Innovative Infrastructure Solutions (2024) 9:412

[Link]

TECHNICAL PAPER

Artificial neural network and soft computing models to predict

the compressive strength in self‑compacting green concrete
Bassam Abd Ali1 · Ahmed Salih Mohammed2 · Moslih Amer Salih3 · Sozin Hama Salih1 · Militan Omer1 ·
Rawaz Kurda4,5,6 · Wael Emad7 · Panagiotis G. Asteris8 · Parveen Sihag9

Received: 7 June 2024 / Accepted: 17 September 2024

© Springer Nature Switzerland AG 2024

Abstract
This study explores self-compacting concrete (SCC) enhancement by incorporating silica fume, fly ash, or both across vari-
ous mix compositions. The research evaluates six predictive models—linear regression, nonlinear regression, pure quadratic,
interaction, full quadratic, and artificial neural network (ANN)—to predict the compressive strength of SCC. A dataset
of 330 experimental studies covering a wide range of parameters, such as water/cement ratio, cement content, aggregate
content, superplasticizer content, silica fume content, fly ash content, and curing time, is used. The compressive strength of
the datasets ranges from 4.9 to 87 MPa, while slump flow diameter ranges from 450 to 790 mm. The models are assessed
using objective function, root mean square error (RMSE), scatter index (SI), and mean absolute error (MAE). The ANN is
the most accurate among the models, achieving an ­R2 of 0.94, RMSE of 3.56 MPa, MAE of 2.67 MPa, and SI of 0.09. The
models effectively predict compressive strength across various concrete compositions, although they do not predict slump
flow diameter, as SCC specifications require it to be within 550 to 850 mm.

Keywords Self-compacting concrete · Compressive strength · Statistical analysis · Soft computing models · Mix
proportions · Fly ash · Silica fume

Introduction workability and the ability to fill intricate and congested

reinforcement without segregation or excessive bleeding
Self-compacting concrete (SCC) is a highly fluid type that [2]. SCC typically comprises the same basic ingredients as
can flow and consolidate under its weight without mechani- conventional concrete, including cement, aggregates (such
cal vibration [1]. It is specifically designed to have excellent as fine and coarse aggregates), and water. However, it also

3
* Ahmed Salih Mohammed Technical Institute of Babylon, Al-Furat Al-Awsat Technical
[Link]@[Link] University (ATU), Najaf, Iraq
4
Moslih Amer Salih Department of Highway and Bridge Engineering, Technical
[Link]@[Link] Engineering College, Erbil Polytechnic University,
Erbil 44001, Iraq
Rawaz Kurda
5
[Link]@[Link] Department of Civil Engineering, College of Engineering,
Nawroz University, Duhok 42001, Iraq
Wael Emad
6
[Link]@[Link] CERIS, Civil Engineering, Architecture and Georresources
Department, Instituto Superior Técnico, Universidade de
Panagiotis G. Asteris
Lisboa, Av. Rovisco Pais, 1049‑001 Lisbon, Portugal
panagiotisasteris@[Link]
7
Scientific Research Center of Soran University,
Parveen Sihag
Kurdistan Region, Iraq
parveen12sihag@[Link]
8
Computational Mechanics Laboratory, School of Pedagogical
1
Civil Engineering Department, College of Engineering, and Technological Education, 14121 Heraklion, Athens,
American University of Iraq, Sulimania, Kurdistan, Iraq Greece
2 9
College of Engineering, University of Sulaimani, Sulaimani, Civil Engineering Department, Chandigarh University,
Iraq Mohali, Punjab, India

Vol.:(0123456789)
412 Page 2 of 31 Innovative Infrastructure Solutions (2024) 9:412

incorporates additional components to achieve its unique at various curing times. Construction work is often delayed
properties. These additional components may include high- as further progress cannot proceed until the compressive
range water reducers (superplasticizers) to improve flowa- strength test results are available, typically after 28 days.
bility, viscosity-modifying agents to enhance stability and This testing process incurs both time and cost burdens [16].
prevent segregation, and fine particles (such as fly ash or Estimating the compressive strength of SCC has always
silica fume) to improve cohesion and fill voids. been a major challenge in concrete technology due to the
The development of self-compacting concrete (SCC) significant impact of different mix proportions and materi-
originated in 1986 with the concept introduced by Hajime als on its properties [16]. This challenge is particularly evi-
Okamura. However, in 1988, Ozawa proposed the proto- dent when SCC incorporates pozzolanic materials such as
type of SCC in Japan [1, 3]. Using self-compacting concrete silica fume or fly ash as partial replacements for cement. The
(SCC) has been estimated to accelerate construction speed inclusion of these materials further complicates the task of
by up to 40% compared to conventional concrete [4, 5]. predicting and evaluating the compressive strength of SCC.
Concrete, the most extensively used material in large-scale Therefore, it is necessary to thoroughly analyze the impact
infrastructure projects, is irreplaceable in construction. Con- of any changes made to the materials that constituent con-
sidering the immense quantities of concrete utilized, even crete, such as cement, gravel, sand, superplasticizers (SP),
minor improvements in concrete can significantly impact fine aggregate (FA), coarse aggregate (CA), curing time (t),
the overall industry [6]. To mitigate the negative impacts silica fume (SF%), fly ash (FA%), and water-to-cement ratio
of cement, pozzolanic materials such as silica fume or fly (w/c) on the overall characteristics.
ash are incorporated into self-compacting concrete (SCC). Fly ash, a by-product of coal combustion in energy pro-
The addition of silica fume enhances the workability of duction is an industrial residue with environmental implica-
SCC while simultaneously improving its rheological and tions. However, it is also considered a valuable resource and
mechanical properties. Conversely, including fly ash in SCC the fifth-largest raw material in the world. With its pozzo-
primarily enhances its rheological properties while slightly lanic properties and ability to bind with lime, fly ash finds
reducing its mechanical characteristics [7, 8]. This is also utility in manufacturing cement, building materials, con-
confirmed by the study done by Satish et al., as it concluded crete, and concrete-based products such as SCC. It serves as
that the inclusion of silica fume resulted in higher strength a raw material in concrete production and contributes to the
when compared to mixes containing fly ash [9]. This can sustainable utilization of this industrial by-product [17, 18].
be explained by the increased silica fume content and the However, silica fume is a by-product of the smelting
decreased L-box ratio, indicating improved flowability and process in the silicon and ferrosilicon industries. This sub-
filling ability. Conversely, adding fly ash enhanced the work- stance possesses exceptional pozzolanic characteristics and
ability of SCC, leading to increased ease of placement and possesses useful benefits. The incorporation of silica fume
compaction [8]. in concrete has improved several features. Significantly, it
Similarly, Yazicioglu et al. The study found that self-com- enhances the capacity to withstand compression, adhesion
pacting concrete (SCC) containing silica fume consistently strength, and wear resistance while decreasing permeability
showed the highest values, followed by SCC with fly ash, [19, 20].
and finally, conventional concrete (PC concrete), regardless Aggregates play a significant role in shaping concrete's
of the curing durations and circumstances [10]. rheological and mechanical properties. The size, shape, and
Producing self-compacting concrete (SCC) mixtures proportion of aggregates and the type of aggregate directly
necessitates incorporating higher volumes of paste. How- impact the strength, durability, and workability of concrete.
ever, excessive cement usage can significantly escalate Rougher aggregates result in higher strength due to their
costs and adversely affect other fundamental characteristics enhanced resistance to applied loads. Conversely, smoother
[11–13]. As a result, pozzolanic materials like fly ash or sil- aggregates improve workability and decrease compressive
ica fume are highly recommended for self-compacting con- strength (CS). The size of the aggregates also influences
crete (SCC) to improve its properties and reduce its costs. CS, with smaller fine aggregates (FA) acting as fillers for
The compressive strength (CS) of self-compacting con- voids, while coarse aggregates (CA) contribute to strength
crete (SCC) holds great significance in the design of engi- through their mass.
neering structures. It serves as a critical parameter because By filling the voids and increasing cohesion, FA enhances
SCC directly or indirectly influences other mechanical and concrete's overall strength, enabling it to withstand applied
durability properties influences other mechanical and dura- loads better [21]. The aggregates used in self-compacting
bility properties, making it possible to derive those prop- concrete (SCC) significantly influence its rheological and
erties from the compressive strength [14, 15]. In practical hardened properties. Since aggregates typically comprise
applications, the determination of CS in SCC involves the around 60–70% of the volume of SCC, they play a crucial
preparation and testing of cylindrical or cubical specimens role in determining its overall performance. The coarse
Innovative Infrastructure Solutions (2024) 9:412 Page 3 of 31 412

