www.rsisinternational.org
Page 3894
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
Prediction of Sustainable Concrete Strength Incorporating Fly Ash and
Recycled Aggregates Using Artificial Neural Networks
Aanand Shah
1
, Bholahari Dhungana
2
and Moulya H V
3*
1,2
Research Scholar, Department of Civil Engineering, Nitte (Deemed to be University), Nitte Meenakshi
Institute of Technology (NMIT), Bengaluru, India.
3*
Assistant Professor, Department of Civil Engineering, Nitte (Deemed to be University), Nitte
Meenakshi Institute of Technology (NMIT), Bengaluru, India.
DOI: https://doi.org/10.51583/IJLTEMAS.2026.150600289
Received: 13 July 2026; Accepted: 18 July 2026; Published: 04 August 2026
ABSTRACT
Artificial Neural Networks (ANNs) have been used to predict the compressive strength of sustainable concrete
as an alternative to large laboratory experiments․ Sustainable concrete mixes were designed by replacing natural
coarse aggregate (NCA) with recycled coarse aggregate (RCA) at varying percentages ranging from 0% to
100%․ Ordinary Portland cement was replaced with fly ash at the rate of 0%‚ 5%‚ 10%‚ 15%‚ 20%‚ 25%‚ 30%‚
35% and 40% to obtain samples․ Twenty-five concrete mixes were tested for compressive strength at 3‚ 14 and
28 days․ In contrast‚ for moderate fly ash replacement levels (10-20 percent)‚ strength increased at later ages
due to pozzolanic reactions․ At high fly ash replacement levels‚ strength development was delayed during the
curing period․ Increasing RCA content decreased compressive strength because of the presence of unhydrated
mortar‚ greater porosity‚ lower bonding between the aggregate and paste matrix‚ and greater water absorption
compared to normal aggregate․ An ANN model has been developed to predict the concrete strength‚ using five
parameters as input (NCA content‚ RCA content‚ cement content‚ fly ash content and curing age) and the
concrete compressive strength as an output parameter․ Based on the seventy-five experimental results‚ the
developed model is highly accurate‚ with R²-value of 0․9935 and RMSE (Root Mean Square Error) of 0․6465
MPa․ These results show that the ANN can successfully predict the sustainable concrete compressive strength․
The proposed model is very useful for mix proportion optimization with considerable reductions in the time‚
cost‚ and effort of experimental studies․
Keywords: Artificial Neural Network (ANN), Sustainable Concrete, Recycled Coarse Aggregate, Fly Ash and
Compressive Strength
INTRODUCTION
Concrete is the most widely used construction material in the world because of its versatility, durability, and
comparatively low production cost. However, the extensive consumption of natural resources and the large
amount of carbon dioxide emitted during cement production have raised significant environmental concerns [1]–
[3]. Ordinary Portland cement (OPC) production contributes substantially to greenhouse gas emissions, while
excessive extraction of natural aggregates leads to depletion of natural resources and ecological imbalance.
Consequently, the construction industry is increasingly focusing on sustainable alternatives capable of reducing
environmental impact while maintaining structural performance [1], [4]. Sustainable concrete, incorporating
industrial by-products and recycled materials, has therefore emerged as a promising solution for achieving
environmentally responsible infrastructure development [2], [5]. Recycled concrete aggregates generally reduce
strength because of residual mortar and higher porosity, while fly ash contributes to later-age strength and
sustainability benefits.
