 Comparison between measured and predicted compressive

strength of SCC mixtures containing RP aggregates using Nonlinear Fig. 13 Comparison between measured and predicted compressive
Regression model (NLR), a training data, b testing data, and strength of SCC mixtures containing RP aggregates using Multilinear
c validating data Regression model (MLR), a training data, b testing data, and
c validating data

123
 R. H. Faraj et al.

SCC containing RP aggregates for the MLR model with

various variable parameters:
 3:073 0:13
rc ¼ 0:013ðRPÞ0:012 ðBÞ0:42 w=b ðtÞ ðSPÞ0:008 ðFAÞ0:73 ðCAÞ0:34
ð15Þ
The R2, RMSE, and MAE assessment parameters for
this model are 0.78, 8.67 MPa, and 6.83 MPa, respectively.
Furthermore, the current model’s OBJ and SI values for the
training dataset are 8.46 and 0.167, respectively.

ANN model

The authors investigated various hidden layers, neurons,

momentum, learning rate, and iterations to achieve high
ANN efficiency. Finally, they discovered that when the
ANN contains three hidden layers, ten neurons on each side
(as shown in Fig. 10), 0.1 momentum, 0.2 learning rate,
and 2000 iterations, the CS of the SCC containing RP
aggregates is best predicted. Figures 14a, b, and c show the
comparison of estimated and measured CS of SCC made
with RP aggregates for all phases (training, testing, and
validating datasets). The training datasets have a ? 20
percent and -20 percent error line, which is better than the
other proposed models. This model’s R2, RMSE, and MAE
assessment parameters are 0.97, 5.46 MPa, and 2.31 MPa,
respectively. Furthermore, the present model’s OBJ and SI
values for the training dataset are 4.26 and 0.1,
respectively.

Model comparisons

As previously stated, five different statistical methods were

implemented to classify the efficacy of the developed
models, including SI, MAE, R2, RMSE, and OBJ. Com-
pared to the LR, NLR, and MLR models, the ANN has a
higher R2 and lower RMSE and MAE values, as shown in
Figs. 15, 16, and 17 for R2 values RMSE, and MAE,
respectively. Figure 18 also compares model CS estimates
for SCC mixtures, including RP aggregates based on all
data. In addition, Fig. 19 depicts the residual error for all
phases and models. The forecasted and actual values of CS
for the ANN model are closer in both Figs. 18 and 19,
indicating that the ANN model outperforms other models.
Figure 20 shows the OBJ performance for various sug-
gested models. The LR, NLR, MLR, and ANN models
have 11.52, 11.44, 8.46, and 4.26, respectively. The ANN
model had an OBJ value of 63 percent less than that of the
LR and NLR models, and it is also 50 percent smaller than
that of the MLR model. This also shows that the ANN Fig. 14 Comparison between measured and predicted compressive
strength of SCC mixtures containing RP aggregates using ANN
approach is more robust to forecast the CS of SCC mixes, model, a training data, b testing data, and c validating data
including RP aggregates.

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

Fig. 15 R2 values for different proposing models including training,

testing, and validating datasets
Fig. 18 Comparison between model predictions of compressive
strength of SCC mixtures containing RP aggregates using all data

Fig. 19 Variation in predicted values of compressive strength for

SCC mixtures containing RP aggregates based on four different
Fig. 16 RMSE values for different proposing models including approaches in comparison to observed values
training, testing, and validating datasets

Fig. 20 The OBJ values for all developed models

Fig. 17 MAE values for different proposing models including
training, testing, and validating datasets
However, the SI values for the ANN model were between 0
Figure 21 depicts the SI values throughout the training, and 0.1, indicating that the ANN model performed excel-
validating, and testing stages. Figure 21 shows that, except lently. Furthermore, like the other performance factors, the
for the ANN model, the SI values for all stages and models ANN model has lower SI values than other approaches. In
were between 0.1 and 0.23, indicating good performance. the training phase, the ANN model has a 49% lower SI than

123
 R. H. Faraj et al.

When the previously mentioned assessment values are

changed and decreased significantly, it indicates the
effectiveness of the input parameter. Table 6 illustrates the
outcomes of the sensitivity study. The findings show that
similar to the normal SCC, w/b ratio and curing times
(t) are the most significant and dominant parameters in
forecasting the CS of the SCC mixes, including RP
aggregates. In this research, the w/b ratio for the collected
data varied from 0.25 to 0.45; increasing the w/b ratio
dramatically reduces the CS of SCC mixes containing RP
aggregates. Moreover, increasing the curing time caused a
considerable enhancement in the CS of SCC mixtures. All
experimental findings presented in Table 1 support this.