aggregate content in SCC is usually lower than traditional it does not require any special considerations during the cur-
concrete mixes [11, 13, 22]. ing process. The curing procedure for SCC follows a similar
The water-to-cement ratio (w/c) is another critical factor approach to that of regular concrete. Previous studies have
when analyzing concrete. Water and cement are fundamental identified that the age of concrete samples at the time of test-
components in concrete production, and their ratios signifi- ing is a critical factor that affects the compressive strength
cantly impact the resulting concrete's compressive strength (CS) of SCC [34–36].
(CS). Increasing the cement content generally leads to higher This study investigates sustainable alternatives in the con-
CS, while increasing the water content has the opposite struction sector by assessing the viability of using recycled
effect. A lower w/c ratio corresponds to reduced porosity and concrete aggregates (RCA), coffee husk ash, polypropylene
increased CS, while a higher w/c ratio results in increased fibres, and granite powder (GP) as replacements for conven-
porosity and decreased CS in the concrete [23, 24]. Accord- tional concrete constituents. The study centers on three pri-
ing to Naderpour and Abbasi, reducing the water-to-binder mary domains: An assessment is conducted to determine the
(w/b) ratio from 0.45 to 0.35 led to a significant improve- viability of utilizing Recycled Concrete Aggregates (RCA)
ment in the compressive strength (CS) of self-compacting as a partial substitute for coarse aggregates in concrete. Tests
concrete (SCC) at various curing times (3, 7, 28, 56 days) were conducted on samples containing 10–40% recycled
[25]. Furthermore, Ahmadi et al. highlighted that the w/b concrete aggregate (RCA), and the results showed that the
ratio has a more significant influence on the CS of conven- ideal percentage of RCA is 20%. The concrete parameters,
tional concrete when compared to the CS of SCC [26]. such as slump, unit weight, and air content, were within
Superplasticizers (SP) in concrete play a crucial role in acceptable limits at this percentage. The mechanical char-
achieving consistent strength, improved workability, and acteristics of RCA were consistent, exhibiting a compres-
reduced water-to-cement ratio (w/c) [27, 28]. Incorporating sive strength above 21 MPa. This indicates that RCA is a
superplasticizers (SP) enhances the cohesiveness of con- feasible and environmentally benign choice for construct-
crete, reduces segregation, lowers permeability, and miti- ing sidewalks, slabs, and pavements. The efficacy of using
gates the formation of cracks and fractures. These effects Coffee Husk Ash and Polypropylene Fibres in typical con-
contribute to a stronger bond within the concrete matrix, crete mixes was evaluated. The findings indicated minor
increasing early and ultimate strength [29]. Advancements in decreases in compressive, tensile, and flexural strength
superplasticizer technology have significantly contributed to but enhancements in the modulus of elasticity, indicating
the advancement of self-compacting concrete (SCC). Super- a trade-off between sustainability and mechanical capabil-
plasticizers, as chemical admixtures, are vital in improving ity. Granite Powder (GP): GP was a partial substitute for
numerous properties of self-compacting concrete (SCC). sand in concrete mixtures. The ideal combination was dis-
These advancements in superplasticizers have enabled the covered using 20% GP, resulting in a 13% enhancement in
development of SCC by enhancing its workability, reducing compressive strength while keeping workability within the
water permeability, increasing resistance to carbonation and desired range. This demonstrates the potential of GP as an
chloride attack, and improving concrete's initial and final environmentally friendly ingredient in concrete manufactur-
strength. Without the advancements in superplasticizers, the ing [37–42]. This study highlights the significant benefits
creation of SCC as we know it today would not have been of integrating industrial wastes and metakaolin (MK) into
possible [30, 31]. In a study conducted by Sor N. A., the ultra-high-performance concrete (UHPC). The study shows
effect of various superplasticizer (SP) dosages on the com- that replacing typical Portland cement with other industrial
pressive strength (CS) and fresh behavior of self-compacting wastes, including as silica fume (SF), granulated blast-fur-
concrete (SCC) was investigated. The experiment revealed nace slag (GGBS), and fly ash (FA), combined with MK,
that by increasing the dose of SP from 5.5 to 8.25 kg/m3, leads to substantial cost and carbon emission savings in the
there was a notable enhancement in the compressive strength manufacturing of UHPC. The addition of 15% MK, 50%
of SCC. Specifically, the compressive strength increased GGBS, and 25% MK significantly increases compressive,
from 35 MPa to 45.5 MPa after 28 days [32]. flexural, and splitting tensile strengths and improved resist-
Understanding the curing time is another significant ance to chloride permeability.
parameter to consider, as it determines the period in which These ternary combinations were used to optimize the
the concrete reaches its maximum strength through the mechanical performance of UHPC, resulting in improved
chemical reactions with water or air moisture. The chosen properties compared to previous formulations. A high-den-
method of concrete curing, along with the interplay of vari- sity microstructure and a narrow interfacial transition zone
ous parameters, influences the strength of the concrete at enhanced the concrete's durability and strength, as shown
different ages, ranging from its early to its later stages [33]. by Scanning Electron Microscopy (SEM) and energy-dis-
One of the significant benefits of self-compacting concrete persive X-ray Spectroscopy (EDX) investigations. Life cycle
(SCC) compared to other specialized concrete types is that studies provide additional evidence of these formulations'
412 Page 4 of 31 Innovative Infrastructure Solutions (2024) 9:412

environmental advantages, a decrease in energy usage, lower (SP), fly ash content percentages (FA%), silica fume content
overall costs throughout the life cycle, and a smaller carbon percentages (SF%), and curing time (t). Through a thorough
footprint. By using industrial wastes as construction materi- analysis of data from multiple research papers, statistical
als, this method tackles important environmental concerns modeling techniques will be applied to validate the findings:
associated with industrial by-products and helps create a
more sustainable and economically feasible Ultra-High • To examine how different SCC compositions, including
Performance Concrete (UHPC). The results indicate that modifications with silica fume, fly ash, or both, influ-
using these materials can result in more environmentally ence self-compacting concrete's compressive strength
friendly and effective construction methods in the future (CS) and slump flow diameter.
[43, 44]. Building upon this study, the primary objective • To identify the model that best predicts the compressive
of the current study is to investigate the influence of fly ash strength (CS) and slump flow diameter of SCC modified
and silica fume on the mechanical properties of SCC. The with silica fume, fly ash, or both.
study will involve various combinations of cement, gravel,
sand, and superplasticizers (SP), taking into account factors
such as fine aggregate (FA), coarse aggregate (CA), curing
Methodology
time (t), silica fume content (SF%), fly ash content (FA%),
and water-to-cement ratio (w/c). To analyze the compres-
A total of 330 data points from previous studies were col-
sive strength and slump flow diameter of SCC, multiple
lected and analyzed using statistical methods. These data
modeling techniques will be employed, including nonlinear
points were divided into three groups: a larger group con-
regression (NLR), linear regression (LR), artificial neural
sisting of 231 datasets for model creation and two smaller
network (ANN), full quadratic (FQ), interaction (IA), and
groups with 50 and 49 datasets used for testing and validat-
pure quadratic (PQ) analysis. The goal is to determine the
ing the models. Table 1 provides a summary of the SCC
most suitable and reliable model for predicting the compres-
mixtures database, including the composition of the SCC
sive strength of SCC modified with fly ash and silica fume.
modified with silica fume, fly ash, or both, as well as the
It is worth noting that the available data for this study
parameters of w/c ratio, cement, fine aggregate (FA), coarse
exhibit certain limitations. There are limited ranges or clus-
aggregate (CA), superplasticizers (SP), and curing time.
tering of data points, with approximately 20.3% of the data
The main objective of this study is to predict the com-
lacking both fly ash and silica fume. The data distribution is
pressive strength (CS) of the SCC while also analyzing the
uneven across different compressive strength ranges. Moreo-
slump flow diameter of all SCC mixtures through statistical
ver, creating models for the slump flow diameter is not fea-
methods. This study will employ and evaluate six differ-
sible due to the required range of 550 to 850 mm of SCC.
ent models to select the most reliable and accurate model
Consequently, this study aims to explore the unanalyzed or
to achieve this. The modeling analysis will involve various
limited scope of studies regarding the compressive strength
mathematical operations, as depicted in Fig. 1, to establish
of SCC while accounting for the limitations.
the best-fitting relationship between the collected parameters
and the CS of the SCC modified with silica fume, fly ash,
or both.
Research objectives

This study aims to predict the strength of SCC modified with

silica fume, fly ash, or both by utilizing data collected from Statistical evaluation
various studies. The dataset consists of 330 samples divided
into three groups: training, testing, and validating. Six mod- This section explores the relationship between the input
els will be employed to analyze these datasets, including variables, the independent variables related to the SCC mix
training to determine anticipated variables and constants, design incorporating silica fume, fly ash, or both, and the
testing to assess model performance, and validating to evalu- output variable, the dependent variable. The independent
ate the workability of fixed variables. variables considered in this study include (1) water-to-
The main objective of this study is to investigate the cement ratio (w/c), (2) cement content, (3) fine aggregate
effectiveness of different models in predicting the compres- content (FA), (4) coarse aggregate content (CA), (5) super-
sive strength (CS) and slump flow diameter of SCC modified plasticizers (SP), (6) fly ash percentage (FA%), (7) silica
with silica fume, fly ash, or both. These predictions will be fume percentage (SF%), and (8) curing time (t). The depend-
based on various mix design components, such as water- ent variable is the compressive strength (CS), and the goal
to-cement ratio (w/c), cement content (c), coarse aggregate is to establish a relationship between the independent vari-
(CA) content, fine aggregate (FA) content, superplasticizers ables and the CS to predict its value. Multiple mathematical
Table 1  Summary of the datasets used in the modeling process
References No. of data w/c Cement (kg/m3) Fine Aggregate (kg/ Coarse SP (%) Fly Ash (%) Silica fume (%) Curing time (Day) Compres- Slump Flow (mm)
m3) Aggregate sive Strength
(kg/m3) (MPa)

[43] 35 0.38–0.45 382–450 873–939 873–939 0.76–1.8 0 0–15 7–90 26–64 660–770
[44] 5 0.44 400–455 800–883 771–783 1.26–1.76 0 0–12 7–90 34.5–58.8 650
[45] 2 0.35 250–300 900 600 1–1.24 0 40–50 28 35.1–42.2 –
Innovative Infrastructure Solutions

[46] 39 0.42 225–495 909–1109 612 2–2 0–50 0–20 7–28 12–35 –
[47] 1 0.44 350 960 920 0.72 0 9 7 21.1 750
[48] 4 0.38 378–444 986–1010 777 1.2–1.5 0 0–15 90–180 57–75.8 650
[49] 3 0.37 320–400 1062–1085 761–778 0.86–1.44 0 0–20 7–28 38–40 670–690
[50] 4 0.4 400–436 921–927 891–898 1–1.3 0 3.1–11 28 37.9–45.9 640–660
(2024) 9:412

[51] 2 0.36–0.38 405–427 932 793 1.78 0 5–10 28 36.5–44.1 –

[52] 9 0.42 400 785–827 798–799 1.3–1.74 0 13 7–28 22–43.5 –
[53] 5 0.45 360–400 784–793 1000 0.33–0.4 0 0–10 56–180 55–77 –
[54] 14 0.33 330–550 970 722 0.22–1.5 0–35 0–5 7–90 37.5–60 690–745
[55] 4 0.32 353.6–416 780 954 0 20 0–12 28 42.71–47.1 680–708
[56] 19 0.35–0.50 161–336 739–866 843–1105 0–3.8 0–60 0 1–28 4.9–34.6 450–650
[57] 13 0.42–0.62 325–425 846 300–750 0.3–0.53 0 0 28 20.5–32.4 590–740
[58] 6 0.4 385–495 676–823.7 630.9–640.7 0 0 0 28 30.16–50.2 660–690
[59] 12 0.38–0.44 450 700–730 700–730 0.68 0 0–7.2 7–93 29–61 650–750
[60] 10 0.36 635 847 565 0 0 10 8–90 46–81 –
[61] 12 0.3–0.33 374–395 750–760 450 0.58–1.09 30 0–4 3–7 23.2–42 645–690
[62] 5 0.38 310–620 780 720 0 0–50 0 28 38.12–57.68 –
[63] 16 0.38 250–500 1017–1072 565–595 0 0–50 0–25 28 35–48 640–650
[64] 9 0.36–0.72 267–482 680–704 775 0.48–0.56 4.9–19.5 10–40 7–91 11.2–57.6 585–685
[65] 2 0.4 280 974.9 974.9 1.1 30 0 7–90 36.1–56 –
[66] 5 0.3 165–495 783–836 869–907 1.2–1.6 10–70 0 3–7 18.14–44.42 620–705
[67] 23 0.4 350 846 822 1.5–3 0 30 7–28 23–37 490–790
[68] 9 0.4 152–380 747–1051 934–1171 0 0–20 0 28 24.6–43.4 545–620
[69] 9 0.4 405–450 671.3–941.6 810.1–817.7 1.25 0 0–10 28 30.88–46.48 644–780
[70] 5 0.35 480 925 760 1.3 0 0 1–90 37.5–72 –
[71] 26 0.36–0.4 300–500 893–990 735 1.35–1.6 0–40 0–20 3–130 20–74 701–709
[72] 9 0.37–0.51 350–500 790–1002 732–927 0.6–1.18 0 2.5–9 3–7 24–36 –
[73] 3 0.32 400–450 935–959 656 1.9–2.4 0 10–20 3–180 37–87 –
Page 5 of 31