Fly ash is the most commonly used supplementary cementitious material because of its pozzolanic properties
and ability to partially replace cement in concrete [6][7]․ Fly ash is an industrial by-product of the combustion
of pulverized coal in thermal power plants․ Fly ash reacts with the calcium hydroxide produced during cement
www.rsisinternational.org
Page 3895
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
hydration to form calcium silicate hydrate (C-S-H) gel that improves concrete's long-term mechanical and
durability-related properties [6]․ A moderate level of fly ash replacement improves long-term compressive
strength‚ reduces permeability‚ and improves durability in aggressive environments [7]‚ [8]․ Nevertheless‚ it has
been shown that a large volume of fly ash can delay hydration‚ and thus slow the gain of early-age strength‚ due
to the slow pozzolanic reaction [6]‚ [9]․ Fly ash concrete can provide increased workability and durability‚ but
careful optimization of replacement level is required to produce a balance of early and long-term strength [7]‚
[10]․ Prediction of the strength of fly ash concrete using artificial neural network (ANN) technologies has been
documented in Scopus-indexed literature․
Recycling and reuse of construction and demolition waste as recycled coarse aggregates (RCA)‚ used as an
alternative to natural coarse aggregates for concrete production‚ is another practice with meaningful potential to
reduce the environmental impact and conserve natural resources [4]‚ [11]‚ with the aggregates themselves
generated via crushed concrete waste․ The use of RCA aids in reducing the use of landfill sites‚ virgin aggregate
extraction‚ and promotes the circular economy concept in the construction sector [4] [12]․ However‚ due to the
presence of adhered old mortar‚ RCA produces higher water absorption‚ lower density‚ and poorer interfacial
transition zones which reduce compressive strength‚ tensile strength‚ and elasticity [11] [13]․ While the trend is
clear that increasing the amount of recycled aggregate will decrease compressive strength‚ reasonable results
have been obtained․ The exact amount of recycled aggregate is subject to control [11]‚ [14]․ Studies involving
the addition of fly ash alongside RCA indicate that it can offset the negative effects of RCA and improve long-
term performance․
Concrete compressive strength is the most important design parameter for structural design and quality control
[15]․ It determines the serviceability‚ limit states‚ load-carrying capacity‚ and durability of concrete structures․
However‚ a laboratory test for compressive strength requires a meaningful amount of time‚ expense‚ and
concrete to be conducted․ Various factors affect concrete properties‚ such as the cement content‚ aggregate
properties‚ supplementary cementitious materials‚ water-to-binder ratio‚ curing conditions‚ and age․ Due to
these factors‚ it is complex and expensive to determine the compressive strength experimentally․ Reliable
computational models have been developed to accurately predict the concrete properties and reduce the amount
of experimental work required․
More recently‚ the field of civil engineering has seen a great increase in the use of AI and machine learning
techniques in modeling and predicting material behavior [17]‚ [18]․ Artificial neural networks (ANNs)‚
computational models based on the structure and functionality of the human brain‚ have been widely used in
nonlinear engineering applications [17]․ As smart systems‚ ANNs have the ability to extract relationships
between multiple input variables․ They can predict concrete properties bounded by multiple interacting variables
[18]․ The advantage of ANNs over regression-based models is that there is no need to formulate a relationship
between input variables and output response‚ as ANNs learn from experimental data obtained from the training
process [17]‚ [19]․
Developed ANN models are also used for predicting the compressive strength‚ workability‚ durability‚ and other
engineering properties of concrete [18]‚ [20]․ Another application is predicting compressive strength of self-
compacting concrete and high-performance concrete using high-volume fly ash as a partial replacement for
cement․ The study by Prasad et al․ suggested a good agreement between predicted values and experimental
results [18]․ Or researchers developed ANN models for recycled aggregate concrete with accurate predictions
of compressive strength based on mix and curing parameters [19]․ Similar ANN models were developed for
concrete containing construction and demolition waste and industrial by-products including supplementary
cementitious materials․ The resulting models had high correlations and low prediction errors [17]‚ [20]․ ANN
techniques are well suited to predict the compressive strength of recycled aggregate concrete in large datasets․
While many researchers have studied concrete sustainability‚ there are few studies that utilize ANN to predict
concrete with fly ash and RCA over the course of various curing ages․ This is because the effect of fly ash
replacement and RCA contents on strength development is complex and difficult to represent with a linear
model․ Other prediction models have been developed using artificial neural networks‚ which combine multiple
input parameters to predict nonlinear trends in concrete performance [17]‚ [20]․ In this research‚ a sustainable
concrete was prepared by partially substituting natural coarse aggregates and ordinary Portland cement with
www.rsisinternational.org
Page 3896
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
recycled coarse aggregate (RCA) and fly ash‚ respectively․ Twenty-five concrete mixtures were prepared by