Fig. 21 . Comparing the SI performance parameter of different Environmental and economic assessment
developed models of SCC incorporating RP aggregates
the LR and NLR, and 35% lower SI value in the testing As previously mentioned, the continuous growth of plastic
phase, and a 65.42% smaller SI in the validating. Fur- production caused a significant environmental threat. In
thermore, compared to the MLR model, the ANN had 2020, despite the global pandemic (COVID-19), the world
lower SI values in all stages, such as 40.11% lower in produced 367 million tons of plastics, of which 55 million
training, 17.58% lower in testing, and 55.42% lower in the tons were produced in Europe; of this, 23% ended up in the
validating. This also confirmed that the ANN is more waste stream (Plastics Europe 2021). Figure 22 illustrates
capable and accurate when predicting the CS of SCC the world production of plastic from 2016 to 2021. It is
mixes, including RP aggregates than the LR, NLR, and evident from the figure that the annual production rate of
MLR models. plastic was increased significantly, with an increase of 10%
since 2016. Besides, many countries’ landfill areas are
Sensitivity analysis restricted due to population growth, necessitating the
search for a new way to dispose of plastic waste. Moreover,
A sensitivity analysis was conducted for the models to the use of natural coarse and fine aggregates in the devel-
classify and evaluate the input parameter that mainly opment of new projects depletes natural resources, and
impacts the CS prediction of SCC containing RP aggre- natural aggregate utilization and extraction have a detri-
gates (Mohammed et al. 2020a). To do so, the MLR model mental influence on the ecosystem. As a result, the concrete
was selected because all the input variables should have a industry should establish a new strategy for sustainable
value greater than zero in this model, and it means that the development using this waste material as an aggregate. One
effect of different variables on the CS prediction could be of the potential uses of RP is its incorporation in SCC as a
more evident. All training data were incorporated partial or total substitution of natural aggregates. The uti-
throughout this analysis, and each time a single input lization of RPs from various products in the concrete
variable was extracted. Parameters for interpretation, such industry is an attractive and safe mode of disposal that can
as RMSE, R2, and MAE, were determined separately.

Table 6 Sensitivity analysis

Sr. No Input combination Removed parameter R2 MAE RMSE Ranking
using MLR-based model
1 RP, C, w/b, t, SP, FA, CA – 0.7817 6.8297 8.6746 –
2 B, w/b, t, SP, FA, CA RP 0.7813 6.8156 8.6828 5
3 RP, w/b, t, SP, FA, CA B 0.7815 6.8414 8.6792 6
4 RP, B, t, SP, FA, CA w/b 0.5755 9.8265 12.0981 1
5 RP, B, w/b, SP, FA, CA t 0.7145 7.5326 9.9213 2
6 RP, B, w/b, t, FA, CA SP 0.7817 6.8352 8.6748 7
7 RP, B, w/b, t, SP, CA FA 0.7693 7.0435 8.9193 3
8 RP, B, w/b, t, SP, FA, CA CA 0.7741 6.9891 8.8260 4

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

138 kg/m3. The curing period for data from various

experimental investigations varied from 7 to 400 days.
• The LR, NLR, MLR, and ANN models were estab-
lished in this research to forecast the compression
strength of SCC mixes. According to the various
evaluation criteria, the ANN model outperformed
others with higher R2, smaller OBJ value, smaller
RMSE, smaller SI, and less MAE values for all phases.
• Except for the ANN model, the SI values for all models
and stages were between 0.1 and 0.23, suggesting good
performance. However, the SI values for the ANN
model were between 0 and 0.1, suggesting that the
ANN model performed excellently.
Fig. 22 World plastic production (Plastics Europe 2021) • The OBJ value of the ANN model is 63% smaller than
that of the LR and NLR models, and it is also 50%
resolve the majority of environmental issues caused by
smaller than the MLR model. This also shows that the
plastic wastes.
ANN is more accurate and capable of forecasting the
The RP aggregates give SCC unique characteristics not
CS of sustainable SCC mixes, including RP aggregates.
found in normal SCC, such as lower unit weight and better
• Concerning the curing time, it was discovered via a
ductility. The partial replacement of natural aggregates
sensitivity analysis that it is the most significant
with RP aggregates reduces the density of SCC due to the
parameter for predicting the CS of SCC mixes,
low specific gravity of RP aggregates compared to the
including RP aggregates and different by-product ashes.
natural aggregates; thus, the possibility of producing
Increasing the curing period of SCC mixes with or
lightweight SCC is higher. The benefits of lightweight SCC
without RP aggregates resulted in a substantial
incorporating RP aggregates over normal SCC include the
improvement in the CS.
reduced size of the structural components, reduction in
• The overall findings and analysis showed that specific
dead load of the structure, improved dynamic load
quantities of RP aggregates and industrial by-product
response, more energy absorption and toughness, less
ashes might be utilized effectively in green SCC
reinforcing steel, greater design flexibility, and an overall
manufacture to affect the strength and other aspects of
decrease in the construction costs (Basha et al. 2020).
SCC favorably. On the other hand, the optimal RP
Another significant change when RPs are included in SCC
content is highly dependent on the shape, type, and
is improved thermal insulation, one of the essential criteria
surface texture of RP aggregates.
for energy saving. A previous study reported that a retail
building constructed with concrete that incorporated a high
amount of RPs was shown to use 40% less energy for
Funding This research received no funding.
heating and cooling than a similar building made with
normal concrete (Elzafraney et al. 2005).
Declarations