[74] 5 0.4 450–600 810 660 1.73–2.3 0 0–25 7–28 31–45.1 670–720
[75] 5 0.22–0.26 489 833 700 0.51–0.66 0 15 7–28 43.8–60 758–770
412
412 Page 6 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 1  Flowchart diagram process followed in this study

operations, such as linear, constant, production, quadratic, mean, while variance indicates the spread of the dataset.
and exponential, were employed to develop various models. Additionally, the skewness and Kurtosis of each variable
For each SCC composition, an analysis was conducted were determined. Skewness can be positive or negative
to assess the standard deviation (SD) and variance, which and shows the dataset's asymmetry. However, Kurtosis
provide insights into the dataset's variability. The SD determines whether the dataset has heavy or light tails
measures the extent of variation in the dataset and its and can be positive or negative.
Innovative Infrastructure Solutions (2024) 9:412 Page 7 of 31 412

Water‑to‑cement ratio Cement content

Figure 2a depicts the water-to-cement ratio (w/c) in the dif- The cement content (c) in the different SCC mix designs
ferent SCC mix designs, which ranged from 0.22 to 0.72. ranged from 152 to 635 kg/m3, as depicted in Fig. 2b. The
The analysis also revealed additional statistical measures, analysis also revealed additional statistical measures such as
such as the median value of 0.4, standard deviation (SD) of the median value 385, standard deviation (SD) of 92.7 kg/
0.06, and variance of 0.004. Figure 2a illustrates the rela- m3, and variance of 8593.38. Figure 2b illustrates the rela-
tionship between the compressive strength (CS) and the w/c, tionship between the compressive strength (CS) and the
along with the Kurtosis and Skewness values of 8.122 and cement content, along with the Kurtosis and Skewness val-
1.721, respectively. ues of 0.957 and 0.029, respectively.

Fig. 2  Marginal plots for the

compressive strength (CS) of
concrete modified with fly ash
or silica fume versus, a water/
cement ratio, b cement content,
c fine aggregate content, d
coarse aggregate content, e
superplasticizers, f fly ash, g
Silica fume, and h curing time
 412 Page 8 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 2  (continued)

Fine aggregate content Coarse aggregate content

The fine aggregate content (FA) in the different SCC mix The coarse aggregate content (CA) in the different SCC
designs ranged from 671.33 to 1109 kg/m3, as depicted mix designs ranged from 300 to 1171 kg/m3, as depicted
in Fig. 2c. The analysis also revealed additional statistical in Fig. 2d. The analysis also revealed additional statistical
measures such as the median value of 882, standard devia- measures such as the median value of 735, standard devia-
tion (SD) of 105.73 kg/m3, and variance of 11,178.51. Fig- tion (SD) of 155.84 kg/m3, and variance of 24,285.73. Fig-
ure 2c illustrates the relationship between the compressive ure 2d illustrates the relationship between the compressive
strength (CS) and the fine aggregate content, along with strength (CS) and the coarse aggregate content, along with
the Kurtosis and Skewness values of −0.212 and 0.28, the Kurtosis and Skewness values of 0.688 and −0.383,
respectively. respectively.
Innovative Infrastructure Solutions (2024) 9:412 Page 9 of 31 412

Fig. 2  (continued)

Superplasticizers Fly ash percentage

The superplasticizers (SP) in the different SCC mix designs The fly ash percentage (FA%) in the different SCC mix
ranged from 0 to 3.8%, as depicted in Fig. 2e. The analy- designs ranged from 0 to 70%, as depicted in Fig. 2f. The
sis also revealed additional statistical measures such as the analysis also revealed additional statistical measures such
median value of 1.25%, standard deviation (SD) of 0.774%, as the median value of 0%, standard deviation (SD) of
and variance of 0.6. Figure 2e illustrates the relationship 17.308%, and variance of 299.554. Figure 2f illustrates the
between the compressive strength (CS) and the superplas- relationship between the compressive strength (CS) and the
ticizers, along with the Kurtosis and Skewness values of fly ash percentage, along with the Kurtosis and Skewness
−0.447 and 0.198, respectively. values of 0.972 and 1.437, respectively.
 412 Page 10 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 2  (continued)

Silica fume percentage Curing time

The silica fume percentage (SF%) in the different SCC mix The curing time (t) in the different SCC mix designs ranged
designs ranged from 0 to 50%, as depicted in Fig. 2g. The from 1 to 180 days, as depicted in Fig. 2h. The analysis also
analysis also revealed additional statistical measures such as revealed additional statistical measures such as the median
the median value of 5%, standard deviation (SD) of 9.832%, value of 28 days, standard deviation (SD) of 36.476 days,
and variance of 96.667. Figure 2g illustrates the relationship and variance of 1330.528. Figure 2h illustrates the relation-
between the compressive strength (CS) and the silica fume ship between the compressive strength (CS) and the curing
percentage, along with the Kurtosis and Skewness values of time, along with the Kurtosis and Skewness values of 5.074
1.919 and 1.526, respectively. and 2.228, respectively.
Innovative Infrastructure Solutions (2024) 9:412 Page 11 of 31 412

Compressive strength in the range of 60–87 MPa. Additionally, the kurtosis and
skewness are 0.203 and 0.652, respectively.
The analyzed studies in this study have reported a wide
range of compressive strength (CS) values, ranging from Slump flow diameter
4.9 to 87 MPa. Statistical measures were calculated,
including a median of 35.25 MPa, a standard deviation The slump flow test is commonly used to evaluate the flow
(SD) of 14.837 MPa, and a variance of 220.207. Figure 3a performance of self-compacting concrete (SCC). It measures
illustrates the distribution of CS ranges, with 8.33% of the diameter of the concrete flow during the test. Accord-
the data clustered in the 0–20 MPa range, 44.67% in the ing to the European Federation of National Associations
20–35 MPa range, 48% in the 35–60 MPa range, and 9% Representing for Concrete (EFNARC) criteria for SCC, the
slump flow diameter should meet specific classes defined

Fig. 3  a Summary plot of the

compressive strength (CS, MPa)
concrete modified with Fly ash
or Silica fume. b Summary
plot of the slump flow diameter
(mm) concrete modified with
Fly ash or Silica fume
 412 Page 12 of 31 Innovative Infrastructure Solutions (2024) 9:412

in the standard. These classes are as follows: Class 1 with median slump flow diameter of 681 mm, a standard devia-
a slump flow diameter between 550 and 650 mm, Class 2 tion of 61.06 mm, and a variance of 3728.83. The kurtosis
with a diameter between 650 and 750 mm, and Class 3 with and skewness values are calculated as 1.506 and −0.994,
a diameter between 750 and 850 mm. Slump flow diameters respectively. Figure 3b illustrates the distribution of slump
below 550 mm or above 850 mm are not considered SCC flow diameter ranges. Figure 4 illustrates the variation
according to the EFNARC criteria [2]. between slump flow diameter and the other independent
Based on the 330 data points collected from previous variables.
studies (Table 1), it was found that 96.67% of the mixtures Furthermore, Fig. 5 displays the relationship between
can be classified as SCC since their slump flow diameters slump flow diameter and compressive strength (CS) of SCC
fall within the range of 550 mm to 850 mm; nonetheless, mixtures modified with silica fume or fly ash, indicating dif-
all the experimental data used are SCC. The data reveals a ferent slump flow classes. It is evident from Fig. 5 that there is

Fig. 4  Marginal plots for slump

flow diameter of concrete modi-
fied with fly ash or silica fume
versus, a water/cement ratio, b
cement content, c fine aggregate
content, d coarse aggregate
content, e superplasticizers, f fly
ash, g Silica fume, and h curing
time
Innovative Infrastructure Solutions (2024) 9:412 Page 13 of 31 412

Fig. 4  (continued)

no strong correlation between compressive strength and slump compressive strength (CS) and the slump flow diameter are
flow diameter. SCC mixtures must meet the required slump multivariable functions that depend on several variables. The
flow diameter class, while the compressive strength can vary. matrix analysis reveals the nature of these relationships. A
positive value indicates a direct correlation between two
variables, a negative value indicates an inverse correlation
Correlation matrix between independent between two variables, while a value of zero indicates no
variables and dependent variable relationship.
Upon analyzing the matrix, it can be observed that
The correlation coefficients between variables were calcu- cement and fly ash are inversely correlated. This implies that
lated using matrices, as shown in Fig. 6, where each cell as the cement content increases, the fly ash content tends
represents the relationship between two variables. The to decrease, and vice versa, and this makes sense since the
 412 Page 14 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 4  (continued)

fly ash is substituting the cement. On the other hand, slump mathematical parameters and expressions, and the goal is to
flow diameter is correlated with the curing time and cement identify the best-performing model among them. The SCC
by 0.73 and 0.53, respectively. Similarly, the compressive mixtures, which include water, cement, fine aggregate (FA),
strength is correlated with cement by 0.53. coarse aggregate (CA), superplasticizers (SP), silica fume,
and fly ash, are vital components that influence the CS. The
parameters used to anticipate the CS value are the water-
Models to-cement ratio (w/c), cement content (C), FA content, CA
content, fly ash percentage (FA%), silica fume percentage
Different mathematical models have been employed to pre- (SF%), superplasticizer percentage (SP%), and curing days.
dict the compressive strength (CS) of SCC modified with sil- Each of these compositions and parameters has a distinct
ica fume, fly ash, or both. These models incorporate various impact on the CS, which will be thoroughly analyzed in
Innovative Infrastructure Solutions (2024) 9:412 Page 15 of 31 412