varying the percentage of RCA from 0% to 100% and fly ash from 0% to 40%․ Furthermore‚ the compressive
strength of the specimens was determined at the curing ages of 3‚ 14 and 28 days․ An ANN model was then
developed with five input parameters: percentage of natural coarse aggregate‚ percentage of recycled coarse
aggregate‚ percentage of cement‚ percentage of fly ash‚ and curing age‚ and compressive strength as the output
parameter․ The developed model is expected to predict the performance of sustainable concrete more rapidly‚
accurately and economically‚ helping to reduce experimental work‚ material usage‚ construction waste and
ultimately contribute to sustainable engineering․
Materials And Methods
Materials Used
The materials used for the present study are Ordinary Portland Cement (OPC)‚ fly ash‚ fine aggregate (sand)‚
natural coarse aggregate‚ recycled coarse aggregate and potable water․ For the present study Ordinary Portland
Cement is used as a binder as per the specification recommended by IS codes․ Fly ash has been successfully
used as a portion of the cement component of concrete to make concrete more sustainable by reducing the amount
of cement and the related carbon footprint of concrete․ Well graded fine aggregate was used to improve
workability and packing․ Natural coarse aggregate was used as control material․ The sustainable material was
recycled aggregate made from construction and demolition waste; the water used was potable water free from
any impurities‚ used for mixing and curing purpose․
However‚ adhered mortar from hardened concrete to recycled coarse aggregate causes increases in the absorption
and decreases in the density of the recycled coarse aggregate particles․ Fly ash may improve the packing density
of the particles‚ and strength of the concrete‚ via pozzolanic action․ The physical and chemical properties of all
materials were determined from standard laboratory tests prior to concrete preparation [22]․ Table 1 and Table 2
summarize the physical properties and the chemical compositions of the cement and the fly ash used in this
study‚ respectively․
Table 1: Physical Properties of Materials Used
Property
Cement
(OPC 53
Grade)
Fly Ash
(Class
F)
Fine Aggregate
(M-Sand)
Natural Coarse
Aggregate
(NCA)
Specific Gravity
3.15
—
2.45
2.72
Bulk Density (kg/m³)
—
—
2450
1550
(Compacted) /
1316 (Loose)
Water Absorption (%)
—
—
—
1.90
Fineness / Fineness
Modulus
15%
—
2.98
—
Standard Consistency (%)
31
—
—
—
Initial Setting Time (min)
36
—
—
—
Maximum Aggregate Size
(mm)
—
—
—
20
Aggregate Type
—
—
Manufactured
Sand
Crushed Granite
Aggregate
The physical characteristics of the cement‚ fly ash‚ fine aggregate‚ natural coarse aggregate (NCA)‚ and recycled
coarse aggregate (RCA) were determined based on the relevant Indian Standard specifications before the
preparation of the concrete mix to confirm the suitability of the materials in producing eco-friendly concrete
[23]․ The specific gravity‚ standard consistency‚ and initial setting time of the cement used in this study were
3․15‚ 31%‚ and 36 minutes‚ respectively․ The fineness modulus of the fine aggregate was 2․98․ The recycled
www.rsisinternational.org
Page 3897
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
coarse aggregate had higher water absorption than the natural coarse aggregate because the mortar remained
attached to the surface of the recycled coarse aggregate․
Table 2: Chemical Composition of Cement and Fly Ash
Material
SiO₂
(%)
Al₂O₃
(%)
Fe₂O₃
(%)
CaO
(%)
MgO
(%)
SO₃
(%)
LOI
(%)
Cement (OPC 53 Grade)
20.8
5.4
3.6
63.5
2.5
2.3
1.8
Fly Ash (Class F)
55
25
7
6
2
—
3
The chemical composition of Ordinary Portland Cement (OPC) and fly ash used in the present investigation is
presented in Table 2. OPC primarily consisted of calcium oxide (CaO), silica (SiO₂), alumina (Al₂O₃), and iron
oxide (Fe₂O₃), which contribute to cement hydration and strength development. The fly ash was classified as
Class F fly ash and contained higher silica and alumina contents, indicating strong pozzolanic characteristics.
The presence of reactive silica and alumina in fly ash contributes to secondary hydration reactions and long-term
strength enhancement in concrete.
Mix Design and Material Proportioning
Based on the experimental study‚ concrete mixes of M30 grade were prepared and the role of fly ash and recycled
concrete aggregate (RCA) on compressive strength of green concrete was investigated by replacing 0%‚ 10%‚
20%‚ 30%‚ and 40% of ordinary Portland cement (OPC) with fly ash and 0%‚ 25%‚ 50%‚ 75%‚ and 100% of
natural coarse aggregate (NCA) with RCA․
The water to binder ratio of the mixes remained constant at 0․45 throughout the test period․ The base water
content of all concrete mixes was fixed at 0․629 L‚ with additional water provided for the RCA mixes to
counteract the higher water absorption of the RCA compared with the natural aggregate mix to ensure adequate
and consistent workability of the mixes [24]․ The amount of fine aggregate (sand) was kept constant․
To achieve batching convenience and uniformity of specimen preparation‚ the proportions of the constituent
materials were based on the volume of a concrete cube of size 150 mm x 150 mm x 150 mm․ The mix proportions
arrived for this experimental work are given in Table 3․
Table 3: Representative Material Quantities for One 150 mm Cube Specimen
Parameter
Range Used
Cement (kg)
0.838 – 1.397
Fly Ash (kg)
0.000 – 0.559
Base Water (L)
0.629
Extra Water (L)
0.079 – 0.170
Total Water (L)
0.708 – 0.799
M-Sand (kg)
2.665
Natural Coarse Aggregate (kg)
0.000 – 4.150
Recycled Coarse Aggregate (kg)
0.000 – 4.150
For all concrete mixtures, the total coarse aggregate content was maintained constant while varying the
proportion of natural coarse aggregate and recycled coarse aggregate according to selected replacement levels.