Conclusion Conflict of interest The author has no conflicts of interest to declare

that are relevant to the content of this article.

Accurate and dependable models to forecast the CS may Data availability All data are provided in the manuscript.
result in considerable cost and time savings for the con-
struction industry. It is possible to draw the following Ethical approval The contents of this manuscript are not now under
consideration for publication elsewhere; The contents of this manu-
findings due to the analysis and modeling using the data
script have not been copyrighted or published previously. The con-
gathered from prior research to forecast the CS of SCC tents of this manuscript will not be copyrighted, submitted, or
mixes, including RP aggregates and different by-product published elsewhere, while acceptance by the Journal is under
ashes at 400 various mixed proportions: consideration.

• In the production of SCC mixes, the average proportion Consent to participate Not applicable.
of RP utilized was 17.035 kg/m3. Furthermore, the
Consent for publication Not applicable.
proportion of RP aggregates replacement with natural
aggregates (Fine or coarse aggregates) varied from 0 to

123
 R. H. Faraj et al.

References Ghafor K, Qadir S, Mahmood W, Mohammed A (2020) Statistical

variations and new correlation models to predict the mechanical
behaviour of the cement mortar modified with silica fume.
Ahmed HU, Mohammed AS, Mohammed AA, Faraj RH (2021)
Geomech Geoeng. https://doi.org/10.1080/17486025.2020.
Systematic multiscale models to predict the compressive
1714083
strength of fly ash-based geopolymer concrete at various mixture
Golafshani EM, Behnood A, Arashpour M (2020) Predicting the
proportions and curing regimes. PLoS ONE 16(6):e0253006.
compressive strength of normal and high-performance concretes
https://doi.org/10.1371/journal.pone.0253006
using ANN and ANFIS hybridized with grey wolf optimizer.
Aslani F, Ma G (2018) Normal and high-strength lightweight self-
Constr Build Mater 232:117266. https://doi.org/10.1016/j.con
compacting concrete incorporating perlite, scoria, and polystyr-
buildmat.2019.117266
ene aggregates at elevated temperatures. J Mater Civ Eng
Güneyisi E, Gesoğlu M, Özbay E (2010) Strength and drying
30(12):04018328. https://doi.org/10.1061/(ASCE)MT.1943-
shrinkage properties of self-compacting concretes incorporating
5533.0002538
multi-system blended mineral admixtures. Constr Build Mater
Aslani F, Ma G, Wan DLY, Muselin G (2018) Development of high-
24(10):1878–1887. https://doi.org/10.1016/j.conbuildmat.2010.
performance self-compacting concrete using waste recycled
04.015
concrete aggregates and rubber granules. J Clean Prod
Guo X, Shi H, Dick WA (2010) Compressive strength and
182:553–566. https://doi.org/10.1016/j.jclepro.2018.02.074
microstructural characteristics of class C fly ash geopolymer.
Basha SI, Ali MR, Al-Dulaijan SU, Maslehuddin M (2020) Mechan-
Cement Concr Compos 32(2):142–147. https://doi.org/10.1016/j.
ical and thermal properties of lightweight recycled plastic
cemconcomp.2009.11.003
aggregate concrete. J Build Eng 32:101710. https://doi.org/10.
Hilal NN, Kareem Q, Nawar MT (2018) Influence of polyethylene
1016/j.jobe.2020.101710
waste on some fresh and mechanical properties of self-
Burhan L, Ghafor K, Mohammed A (2020) Enhancing the fresh and
compacting concrete. J Eng Appl Sci 13(14):10901–10911
hardened properties of the early age concrete modified with
Law Yim Wan DS, Aslani F, Ma G (2018) Lightweight self-
powder polymers and characterized using different models. Adv
compacting concrete incorporating perlite, scoria, and polystyr-
Civil Eng Mater 9(1):227–249. https://doi.org/10.1520/
ene aggregates. J Mater Civil Eng 30(8):04018178. https://doi.
ACEM20190087
org/10.1061/(ASCE)MT.1943-5533.0002350
Chunchu BRK, Putta J (2019) Rheological and strength behavior of
Li MF, Tang XP, Wu W, Liu HB (2013) General models for
binary blended SCC replacing partial fine aggregate with plastic