Fig. 4  (continued)

this section. The dataset has been divided into three sets: fume percentage, t for curing time, and β0 to βn as the model
a training set comprising 70% of the total dataset and two parameters.
validation and testing sets containing 15% of the entire
dataset. The accuracy and credibility of each model will Linear regression model (LR)
be assessed based on the R­ 2 (coefficient of determination),
RMSE (root mean square error), and maximum absolute The compressive strength (CS) of the SCC, considering
error (MAE) values associated with each dataset. To clarify different mix design components, was estimated using a
the equations, specific notations have been defined: w/c for linear model, as presented in Table 1. Equation (1) repre-
water-to-cement ratio, C for cement content, FA for fine sents the mathematical expression utilized in this model,
aggregate content, CA for coarse aggregate content, SP for incorporating the parameters that influence the CS in lin-
superplasticizers, FA% for fly ash percentage, SF% for silica ear and constant terms. This linear model assumes a steady
412 Page 16 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 5  The relationship between

the slump flow diameter and
compressive strength of self-
compacting concrete (SCC)
mixtures modified with fly ash
(FA) and silica fume (SF)

rate of change and is relatively simple to implement due Pure quadratic (PQ)
to its straightforward mathematical operations. However, it
is essential to note that this model may have limitations in The pure quadratic (PQ) model, Eq. 3, will establish a rela-
accurately fitting the data, particularly when encountering tionship between the SCC components and predict their
outliers. Its main advantage lies in its ease of implemen- strength. This model incorporates linear terms, constant
tation, but its accuracy and reliability are relatively lower variables, and squared parameters to accurately estimate
than models. Based on the observed results, the linear model the compressive strength (CS) of the SCC. The relation-
assumes a linear relationship between the input and output ship between the independent variables (SCC mix propor-
variables [70]. tions) and the dependent variable (CS) can be effectively
captured by manipulating the mathematical parameters
CS = 𝛽0 + 𝛽1 ∗ (w∕c) + 𝛽2 ∗ (C) + 𝛽3 ∗ (FA) + 𝛽4 ∗ (CA) + 𝛽5 ∗ (SP)
within the PQ model. It should be noted that this model
+ 𝛽6 ∗ (FA%) + 𝛽7 ∗ (SF%) + 𝛽8 ∗ (t) presents a disadvantage in terms of its complex combina-
(1) tion of mathematical terms. The intricate nature of the
where β0 to β8 are defined as the model parameters. formulation of the PQ model may require a more thorough
understanding of the underlying mathematical concepts
and calculations [70, 75, 76].
Nonlinear regression model (NLR)
CS = 𝛽0 + 𝛽1 ∗ (w∕c) + 𝛽2 ∗ (C) + 𝛽3 ∗ (FA)
The compressive strength (CS) of the SCC, considering + 𝛽4 ∗ (CA) + 𝛽5 ∗ (SP) + 𝛽6 ∗ (FA%)
different mix design components, was predicted using the + 𝛽7 ∗ (SF%) + 𝛽8 ∗ (t) + 𝛽9 ∗ (w∕c)2
nonlinear regression (NLR) model. Equation (2) represents (3)
+ 𝛽10 ∗ (C)2 + 𝛽11 ∗ (FA)2
the mathematical expression utilized in this model, involving
the parameters influencing the CS of the SCC in exponential + 𝛽12 ∗ (CA)2 + 𝛽13 ∗ (SP)2
and constant terms. + 𝛽14 ∗ (FA%)2 + 𝛽15 ∗ (SF%)2 + 𝛽16 ∗ (t)2

CS = 𝛽0 + 𝛽1 ∗ (w∕c)𝛽2 + 𝛽3 ∗ (C)𝛽4 where β0 to β16 are defined as the model parameters.

+ 𝛽5 ∗ (FA)𝛽6 + 𝛽7 ∗ (CA)𝛽8
(2)
+ 𝛽9 ∗ (SP)𝛽10 + 𝛽11 ∗ (FA%)𝛽12
+ 𝛽13 ∗ (SF%)𝛽14 + 𝛽15 ∗ (t)𝛽16

where β0 to β16 are defined as the model parameters.

Innovative Infrastructure Solutions (2024) 9:412 Page 17 of 31 412

Coarse
Aggregate, -0.20
CA (kg/m3)

Fine
Aggregate, -0.15 -0.13
FA (kg/m3)

Water to
cement ratio, -0.21 0.04 -0.11
w/c

Silica Fume,
0.01 0.08 -0.14 0.12
SF%

Fly Ash,
-0.62 -0.09 0.04 -0.16 -0.36
FA%

Superplastici
-0.17 0.05 0.22 -0.07 0.27 0.14
zer, SP

Curing Time,
0.10 0.13 0.04 0.07 0.03 -0.18 -0.07
t (Day)

Slump Flow
0.53 0.10 0.00 -0.32 0.00 -0.39 -0.30 0.73
(mm)

Compressive
Strength, CS 0.53 -0.11 0.09 -0.23 -0.08 -0.30 0.10 0.09 0.34
(MPa)
Cement, C Coarse Fine Water to Silica Fly Ash, Superplasti Curing Slump
3 Aggregate, Aggregate, cement Fume, SF% FA% cizer, SP Time, t Flow (mm)
(kg/m )
3 3
CA (kg/m ) FA (kg/m ) ratio, w/c (Day)

Fig. 6  Correlation matrix plot for input and output variables of the concrete modified with FA or SF

Interaction (IA) quantitative variables of the SCC, incorporating linear

and constant terms to accurately predict the compressive
The interaction (IA) model will establish a relationship strength (CS). Equation (4) illustrates the mathematical
between the SCC components and predict the strength. parameters involved in the IA model, capturing the inter-
The IA model represents the interaction between the play between the independent variables and their impact
on the CS [77].

CS = 𝛽0 + 𝛽1 ∗ (w∕c) + 𝛽2 ∗ (C) + 𝛽3 ∗ (FA) + 𝛽4 ∗ (CA) + 𝛽5 ∗ (SP) + 𝛽6 ∗ (FA%)

+ 𝛽7 ∗ (SF%) + 𝛽8 ∗ (t) + 𝛽9 ∗ (w∕c ∗ C) + 𝛽10 ∗ (w∕c ∗ FA) + 𝛽11 ∗ (w∕c ∗ CA)
+ 𝛽12 ∗ (w∕c ∗ SP)𝛽13 ∗ (w∕c ∗ FA%) + 𝛽14 ∗ (w∕c ∗ SF%) + 𝛽15 ∗ (w∕c ∗ t)
+ 𝛽16 ∗ (C ∗ FA) + 𝛽17 ∗ (C ∗ CA) + 𝛽18 ∗ (C ∗ SP) + 𝛽19 ∗ (C ∗ FA%) + 𝛽20 ∗ (C ∗ SF%) (4)
+ 𝛽21 ∗ (C ∗ t) + 𝛽22 ∗ (FA ∗ CA) + 𝛽23 ∗ (FA ∗ SP) + 𝛽24 ∗ (FA ∗ FA%) + 𝛽25 ∗ (FA ∗ SF%)
+ 𝛽26 ∗ (FA ∗ t) + 𝛽27 ∗ (CA ∗ SP) + 𝛽28 ∗ (CA ∗ FA%) + 𝛽29 ∗ (CA ∗ SF%) + 𝛽30 ∗ (CA ∗ t)
+ 𝛽31 ∗ (SP ∗ FA%) + 𝛽32 ∗ (SP ∗ SF%) + 𝛽33 ∗ (SP ∗ t) + 𝛽34 ∗ (FA% ∗ SF%)
+ 𝛽35 ∗ (FA% ∗ t) + 𝛽36 ∗ (SF% ∗ t)
412 Page 18 of 31 Innovative Infrastructure Solutions (2024) 9:412

where β0 to β36 are defined as the model parameters. absolute error), scatter index (SI), and the objective function
(OBJ). These criteria were determined using specific equa-
tions, which are as follows [70, 75, 78, 79]:
Full quadratic (FQ) 2
⎛ p �
⎞
⎜ ∑ �� � ⎟
The full quadratic (FQ) model will establish a relation- ⎜ tp − t �
yp − y�
⎟
ship between the SCC components and accurately predict R2 = ⎜ ��
p=1
⎟ (6)
their strength. The FQ model incorporates the product/ ⎜� p �� �⎟
�
⎜� ∑ � �2 ∑�
p �2 ⎟
interaction of the quantitative variables of the SCC, linear ⎜ tp − t � yp − y� ⎟
terms, constant variables, and squared parameters. This ⎝ p=1 p=1 ⎠
advanced mathematical combination aims to produce the
best-fitting predictions for the SCC compressive strength �∑
p
(yp − tp )2
(CS). Equation (5) illustrates the mathematical parameters RMSE =
p=1
(7)
involved in the FQ model, capturing the intricate relation- p
ship between the independent variables and their impact
on the CS. Although the FQ model is known for its accu- ∑p
�(yp − tp )�
racy, its drawback lies in the complexity of the mathemati-
p=1
MAE = (8)
cal terms involved [70, 75, 78]. p

CS = 𝛽0 + 𝛽1 ∗ (w∕c) + 𝛽2 ∗ (C) + 𝛽3 ∗ (FA) + 𝛽4 ∗ (CA) + 𝛽5 ∗ (SP) + 𝛽6 ∗ (FA%)

+ 𝛽7 ∗ (SF%) + 𝛽8 ∗ (t) + +𝛽9 ∗ (w∕c ∗ C) + 𝛽10 ∗ (w∕c ∗ FA) + 𝛽11 ∗ (w∕c ∗ CA)
+ 𝛽12 ∗ (w∕c ∗ SP) + 𝛽13 ∗ (w∕c ∗ FA%) + 𝛽14 ∗ (w∕c ∗ SF%) + 𝛽15 ∗ (w∕c ∗ t) + 𝛽16 ∗ (C ∗ FA)
+ 𝛽17 ∗ (C ∗ CA) + 𝛽18 ∗ (C ∗ SP) + 𝛽19 ∗ (C ∗ FA%) + 𝛽20 ∗ (C ∗ SF%) + 𝛽20 ∗ (C ∗ t)
+ 𝛽21 ∗ (FA ∗ CA) + 𝛽22 ∗ (FA ∗ SP) + 𝛽23 ∗ (FA ∗ FA%) + 𝛽24 ∗ (FA ∗ SF%) + 𝛽25 ∗ (FA ∗ t) (5)
+ 𝛽26 ∗ (CA ∗ SP) + 𝛽27 ∗ (CA ∗ FA%) + 𝛽28 ∗ (CA ∗ SF%) + 𝛽29 ∗ (CA ∗ t)
+ 𝛽30 ∗ (SP ∗ FA%) + 𝛽31 ∗ (SP ∗ SF%) + 𝛽32 ∗ (SP ∗ t) + 𝛽33 ∗ (FA% ∗ SF%)
+ 𝛽34 ∗ (FA% ∗ SF%) + 𝛽35 ∗ (FA% ∗ t) + 𝛽36 ∗ (SF% ∗ t) + 𝛽37 ∗ (w∕c)2 + 𝛽38 ∗ (C)2
+ 𝛽39 ∗ (FA)2 + 𝛽40 ∗ (CA)2 + 𝛽41 ∗ (SP)2 + 𝛽42 ∗ (FA%)2 + 𝛽43 ∗ (SF%)2 + 𝛽44 ∗ (t)2