Similarly, fly ash was introduced by replacing an equivalent proportion of cement while maintaining total binder
content nearly constant. The adopted material proportioning enabled systematic investigation of both the
individual and combined effects of fly ash incorporation and recycled aggregate utilization on sustainable
concrete performance.
Specimen Preparation and Testing
The concrete raw materials were weighed for the different compositions and mixed to obtain a homogeneous
mixture․ The dry mixed materials consisting of Ordinary Portland Cement (OPC)‚ fly ash‚ manufactured sand
www.rsisinternational.org
Page 3898
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
(M-sand)‚ natural coarse aggregate (NCA)‚ and recycled coarse aggregate (RCA) were first mixed dry until a
homogeneous mixture was obtained․ Water was carefully added and mixed to produce a homogenous and
workable concrete mix․ Due to the high water absorption capacity of the RCA‚ owing to mortar residue
remaining on the surface‚ water was added to each mix according to the percentage of RCA replacement in order
to ensure the same workability for each matrix [25]․ The concrete was mixed until a homogeneous material was
obtained‚ with no visible segregation․
The freshly prepared concrete was immediately subjected to the slump cone test for assessment of workability․
For testing concrete for compressive strength‚ it was filled into steel cube moulds of 150 mm × 150 mm × 150
mm dimensions and poured in layers compacted by a poker to disperse air bubbles and achieve full compaction․
Specimens were cast in the laboratory‚ demoulded after approximately 24 h and cured in water until 3‚ 14 and
28 days of testing․ The complete sequence of specimen preparation and casting are shown in Fig․ 1(a)-(c)․
Fig 1(a) Mould Preparation
Fig 1(b): Casting
Fig 1(c): Specimen and Curing
Artificial Neural Network (ANN) Modeling
The Artificial Neural Network (ANN) model is proposed to predict the sustainable concrete compressive strength
made of fly ash and RCA․ ANN is a computational technique‚ which can effectively model and predict the
complex nonlinear relationships amongst input variables and target variables for estimating the mechanical
behavior of concrete materials [26]‚ [27]․ The measured values of compressive strength were used to train and
validate the prediction model developed in this study․ The input variables to the model were the natural coarse
aggregate (NCA) content‚ the recycled coarse aggregate (RCA) content‚ the cement content‚ the fly ash content
and the curing age․ The output variable was the compressive strength․ The database comprised 75 experimental
observations corresponding to 25 concrete mix combinations tested at 3‚ 14 and 28 days of curing․
A feed-forward backpropagation neural network was chosen for the prediction model‚ as it is composed of an
input layer‚ one or more hidden layer(s)‚ and one output representing the compressive strength․ The number of
hidden neurons was determined by trial-and-error to be the minimum to achieve maximum prediction accuracy
and minimum model error․ To reduce prediction time and the effect of several orders of magnitude‚ the data
inputs were normalized before training [28]․
The dataset was randomly divided into training‚ validation‚ and test sets in order to allow learning‚ validation
of generalization‚ and prevention of overfitting․ Then‚ the model was trained until the prediction error became
constant and the predicted and measured compressive strength values exhibited an acceptable level of agreement․
Coefficient of determination (R²) and root mean square error (RMSE) were used to check the performance of
the model in order to understand the accuracy and reliability of developed ANN model․
www.rsisinternational.org
Page 3899
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
The complete ANN model was developed using the MATLAB Neural Network Toolbox․ The predicted
compressive strength values were compared with the experimental compressive strength values to validate the
performance of the developed prediction model․
Fig. 2. Workflow of Artificial Neural Network (ANN) modeling adopted for compressive strength
RESULTS AND DISCUSSION
Workability
Figure 3 shows slump values for all mixtures․ It can be seen that both the fly ash content and the percentage of
replacement with recycled coarse aggregate (RCA) had a meaningful effect on the workability of fresh concrete
mixtures․ With the increase in RCA content‚ slump was found to decrease gradually․ The control concrete mix
(M1) had maximum slump of 88 mm while the minimum slump was observed in 100% RCA concrete mix (M5)
with slump value of 66 mm․
Decreased workability of recycled coarse aggregate concrete is mainly associated with the recycled aggregate
properties (the amount of attached mortar)․ Recycled aggregate typically has a considerably higher water
absorption and rougher surface texture compared to natural aggregates․ This causes an increased amount of
mixing water to be sorbed by the RCA‚ thus reducing the amount of mixing water available for lubricating and
plasticizing the concrete matrix․ Likewise‚ the high surface roughness of the RCA increases interparticle friction‚
which in turn causes a reduction in slump values with an increase in replacement level․
www.rsisinternational.org
Page 3900
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
Fig 3: Variation of slump value for different concrete mixes
In contrast, the incorporation of fly ash improved workability characteristics. Mixes containing higher fly ash
percentages exhibited comparatively higher slump values at corresponding RCA replacement levels. The
maximum slump value of 96 mm was recorded for mix M21 containing 40% fly ash and 0% RCA. The
improvement in workability is associated with the spherical particle morphology and finer particle size of fly
ash, which enhanced particle packing and reduced interparticle friction.