estimating daily global solar radiation for different solar
E-waste as high impact polystyrene. Buildings 9(2):50. https://
radiation zones in mainland China. Energy Convers Manage
doi.org/10.3390/buildings9020050
70:139–148. https://doi.org/10.1016/j.enconman.2013.03.004
Deilami S, Aslani F, Elchalakani M (2019) An experimental study on
Mohamed O (2018) Durability and compressive strength of high
the durability and strength of SCC incorporating FA, GGBS and
cement replacement ratio self-consolidating concrete. Buildings
MS. Proc Inst Civil Eng-Struct Buildings 172(5):327–339.
8(11):153. https://doi.org/10.3390/buildings8110153
https://doi.org/10.1680/jstbu.17.00129
Mohammed MK, Al-Hadithi AI, Mohammed MH (2019) Production
Demircan E, Harendra S, Vipulanandan C (2011) Artificial neural
and optimization of eco-efficient self-compacting concrete SCC
network and nonlinear models for gelling time and maximum
with limestone and PET. Constr Build Mater 197:734–746.
curing temperature rise in polymer grouts. J Mater Civ Eng
https://doi.org/10.1016/j.conbuildmat.2018.11.189
23(4):372–377. https://doi.org/10.1061/(ASCE)MT.1943-5533.
Mohammed A, Rafiq S, Mahmood W, Al-Darkazalir H, Noaman R,
0000172
Qadir W, Ghafor K (2020a) Artificial Neural Network and NLR
Deshpande N, Londhe S, Kulkarni S (2014) Modeling compressive
techniques to predict the rheological properties and compression
strength of recycled aggregate concrete by artificial neural
strength of cement past modified with nanoclay. Ain Shams Eng
network, model tree and nonlinear regression. Int J Sustain Built
J. https://doi.org/10.1016/j.asej.2020.07.033
Environ 3(2):187–198. https://doi.org/10.1016/j.ijsbe.2014.12.
Mohammed A, Rafiq S, Sihag P, Kurda R, Mahmood W (2020b) Soft
002
computing techniques: systematic multiscale models to predict
Duan ZH, Kou SC, Poon CS (2013) Prediction of compressive
the compressive strength of HVFA concrete based on mix
strength of recycled aggregate concrete using artificial neural
proportions and curing times. J Build Eng. https://doi.org/10.
networks. Constr Build Mater 40:1200–1206. https://doi.org/10.
1016/j.jobe.2020.101851
1016/j.conbuildmat.2012.04.063
Mohammed AM, Asaad DS, Al-Hadithi AI (2021) Experimental and
Elzafraney M, Soroushian P, Deru M (2005) Development of energy-
statistical evaluation of rheological properties of self-compacting
efficient concrete buildings using recycled plastic aggregates.
concrete containing fly ash and ground granulated blast furnace
J Archit Eng 11(4):122–130. https://doi.org/10.1061/
slag. J King Saud Univ Eng Sci. https://doi.org/10.1016/j.jksues.
(ASCE)1076-0431(2005)11:4(122)
2020.12.005
Faraj RH, Sherwani AFH, Daraei A (2019) Mechanical, fracture and
Neville AM (1995) Properties of concrete’’, five, and final Edition,
durability properties of self-compacting high strength concrete
Wiley, New York and Longman, London, pp 844
containing recycled polypropylene plastic particles. J Build Eng
PlasticsEurope EPRO (2021). Plastics—the facts 2021. An analysis of
25:100808. https://doi.org/10.1016/j.jobe.2019.100808
european plastics production, demand and waste data. Plast Eur
Faraj RH, Mohammed AA, Mohammed A, Omer KM, Ahmed HU
https://plasticseurope.org/knowledge-hub/plastics-the-facts-
(2021) Systematic multiscale models to predict the compressive
2021/
strength of self-compacting concretes modified with nanosilica at
Qadir W, Ghafor K, Mohammed A (2019) Characterizing and
different curing ages. Eng Comput. https://doi.org/10.1007/
modeling the mechanical properties of the cement mortar
s00366-021-01385-9
modified with fly ash for various water-to-cement ratios and
Gao W, Karbasi M, Derakhsh AM, Jalili A (2019) Development of a
curing times. Adv Civil Eng. https://doi.org/10.1155/2019/
novel soft-computing framework for the simulation aims: a case
7013908
study. Eng Comput 35(1):315–322. https://doi.org/10.1007/
Ranjbar MM, Mousavi SY (2015) Strength and durability assessment
s00366-018-0601-y
of self-compacted lightweight concrete containing expanded