where β0 to β44 are defined as the model parameters. RMSE

SI = (9)
t�
Artificial neural networks (ANN)
( ) ( )
ntr RMSEtr + MAEtr ntst RMSEts + MAEts
OBJ = ∗ + ∗
An artificial neural network (ANN) is a computational model nall R2tr + 1 nall R2ts + 1
that draws inspiration from the structure and functionality of ( )
nval RMSEva + MAEva
the neural networks of the human brain. ANNs are designed + ∗
nall R2va + 1
to process and learn from large datasets, making them valu- (10)
able tools for predicting outcomes based on input data. These
In these equations, "yp" and "tp" represent the predicted
networks can be trained and validated using efficient methods,
and measured values for the ­pth pattern, while "t'" and "y'"
resulting in accurate predictions. The ability of ANN to learn
indicate the averages of the measured and predicted values.
and make connections within data sets makes them effective
The terms "tr," "ts," and "va" refer to the training, testing,
analytical tools for generating reliable forecasts [75].
and validation datasets, respectively, and "n" represents the
number of patterns in each dataset.
Assessment criteria
For all assessment parameters except R ­ 2, the ideal value is
2
zero. However, for R­ , the ideal value is one. The SI param-
The performance evaluation of the models was conducted
eter can be used to evaluate the performance of the model.
using various parameters to compare their outputs. The
A model is seen to have subpar performance when the Struc-
effectiveness, credibility, and reliability of each model were
tural Index (SI) exceeds 0.3. It is considered to have moder-
assessed based on several criteria, including RMSE (root mean
ate performance when the SI falls between 0.2 and 0.3. A
­ 2 (coefficient of determination), MAE (mean
square error), R
Innovative Infrastructure Solutions (2024) 9:412 Page 19 of 31 412

model performs satisfactorily when the SI ranges from 0.1 CS = 23.45 − 74.43 ∗ w∕c + 0.059 ∗ C
to 0.2. Finally, a model is classified as having exceptional
+ 0.011 ∗ FA + 0.01 ∗ CA
performance when the SI is less than 0.1[80, 81]. (11)
In addition, the OBJ parameter, as mentioned in − 4.67 ∗ SP − 0.019 ∗ FA%
Eq. (10), was used as an integrated performance parameter + 0.116 ∗ SF% + 0.274 ∗ t
to assess the efficiency of the proposed models.
The LR statistical analysis yielded values of 5.08 MPa for
RMSE and 0.88 for ­R2, as depicted in Fig. 7. These values
indicate that the performance of the LR model in predicting
Results and discussion the compressive strength (CS) of the SCC modified with
silica fume, fly ash, or both is capable of predicting the CS
Predicted and measured CS relationship of SCC. It should be noted that this model serves as the
foundation for all other models. However, when evaluating
Linear regression model (LR) the LR model using the testing and validating databases, the
RMSE values were found to be 5.37 MPa and 5.55 MPa,
The Linear Regression (LR) model was utilized to predict with corresponding ­R2 values of 0.87 and 0.86, as shown
SCC mix designs' compressive strength (CS), specifically in Fig. 7.
those modified with NS. However, due to its simplistic
mathematical expression, the LR model proved to be one Nonlinear regression model (NLR)
of the least effective models for CS prediction. This model
is based solely on linear and constant terms for its math- The NLR model was utilized to predict the compressive
ematical parameters. Equation (11) illustrates the relation- strength (CS) of the SCC modified with silica fume, fly ash,
ship between the parameters in detail. The LR model was or both. This model proves more effective than the linear
trained using a database and then applied to the testing and model due to its advanced mathematical expressions. The
validation datasets to evaluate its credibility and reliabil- NLR model incorporates constant terms and terms raised to
ity. Notably, the parameters that significantly influenced the power of constant variables as its mathematical param-
the predicted CS in this Equation were the cement content, eters. Equation (12) outlines these parameters and their
which contributed to its increment, and the curing time, relationships in detail. The Equation was derived from the
which led to its decrement. training database and applied to the testing and validating

Fig. 7  Comparison between

measured and predicted com-
pressive strength of concrete
modified with Fly ash or silica
fume using linear regression
(LR) model
412 Page 20 of 31 Innovative Infrastructure Solutions (2024) 9:412

databases to assess its credibility and reliability. Notably, in parameters and their relationships in detail. This Equation
this Equation, the parameters FA, cement, and curing time was derived from the training database and applied to the
significantly impact the predicted CS, influencing its incre- testing and validating databases to assess its credibility and
ment or decrement. reliability. Notably, the FA content, water-to-cement ratio

CS = − 30.87 + 27.34 ∗ w∕c−0.65 + 1.98 ∗ FA−3.52 + 1.95 ∗ CA−3.66

(12)
− 2.94 ∗ SP1.13 − 4.74 ∗ FA%−0.079 + 1.3 ∗ SF%0.012 + 1.19 ∗ t0.72

The NLR statistical analysis yielded the following results: (w/c), cement content, and curing time are the prominent
an RMSE value of 5.34 MPa and an ­R2 value of 0.87, as parameters influencing the predicted CS, with each param-
depicted in Fig. 8. These values indicate that the NLR model eter either increasing or decreasing the CS value.
performs credibly and reliably in predicting the compressive
strength (CS) of the SCC modified with silica fume, fly ash, CS = − 26.78 − 129.2 ∗ (w∕c) + 0.072 ∗ (C)
or both. Furthermore, the testing and validating databases + 0.133 ∗ (FA) + 0.031 ∗ (CA)
were examined, revealing RMSE values of 5.81 MPa and − 5.87 ∗ (SP) − 0.115 ∗ (FA%) (13)
5.42 MPa, along with R ­ 2 values of 0.84 and 0.87, respec-
+ 0.177 ∗ (SF%) + 0.354 ∗ (t)
tively, as shown in Fig. 8. These results reinforce the effec-
tiveness and reliability of the NLR model in predicting CS. + 62.1 ∗ (w∕c)2 + 0.445 ∗ (SP)2

The PQ statistical analysis has yielded RMSE and ­R2 val-

Pure quadratic (PQ)
ues of 4.91 MPa and 0.89, respectively, as depicted in Fig. 9.
These values indicate that the PQ model performs reason-
The PQ model was employed to forecast the compressive
ably well in predicting the compressive strength (CS) of
strength (CS) of SCC mix designs, particularly those modi-
SCC modified with silica fume, fly ash, or both. It represents
fied with fly ash, silica fume, or both. This model incorpo-
an enhanced version of the LR and NLR models. However,
rates various mathematical parameters, including constant
when applied to the testing and validating databases, the PQ
terms, linear terms, and squared variables. The PQ model is
model yielded RMSE and ­R2 values of 5.14 MPa and 0.88
derived from the mathematical parameters utilized in the FQ
and 5.00 MPa and 0.89, respectively, as shown in Fig. 9.
and LR models for CS prediction. Equation (13) presents the

Fig. 8  Comparison between

measured and predicted com-
pressive strength of concrete
modified with Fly ash or silica
fume using nonlinear regression
(NLR) model
Innovative Infrastructure Solutions (2024) 9:412 Page 21 of 31 412

Fig. 9  Comparison between

measured and predicted com-
pressive strength of concrete
modified with Fly ash or silica
fume using pure quadratic (PQ)
model

These results suggest that the model's performance remains values indicate that the IA model exhibits high credibility
consistent and reliable across different datasets. and reliability in predicting the CS of SCC modified with fly
ash, silica fume, or both. However, when evaluated using the
Interaction (IA) testing and validating databases, the IA model has RMSE
and ­R2 values of 4.95 MPa and 0.89 and 5.42 MPa and 0.87,
The IA model was employed to predict SCC mix designs' respectively, as shown in Fig. 10. These results demonstrate
compressive strength (CS), particularly those modified with that the model maintains consistent and reliable performance
silica fume, fly ash, or both. This model proved to be one across different datasets.
of the most effective models in CS prediction due to its
advanced mathematical expression. It was derived based on Full quadratic (FQ)
constant, linear, and variable product terms, allowing for
direct variation in predicted CS values. Equation (14) details The FQ model was employed to predict the compressive
the relationship and parameters of the IA model. The Equa- strength (CS) of SCC modified with silica fume, fly ash,
tion was derived from the training database and applied to or both. This model has proven to be the most effective in
the testing and validating databases to assess its credibility predicting CS due to its advanced mathematical expression.
and reliability. Notable parameters in this Equation include It incorporates constant, linear, variable product terms/inter-
fly ash content, water-to-cement ratio (w/c), cement content, action and squared variables as its mathematical parameters.
coarse aggregate (CA) content, and curing time, all contrib- The relationship between these parameters is detailed in
uting to the increase or decrease in predicted CS. Eq. (15). The Equation was derived from the training data-

CS = − 0.012 + 0.777 ∗ w∕c + 0.091 ∗ C + 0.0249 ∗ FA − 0.064 ∗ CA + 26.73 ∗ SP

− 0.199 ∗ FA% + 0.225 ∗ SF% + 1.049 ∗ t − 0.091 ∗ w∕c ∗ C − 0.033 ∗ w∕c ∗ FA
+ 0.304 ∗ w∕c ∗ CA − 25.56 ∗ w∕c ∗ SP − 0.201 ∗ w∕c ∗ FA% + 0.17 ∗ w∕c ∗ SF% (14)
− 0.501 ∗ w∕c ∗ t − 0.027 ∗ C ∗ SP − 0.012 ∗ FA ∗ SP − 0.0004 ∗ FA ∗ t
+ 0.004 ∗ CA ∗ SP − 0.165 ∗ SP ∗ FA% − 0.114 ∗ SP ∗ SF%

­ 2 values of
The IA statistical analysis yielded RMSE and R base and applied to the testing and validating databases to
4.60 MPa and 0.90, respectively, as shown in Fig. 10. These assess its credibility and reliability. Notable parameters in
412 Page 22 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 10  Comparison between

measured and predicted com-
pressive strength of concrete
modified with Fly ash or silica
fume using interactive (IA)
model

this Equation include the water-to-cement ratio (w/c), fly Consequently, it is impossible to identify which variables
ash (FA) content, coarse aggregate (CA) content, and silica affected the compressive strength the most. Figure 12a pro-
fume (SF%) content, all of which contribute to the increase vides a visual representation of the functioning of the ANN
or decrease in predicted CS. model. The neural network training was conducted using the

S = − 0.0394 + 0.071 ∗ w∕c − 0.07 ∗ C + 0.192 ∗ FA − 0.164 ∗ CA + 0.669 ∗ SP

− 1.34 ∗ FA% + 2.42 ∗ SF% + 1.918 ∗ t − 0.132 ∗ w∕c ∗ C − 0.131 ∗ w∕c ∗ FA
(15)
+ 0.156 ∗ w∕c ∗ C − 0.445 ∗ w∕c ∗ SP − 0.777 ∗ w∕c ∗ FA% − 0.338 ∗ w∕c ∗ SF%
− 0.959 ∗ w∕c ∗ t + 1.454 ∗ C ∗ FA − 0.024 ∗ C ∗ SP + 0.0025 ∗ C ∗ FA%
− 0.0012 ∗ C ∗ t + 0.007 ∗ CA ∗ SP + 0.13 ∗ (w∕c)2

The FQ statistical analysis yielded a root mean square neural fitting feature in MATLAB.
error (RMSE) value of 4.37 MPa and an ­R2 value of 0.91, The ANN statistical analysis yielded a root mean square
as shown in Fig. 11. These values indicate that the FQ error (RMSE) value of 3.44 MPa and an R ­ 2 value of 0.95, as
model performs credibly and reliably in predicting the shown in Fig. 12. These values indicate that the ANN model
compressive strength (CS) of SCC modified with silica performs credibly and reliably in predicting the compressive
fume, fly ash, or both. Furthermore, when evaluated with strength (CS) of SCC modified with silica fume, fly ash, or
the testing and validating databases, the FQ model pro- both. Furthermore, when evaluated with the testing and vali-
duced RMSE values of 4.26 MPa and 4.86 MPa, along dating databases, the ANN model produced RMSE values
with ­R2 values of 0.91 and 0.90, respectively, as shown of 3.44 MPa and 4.18 MPa, along with ­R2 values of 0.94
in Fig. 11. These results further affirm the credibility and and 0.92, respectively, as shown in Fig. 12b. These results
reliability of the model in predicting CS. further affirm the credibility and reliability of the model in
predicting CS.
Artificial neural networks (ANN)
Models comparison
An artificial neural network was also utilized to predict the
compressive strength of the SCC modified with fly ash, The efficiency of the proposed models was evaluated using
silica fume, or both. A neural network of 10 nodes was five statistical tools: RMSE, MAE, SI, OBJ, and ­R2. Among
trained using the Levenberg–Marquardt training algorithm. the analyzed models, the ANN model outperformed the FQ,
Innovative Infrastructure Solutions (2024) 9:412 Page 23 of 31 412

Fig. 11  Comparison between

measured and predicted com-
pressive strength of concrete
modified with Fly ash or silica
fume using full quadratic (FQ)
model

IA, PQ, LR, and NLR models, exhibiting higher R ­ 2 and This indicates that the ANN model is more efficient in esti-
lower RMSE and MAE values. Using testing data, Fig. 13a mating the compressive strength of SCC mixtures modified
compares the predicted CS of the various models for SCC with silica fume or fly ash.
mixes modified with silica fume or fly ash. Figure 13b shows The SI assessment parameter values for the proposed
the median line of the CS and outliers of all the models in models in the training, validating, and testing phases are
this study. Additionally, Fig. 14 illustrates the residual errors depicted in Fig. 16. Figure 16 shows that all models per-
for all models using training, testing, and validating datasets. formed well across phases, with SI values between 0.1 and
It can be observed from both figures that the ANN model 0.2. However, similar to other performance parameters, the
demonstrates a closer alignment between the predicted and ANN model exhibited lower SI values than the other models
measured values of compressive strength, indicating its across the training, testing, and validation datasets and has
superior performance compared to other models. Likewise, less than 0.1 SI for the training and testing datasets.
it is evident from the figures that the LR model is the poorest In the training phase, the ANN model exhibited SI val-
performer among the models. This can be attributed to the ues that were 21.2% lower than the FQ model, 25.1% lower
simplicity of the linear model. than the IA model, 29.9% lower than the PQ model, 32.2%
Furthermore, the ANN model accurately predicts the lower than the LR model, and 35.5% lower than the NLR
CS of SCC across all ranges. In the ranges of 0–20 MPa, model. Similarly, in the testing phase, the ANN model dis-
20–35 MPa, 35–60 MPa, and 60–87 MPa, the ANN model played SI values that were 19.2% lower than the FQ model,
exhibits the highest ­R2 values and the lowest RMSE, MAE, 30.5% lower than the IA model, 33.2% lower than the PQ
and SI values (Table 2), making it the superior model. How- model, 36% lower than the LR model, and 40.9% lower than
ever, between 20–35 MPa, the MAE of the FQ model is the NLR model. Lastly, in the validating dataset, the ANN
1.88 MPa, the smallest compared to the other models. model showcased SI values that were 14% lower than the FQ
Figure 15 provides the OBJ values for all proposed mod- model, 16.5% lower than the PQ model, 22.9% lower than
els for the training dataset. The OBJ values for LR, NLR, the IA model, 22.9% lower than the NLR model, and 24.7%
PQ, IA, FQ, and ANN are 6.65, 6.97, 6.33, 5.98, 5.5, and lower than the LR model in the validating phase.
4.18 MPa, respectively. The ANN model exhibits a lower This analysis highlights that the ANN model is the most
OBJ value, which is 24.1% less than the FQ model, 30.2% effective and performs better than the other models in pre-
less than the IA model, 34% less than the PQ model, 37.2% dicting the CS of SCC mixtures modified with silica fume
less than the LR model, and 40.1% less than the NLR model. or fly ash.
412 Page 24 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 12  a ANN Model b Com-

parison between measured and
predicted compressive strength
of concrete modified with
Fly ash or silica fume using
artificial neural network (ANN)
model

Sensitivity analysis the output element in the FQ model since the ANN model
can be trained to predict the values of CS effectively even
The sensitivity analysis aimed to evaluate the impact of when one of the parameters is removed. Therefore, the FQ
each input parameter of the mix design on the behavior of model is chosen as the best performer for predicting the
Innovative Infrastructure Solutions (2024) 9:412 Page 25 of 31 412

Fig. 13  a Variation in predicted

and measured compressive
strength values of concrete
modified with Fly ash or Silica
fume based on six different
models (b) box plot of all the
models

CS of SCC to evaluate which parameter affects the CS Limitations of this research

of SCC, as indicated in Table 3. The sensitivity analysis
showed that removing the fine aggregate parameter sig- The limitations of this study include its exclusive focus on
nificantly influenced the CS values of SCC when using predicting compressive strength (CS) of self-compacting
the FQ model since the RMSE and MAE have risen to concrete (SCC) modified with silica fume, fly ash, or both
70.21 and 67.20 MPa, respectively. On the other hand, the using limited variables. The study's reliance on mathemati-
percentage of silica fume (SF%) had the most negligible cal models and statistical analysis poses another limitation,
impact on the predicted CS values, as evidenced by lower as these models might not fully account for the complex
RMSE and MAE and higher R ­ 2 compared to the exclusion real-world behavior of concrete due to simplifications and
of other parameters. The exclusion of SF% has resulted assumptions. Uncertainties in material properties, envi-
in higher RMSE and MAE, 4.89 and 4 MPa, respectively, ronmental conditions, and construction practices could
and lower ­R2, 0.89. affect prediction accuracy. Additionally, the study could
 412 Page 26 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 14  Residual error of com- LR NLR

pressive strength of concrete 15
modified with Fly ash or Silica (a) Training Testing Validating
fume using training, testing,
and validating datasets of (a) 10
linear regression (LR), nonlin-

Residual Error (MPa)

ear regression (NLR), b pure 5
quadratic (PQ), interaction (IA),
and c artificial neural network
(ANN), full quadratic (FQ) 0
models
-5

-10

-15
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320
Data Set Number

QR IA
15
(b) Training Testing Validating

10
Residual Error (MPa)

-5

-10

-15
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320
Data Set Number

FQR ANN
25
(c) Training Testing Validating
20
15
Residual Error (MPa)

10
5
0
-5
-10
-15
-20
-25
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320
Data Set Number
Innovative Infrastructure Solutions (2024) 9:412 Page 27 of 31 412

Table 2  The comparison of Compressive Number of data R2 RMSE (MPa) MAE (MPa) SI Remarks
models in different compressive strength range
strength ranges (MPa)

LR 0–20 25 0.95 4.99 3.91 0.34 Excellent

20–35 134 0.75 5.08 0.65 0.17 Good
35–60 144 0.79 5.03 0.45 0.12 Good
60–90 27 0.96 6.60 5.60 0.10 Excellent
NLR 0–20 25 0.97 4.17 3.37 0.28 Excellent
20–35 134 0.72 5.36 2.80 0.18 Good
35–60 144 0.74 5.55 2.02 0.13 Good
60–90 27 0.97 6.05 4.61 0.09 Excellent
PQ 0–20 25 0.96 4.43 3.72 0.30 Excellent
20–35 134 0.79 4.67 1.07 0.16 Good
35–60 144 0.78 5.16 1.22 0.12 Good
60–90 27 0.97 5.73 4.45 0.08 Excellent
FQ 0–20 25 0.97 4.08 3.15 0.28 Excellent
20–35 134 0.83 4.22 1.88 0.15 Very good
35–60 144 0.81 4.72 2.83 0.11 Very good
60–90 27 0.98 4.26 2.75 0.06 Excellent
ANN 0–20 25 0.98 2.89 2.45 0.20 Excellent
20–35 134 0.91 3.18 2.97 0.11 Excellent
35–60 144 0.85 3.99 2.54 0.09 Very good
60–90 27 0.99 3.50 2.43 0.05 Excellent
IA 0–20 25 0.95 4.86 3.73 0.33 Excellent
20–35 134 0.79 4.61 1.95 0.16 Good
35–60 144 0.82 4.57 2.03 0.11 Very good
60–90 27 0.96 6.39 5.36 0.09 Excellent

Conclusions
ANN LR
LR 12% 19%
FQ This study aimed to develop models for predicting the com-
NLR 15%
pressive strength (CS) of self-compacting concrete (SCC)
NLR
PQ 19% modified with silica fume and fly ash using 330 datasets.
IA IA Several statistical models, including LR, NLR, PQ, FQ, and
17% PQ
FQ 18% ANN, were implemented and evaluated for their perfor-
ANN mance in predicting CS.