123
Soft computing techniques to predict the compressive strength of green self-compacting concrete…

polystyrene. Mater Struct 48(4):1001–1011. https://doi.org/10. cement. Eng Comput. https://doi.org/10.1007/s00366-020-

1617/s11527-013-0210-6 01081-0
Sadrmomtazi A, Dolati-Milehsara S, Lotfi-Omran O, Sadeghi-Nik A Sihag P, Jain P, Kumar M (2018) Modelling of impact of water
(2016) The combined effects of waste polyethylene terephthalate quality on recharging rate of storm water filter system using
(PET) particles and pozzolanic materials on the properties of various kernel function based regression. Model Earth Syst
self-compacting concrete. J Clean Prod 112:2363–2373. https:// Environ 4(1):61–68. https://doi.org/10.1007/s40808-017-0410-0
doi.org/10.1016/j.jclepro.2015.09.107 Spiesz P, Rouvas S, Brouwers HJH (2016) Utilization of waste glass
Saikia N, Brito JD (2013) Waste polyethylene terephthalate as an in translucent and photocatalytic concrete. Constr Build Mater
aggregate in concrete. Mater Res 16(2):341–350. https://doi.org/ 128:436–448. https://doi.org/10.1016/j.conbuildmat.2016.10.063
10.1590/S1516-14392013005000017 Uysal M, Sumer M (2011) Performance of self-compacting concrete
Salih A, Rafiq S, Sihag P, Ghafor K, Mahmood W, Sarwar W (2020) containing different mineral admixtures. Constr Build Mater
Systematic multiscale models to predict the effect of high- 25(11):4112–4120. https://doi.org/10.1016/j.conbuildmat.2011.
volume fly ash on the maximum compression stress of cement- 04.032
based mortar at various water/cement ratios and curing times. Yang S, Yue X, Liu X, Tong Y (2015) Properties of self-compacting
Measurement 171:108819. https://doi.org/10.1016/j.measure lightweight concrete containing recycled plastic particles. Constr
ment.2020.108819 Build Mater 84:444–453. https://doi.org/10.1016/j.conbuildmat.
Sarwar W, Ghafor K, Mohammed A (2019) Modeling the rheological 2015.03.038
properties with shear stress limit and compressive strength of Yu QL (2019). Application of nanomaterials in alkali-activated
ordinary Portland cement modified with polymers. J Build Pathol materials. In Nanotechnology in Eco-efficient Construction (pp
Rehabil 4(1):25. https://doi.org/10.1007/s41024-019-0064-6 97–121). Woodhead Publishing. https://doi.org/10.1016/B978-0-
Shahbazpanahi S, Manie S, Faraj RH, Seraji M (2021) Feasibility 08-102641-0.00005-0
study on the use of tagouk ash as pozzolanic material in Zain MFM, Abd SM (2009) Multiple regression model for compres-
concrete. Clean Technol Environ Policy 23(4):1283–1294. sive strength prediction of high: performance concrete. J Appl
https://doi.org/10.1007/s10098-020-02021-8 Sci 9:155–160. https://doi.org/10.3923/jas.2009.155.160
Shariati M, Mafipour MS, Ghahremani B, Azarhomayun F, Ahmadi
M, Trung NT, Shariati A (2020) A novel hybrid extreme learning Publisher’s Note Springer Nature remains neutral with regard to
machine–grey wolf optimizer (ELM-GWO) model to predict jurisdictional claims in published maps and institutional affiliations.
compressive strength of concrete with partial replacements for

123