(1) Based on the analysis and comparison of the models,

the ANN model emerged as the most effective in pre-
Fig. 15  The objective function (OBJ) function values in (MPa) for all dicting the CS of SCC. It exhibited higher ­R2 values
developed models for compressive strength of concrete modified with and lower RMSE, MAE, and SI values than the other
Fly ash or Silica fume
models. The ANN model demonstrated its superiority
across 0 – 87 MPa CS.
have incorporated more CS-determining factors like curing (2) Among the proposed models, the ANN model achieved
temperature, fineness, aggregate particle diameter, and air the lowest OBJ value, indicating its superior perfor-
entrainment. However, the absence of consistent data across mance in minimizing prediction errors compared to the
experimental datasets hindered the comprehensive inclusion other models. The lower OBJ value of the ANN model
of these parameters.. suggests that it provides a more accurate estimation of
the CS of SCC mixtures modified with silica fume or
fly ash. This highlights the effectiveness of the ANN
 412 Page 28 of 31 Innovative Infrastructure Solutions (2024) 9:412

Fig. 16  Comparison of the SI 0.4

performance parameter of dif-
Training Testing Validating

Poor
ferent developed models for the
training, testing, and validating 0.35
datasets for all the models
0.3

Fair
0.25

Scatter Index
0.2

Good
0.15

0.1

Excelent
0.05

0
LR NLR QR FQR IA ANN
Models

Table 3  Sensitivity analysis using full quadratic model

No. Input parameters Removed parameters R2 RMSE (MPa) MAE (MPa) SI SD Ranking

1 Cement, coarse aggregate, fine aggregate, w/c, silica None 0.91 4.43 3.62 0.12 14.26 9
fume, fly ash, superplastizer, curing time
2 Coarse aggregate, fine aggregate, w/c, silica fume, fly Cement 0.58 26.53 19.91 0.71 33.89 3
ash, superplastizer, curing time
3 Cement, fine aggregate, w/c, silica fume, fly ash, super- Coarse aggregate 0.72 14.26 11.19 0.38 24.38 5
plastizer, curing time
4 Cement, coarse aggregate, w/c, silica fume, fly ash, Fine aggregate 0.60 70.21 67.20 1.87 30.22 1
superplastizer, curing time
5 Cement, coarse aggregate, fine aggregate, silica fume, w/c 0.58 41.58 37.3 1.11 27.67 2
fly ash, superplastizer, curing time
6 Cement, coarse aggregate, fine aggregate, w/c, fly ash, Silica fume 0.89 4.89 4.00 0.13 14.21 8
superplastizer, curing time
7 Cement, coarse aggregate, fine aggregate, w/c, silica Fly ash 0.77 7.49 5.41 0.20 13.59 6
fume, superplastizer, curing time
8 Cement, coarse aggregate, fine aggregate, w/c, silica Superplasticizer 0.86 6.37 5.06 0.17 13.70 7
fume, fly ash, curing time
9 Cement, coarse aggregate, fine aggregate, w/c, silica Curing time 0.43 15.49 11.65 0.41 9.63 4
fume, fly ash, superplastizer

model in achieving the optimization goal and minimiz- affecting the accuracy of the FQ model. On the other
ing the discrepancies between predicted and actual CS hand, the Silica Fume percentage (SF%) has the most
values. negligible effect on the predicted CS values.
(3) Sensitivity analysis was conducted to assess the influ- (4) Predicting the slump flow diameter of SCC is impossi-
ence of individual mix design parameters on the model ble using the current models since it has to be between
predictions. It was found that fine aggregate (FA) sig- 550 and 850 mm; the inclusion of more variables and
nificantly impacts the predicted CS values, negatively
Innovative Infrastructure Solutions (2024) 9:412 Page 29 of 31 412

larger datasets might make it possible to predict it using 16. Shariati M, Mafipour MS, Ghahremani B, Azarhomayun
deep learning or ensemble methods. F, Ahmadi M, Trung NT, Shariati A (2022) A novel hybrid
extreme learning machine–grey wolf optimizer (ELM-GWO)
model to predict compressive strength of concrete with partial
replacements for cement. Eng Comput 38:1–23
Funding This work had no funding. 17. Ahmaruzzaman M (2010) A review on the utilization of fly ash.
Prog Energy Combust Sci 36(3):327–363
Data availability The data supporting the conclusions of this article 18. Mukherjee AB, Zevenhoven R, Bhattacharya P, Sajwan KS,
are included in the article. Kikuchi R (2008) Mercury flow via coal and coal utilization
by-products: a global perspective. Resour Conserv Recycl
Declarations 52(4):571–591
19. Khan MI, Siddique R (2011) Utilization of silica fume in con-
Conflict of interest The authors declare that they have no competing crete: review of durability properties. Resour Conserv Recycl
interests. 57:30–35
20. Siddique R (2011) Utilization of silica fume in concrete: review
of hardened properties. Resour Conserv Recycl 55(11):923–932
21. Beshr H, Almusallam AA, Maslehuddin M (2003) Effect of
References coarse aggregate quality on the mechanical properties of high
strength concrete. Constr Build Mater 17(2):97–103
1. Ozawa K, Maekawa HK, Kunishima M and Okamura H (1989) 22. Güneyisi E, Gesoglu M, Al-Goody A, İpek S (2015) Fresh and
Performance of concrete based on the durability design of con- rheological behavior of nano-silica and fly ash blended self-
crete structures. In: Proceedings of the second east Asia‐Pacific compacting concrete. Constr Build Mater 95:29–44
conference on structural engineering and construction 23. Faraj RH, Mohammed AA, Omer KM, Ahmed HU (2022) Soft
2. EFNARC S (2002) Guidelines for self-compacting concrete. computing techniques to predict the compressive strength of
Association House: London, UK, pp 32–34 green self-compacting concrete incorporating recycled plastic
3. Okamura H (1997) Self-compacting high-performance concrete. aggregates and industrial waste ashes. Clean Technol Environ
Concr Int 19(7):50–54 Policy 24(7):2253–2281
4. Yahia A, Tanimura M, Shimabukuro A and Shimoyama Y 24. Faraj RH, Ahmed HU, Rafiq S, Sor NH, Ibrahim DF, Qaidi SM
(1999) Effect of rheological parameters on self compactability (2022) Performance of self-compacting mortars modified with
of concrete containing various mineral admixtures. The first nanoparticles: a systematic review and modeling. Clean Mater
RILEM international symposium on self-compacting concrete. 4:100086
In: Proceedings book, Stockholm, pp 523–535 25. Dinakar P, Manu SN (2014) Concrete mix design for high
5. Perssoiv B (1998) Technical report, japan society for the promo- strength self-compacting concrete using metakaolin. Mater Des
tion of science, report 9803: 12, Division of building materials, 60:661–668
Lund institute of technology, Lund 26. Ahmadi MA, Alidoust O, Sadrinejad I, Nayeri M (2007)
6. Flatt RJ, Roussel N, Cheeseman CR (2012) Concrete: an Development of mechanical properties of self compacting
eco material that needs to be improved. J Eur Ceram Soc concrete contain rice husk ash. Int J Comput Inf Syst Sci Eng
32(11):2787–2798 1(4):259–262
7. Gokulnath V, Ramesh B and Reddy SS (2020) Addition of 27. Ahmed HU, Mohammed AA, Mohammed AS (2022) The role of
reinforcing materials in self compacting concrete. In: Materi- nanomaterials in geopolymer concrete composites: a state-of-the-
als today: proceedings, 22: 722–725 art review. J Build Eng 49:104062
8. Dinesh A, Harini S, Jeba JP, Jincy J, Shagufta J (2017) Experi- 28. Yildirim G, Sahmaran M, Ahmed HU (2015) Influence of
mental study on self-compacting concrete. Int J Eng Sci Res hydrated lime addition on the self-healing capability of high-
Technol 6(3):42–50 volume fly ash incorporated cementitious composites. J Mater
9. Satish K, Kumar S, Rai B (2017) Self compacting concrete Civ Eng 27(6):04014187
using fly ash and silica fumes as pozzolanic material. J Eng 29. Aslam F, Zaid O, Althoey F, Alyami SH, Qaidi SM, de Prado Gil
Technol 6(2):394–407 J, Martínez-García R (2022) Evaluating the influence of fly ash
10. Yazicioglu S, Caliskan S, Türk K (2006) Effect of curing condi- and waste glass on the characteristics of coconut fibers reinforced
tions on the engineering properties of self-compacting concrete. concrete. Struct Concr 24:2440
Indian J Eng Mater Sci 13:25–29 30. Okamura H and Ouchi M (2003) Applications of self-compacting
11. Faraj RH, Sherwani AFH, Daraei A (2019) Mechanical, frac- concrete in Japan. In: The 3rd international RILEM symposium
ture and durability properties of self-compacting high strength on self-compacting concrete. In: Wallevik OH, Nielsson I (Eds.),
concrete containing recycled polypropylene plastic particles. J RILEM Publications SARL: Bagneux, France, pp 3–5
Build Eng 25:100808 31. Siddique R (2011) Properties of self-compacting concrete contain-
12. Faraj RH, Sherwani AFH, Jafer LH, Ibrahim DF (2021) Rheo- ing class F fly ash. Mater Des 32(3):1501–1507
logical behavior and fresh properties of self-compacting high 32. Nadhim Hamah SOR, Abdulwahid N (2018) The effect of super-
strength concrete containing recycled PP particles with fly ash plasticizer dosage on fresh properties of self-compacting light-
and silica fume blended. J Build Eng 34:101667 weight concrete produced with coarse pumice aggregate. J Gar-
13. Faraj RH, Ali HFH, Sherwani AFH, Hassan BR, Karim H mian Univ 5(2):190–209
(2020) Use of recycled plastic in self-compacting concrete: a 33. Barkhordari MS, Barkhordari MM, Armaghani DJ, Rashid ASA,
comprehensive review on fresh and mechanical properties. J Ulrikh DV (2022) Hybrid wavelet scattering network-based model
Build Eng 30:101283 for failure identification of reinforced concrete members. Sustain-
14. Neville AM, Brooks JJ (1987) Concrete technology, vol 438. ability 14(19):12041
Longman Scientific & Technical, England 34. Quercia G, Spiesz P, Hüsken G, Brouwers HJH (2014) SCC modi-
15. Neville AM (1995) Properties-of-concrete-AM 149, 4th edn. fication by use of amorphous nano-silica. Cement Concr Compos
Addison Wesley Longman Ltd., Essex 45:69–81
 412 Page 30 of 31 Innovative Infrastructure Solutions (2024) 9:412

35. Corinaldesi V, Moriconi G (2011) Characterization of self-com- 52. Kennouche S, Zerizer A, Benmounah A, Hami B, Mahdad M,
pacting concretes prepared with different fibers and mineral addi- Benouali H, Bedjou S (2013) Formulation and characterization
tions. Cement Concr Compos 33(5):596–601. [Link] of self-compacting concrete with silica fume. J Eng Technol Res
1016/j.​cemco​ncomp.​2011.​03.​007 5(5):160–169. [Link]
36. Şahmaran M, Yaman İÖ, Tokyay M (2009) Transport and mechan- 53. Khodabakhshian A, Brito J, Ghalehnovi M, Shamsabadi EA
ical properties of self consolidating concrete with high volume fly (2018) Mechanical, environmental and economic performance
ash. Cement Concr Compos 31(2):99–106 of structural concrete containing silica fume and marble industry
37. Muñoz Pérez SP, Pardo Becerra JM, García Chumacero JM, waste powder. Constr Build Mater 169:237–251. [Link]
Sánchez Diaz E, Diaz Ortiz EA, Laffite EDR, Briceño Mendoza 10.​1016/j.​conbu​ildmat.​2018.​02.​192
YM (2024) Evaluation of the physical and mechanical properties 54. Choudhary R, Gupta R, Nagar R (2020) Impact on fresh, mechani-
of concrete with the incorporation of recycled concrete aggregate. cal, and microstructural properties of high strength self-compact-
Innov Infrastruct Solut 9(6):1–12 ing concrete by marble cutting slurry waste, fly ash, and silica
38. Pérez SM, Segura MR, Bustamante YA, Zapata LV (2024) Study fume. Constr Build Mater 239:117888. [Link]
of the combined effect of coffee husk ash and polypropylene fibres conbu​ildmat.​2019.​117888
on the mechanical properties of concrete. J Appl Res Technol 55. Mahalakshmi SHV and Khed VC (2020) Experimental study on
22(1):32–41 M-sand in self-compacting concrete with and without silica fume.
39. Gamboa O, Leonardo J, Pérez SPM, Piedra JLL, Garcia J, Alvarez In: Materials today: proceedings, 27:1061-1065
J (2024) Effect of corn stover ash reinforced with cabuya fiber on 56. Bouzoubaâ N, Lachemi M (2001) Self-compacting concrete
the mechanical properties of concrete. J Sustain Archit Civil Eng incorporating high volumes of class F fly ash: preliminary
35(2):103–116 results. Cem Concr Res 31(3):413–420
40. Muñoz Perez SP, Garcia Chumacero JM, Charca Mamani S, Vil- 57. AbdelAleem BH, Hassan AA (2018) Development of self-con-
lena Zapata LI (2023) Influence of the secondary aluminum chip solidating rubberized concrete incorporating silica fume. Constr
on the physical and mechanical properties of concrete. Innov Build Mater 161:389–397
Infrastruct Solut 8(1):45 58. Tahwia AM, Abdelraheem AH, Taha TE (2018) Effect of silica
41. Muñoz Pérez SP, Zamora A (2023) Effects of incorporating gran- fume on mechanical and durability properties of selfcompact-
ite powder in the mechanical properties of concrete. Przegląd ing concrete. Int J Innov Res Sci Eng Technol 7(5):6027–6039.
Naukowy Inżynieria i Kształtowanie Środowiska 32:18 [Link]
42. Chicoma A, Quiroz R, Muñoz S, Villena L (2023) Influence of the 59. Garcia-Troncoso N, Li L, Cheng Q, Mo KH, Ling TC (2021)
physical and mechanical properties of concrete by adding rubber Comparative study on the properties and high temperature
powder and silica fume. J Sustain Archit Civil Eng 32(1):205–223 resistance of self-compacting concrete with various types of
43. Abdellatief M, Al-Tam SM, Elemam WE, Alanazi H, Elgendy recycled aggregates. Case Stud Constr Mater 15:e00678
GM, Tahwia AM (2023) Development of ultra-high-performance 60. Liu X, Wu T, Yang X, Wei H (2019) Properties of self-compact-
concrete with low environmental impact integrated with metakao- ing lightweight concrete reinforced with steel and polypropyl-
lin and industrial wastes. Case Stud Constr Mater 18:e01724 ene fibers. Constr Build Mater 226:388–398
44. Abdellatief M, Abd Elrahman M, Elgendy G, Bassioni G, Tahwia 61. Leung HY, Kim J, Nadeem A, Jaganathan J, Anwar MP (2016)
AM (2023) Response surface methodology-based modelling and Sorptivity of self-compacting concrete containing fly ash and
optimization of sustainable UHPC containing ultrafine fly ash and silica fume. Constr Build Mater 113:369–375
metakaolin. Constr Build Mater 388:131696 62. Ardalan RB, Joshaghani A, Hooton RD (2017) Workability
45. Asteris PG, Kolovos KG (2017) Self-compacting concrete retention and compressive strength of self-compacting concrete
strength prediction using surrogate models. Neural Comput Appl incorporating pumice powder and silica fume. Constr Build
31:409–424. [Link] Mater 134:116–122
46. Mohamed HA (2011) Effect of fly ash and silica fume on com- 63. Simões B, da Silva PR, Silva RV, Avila Y, Forero JA (2020)
pressive strength of self-compacting concrete under different cur- Ternary Mixes of Self-Compacting Concrete with Fly Ash
ing conditions. Ain Shams Eng J 2(2):79–86 and Municipal Solid Waste Incinerator Bottom Ash. Appl Sci
47. Faez A, Sayari A, Manie S (2020) Mechanical and rheological 11(1):107
properties of self-compacting concrete containing Al2O3 nan- 64. Dehwah HAF (2012) Mechanical properties of self-compacting
oparticles and silica fume. Iran J Sci Technol Trans Civil Eng concrete incorporating quarry dust powder, silica fume or fly
44:217–227. [Link] ash. Constr Build Mater 26(1):547–551
48. Behfarnia K, Farshadfar O (2013) The effects of pozzolanic bind- 65. Dinakar P, Reddy MK, Sharma M (2013) Behaviour of self
ers and polypropylene fibers on durability of SCC to magnesium compacting concrete using Portland pozzolana cement with dif-
sulfate attack. Constr Build Mater 38:64–71. [Link] ferent levels of fly ash. Mater Des 46:609–616
1016/j.​conbu​ildmat.​2012.​08.​035 66. Zeyad AM, Almalki A (2020) Influence of mixing time and
49. Gholhaki M, Kheyroddin A, Hajforoush M, Kazemi M (2017) superplasticizer dosage on self-consolidating concrete proper-
An investigation on the fresh and hardened properties of self- ties. J Market Res 9(3):6101–6115
compacting concrete incorporating magnetic water with various 67. Long WJ, Gu Y, Liao J, Xing F (2017) Sustainable design and
pozzolanic material. Constr Build Mater 158:173–180. [Link] oi.​ ecological evaluation of low binder self-compacting concrete.
org/​10.​1016/j.​conbu​ildmat.​2017.​09.​135 J Clean Prod 167:317–325
50. Hassan AAA, Lachemi M, Hossain KMA (2012) Effect of 68. Bušić R, Benšić M, Miličević I, Strukar K (2020) Prediction
metakaolin and silica fume on the durability of self-consolidating models for the mechanical properties of self-compacting con-
concrete. Cement Concr Compos 34(6):801–807. [Link] crete with recycled rubber and silica fume. Materials 13(8):1821
10.​1016/j.​cemco​ncomp.​2012.​02.​013 69. Al-Mughanam T, Aldhyani TH, Alsubari B, Al-Yaari M (2020)
51. Karatas M, Turk K, Ulucan ZC (2010) Investigation of bond Modeling of compressive strength of sustainable self-compact-
between lap-spliced steel bar and self-compacting concrete: the ing concrete incorporating treated palm oil fuel ash using arti-
role of silica fume. Can J Civ Eng 37(3):420–428. [Link] ficial neural network. Sustainability 12(22):9322
10.​1139/​l09-​159
Innovative Infrastructure Solutions (2024) 9:412 Page 31 of 31 412

70. Turk K, Karatas M, Gonen T (2013) Effect of fly ash and silica 77. Abdalla A, Mohammed AS (2022) Hybrid MARS-, MEP-, and
fume on compressive strength, sorptivity and carbonation of ANN-based prediction for modeling the compressive strength of
SCC. KSCE J Civ Eng 17:202–209 cement mortar with various sand size and clay mineral metakaolin
71. R'mili A, Ben-Ouezdou M, Added M and Ghorbel E (2009) content. Arch Civil Mech Eng 22(4):194
Prediction of the compressive strengths of self-compacting con- 78. Tahr Z, Mohammed A, Ali JA (2022) Surrogate models to predict
crete. In: Proceedings of international conference, INVACO, initial shear stress of clay bentonite drilling fluids incorporated
Hammamet, Tunisia, pp 195–204 with polymer under various temperature conditions. Arab J Geosci
72. Sabet FA, Libre NA, Shekarchi M (2013) Mechanical and 15(17):1449
durability properties of self-consolidating high performance 79. Asteris PG, Mokos VG (2020) Concrete compressive strength
concrete incorporating natural zeolite, silica fume and fly ash. using artificial neural networks. Neural Comput Appl
Constr Build Mater 44:175–184. [Link] 32(15):11807–11826
ildmat.​2013.​02.​069 80. Kooshkaki A, Eskandari-Naddaf H (2019) Effect of porosity on
73. Vivek SS, Dhinakaran G (2017) Fresh and hardened properties predicting compressive and flexural strength of cement mortar
of binary blend high strength self compacting concrete. Eng Sci containing micro and nano-silica by multi-objective ANN mod-
Technol Int J 20(3):1173–1179. [Link] eling. Constr Build Mater 212:176–191
2017.​05.​003 81. Golafshani EM, Behnood A, Arashpour M (2020) Predicting the
74. Zende A, Khadiranaikar RB (2018) Experimental investigation compressive strength of normal and high-performance concretes
of high-strength self-compacting fibre-reinforced concrete. Sus- using ANN and ANFIS hybridized with Grey Wolf optimizer.
tain Constr Build Mater. [Link] Constr Build Mater 232:117266
3317-0_​32
75. Akeed MH, Qaidi S, Ahmed HU, Emad W, Faraj RH, Mohammed Springer Nature or its licensor (e.g. a society or other partner) holds
AS, Azevedo AR (2022) Ultra-high-performance fiber-reinforced exclusive rights to this article under a publishing agreement with the
concrete. Part III: fresh and hardened properties. Case Stud Constr author(s) or other rightsholder(s); author self-archiving of the accepted
Mater 17:e01265 manuscript version of this article is solely governed by the terms of
76. Piro NS, Mohammed AS, Hamad SM (2022) The impact of GGBS such publishing agreement and applicable law.
and ferrous on the flow of electrical current and compressive
strength of concrete. Constr Build Mater 349:128639