Although RCA incorporation reduced slump values, all concrete mixes exhibited medium workability suitable
for casting and compaction operations. The results demonstrate that fly ash effectively compensated for the
adverse effects of RCA on fresh concrete behavior.
COMPRESSIVE STRENGTH
All the concrete mixes were subjected to compressive strength tests to investigate the effects of fly ash
replacement level‚ recycled coarse aggregate (RCA) content and curing period as independent variables on the
compressive strength performance of the concrete mixes․ The 3-‚ 14- and 28-day results were summarized in a
bar chart and a line chart (shown in figure 4)‚ to help with comparing the material results for different curing
ages․ The results showed that the compressive strength increased continuously as the curing age of all concrete
mixes increased due to continuous cement hydration and the pozzolanic reaction between fly ash and cement
which was delayed for a while․ However‚ an increase in the content of RCA decreased the compressive strength․
The highest 28-day compressive strength was achieved for the M11 mix‚ which used a combination of fly ash
and RCA in the concrete mix‚ showing its effectiveness․ The M25 mix had the lowest 28-day compressive
strength due to the combination of the highest fly ash replacement and the use of 100% recycled coarse
aggregate‚ which reduced the mechanical properties of the concrete mixes․
Fig 4 : Compressive Strength Results at 3, 14, and 28 Days
www.rsisinternational.org
Page 3901
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
Figure 5 presents the relationship between the compressive strength at 28 days and the amount of recycled coarse
aggregate (RCA) where the compressive strength decreases as the RCA percentage of substitution increases․
This reduction can be attributed to factors such as the increased porosity and water absorption of the recycled
aggregate and the weaker interfacial transition zone of the aggregate-cement paste․ In spite of this‚ moderate use
of RCA in concrete has reached compressive strengths adequate for many sustainable concrete applications․
Fig 5 Variation of 28-Day Strength with RCA Content
Figure 6 also shows that 10-20% fly ash replacement has been seen to increase the 28-day compressive strength
due to further pozzolanic reaction and refinement of the pore structure (i․e․ lower porosity) of the concrete․ In
contrast‚ higher replacement levels of fly ash (>30%) showed a decrease in compressive strength‚ potentially
due to the reduction in cement content and the slower rate of hydration․
Figure 7 shows that the compressive strength continued to increase from 3 to 28 days of curing for all concrete
mixes‚ indicating that the reaction of the cementitious materials continued over this period․ The strength gain is
more pronounced during the first 14 days and less so from 14 days to 28 days as the processes of hydration are
nearly complete․
Fig 6: Effect of Fly Ash on Compressive Strength
Fig 7 Strength Development with Curing Age
www.rsisinternational.org
Page 3902
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
ANN Performance Analysis
ANN Training Performance
The ANN model was developed in MATLAB R2026a using a feed-forward backpropagation architecture trained
with the Levenberg–Marquardt (trainlm) algorithm for predicting compressive strength of sustainable concrete
containing RCA and fly ash. The dataset was randomly divided into training, validation, and testing subsets using
dividerand, and MSE was used as the performance metric. As shown in Fig. 8, the training performance curve
indicates a rapid reduction in MSE during initial epochs, followed by gradual stabilization[29]-[30]. The best
validation performance of 0.0076424 at epoch 18 (Fig. 8 indicates optimal generalization of the network. Beyond
this point, validation error slightly increased while training error continued to decrease, indicating early signs of
overfitting. MATLAB automatically terminated training using early stopping, ensuring optimum model selection.
Fig 8: Performance Plot
Fig 9: Training State / Epoch Graph
The training state plot (Fig. 5.7) shows stable convergence within 24 epochs with an elapsed time of ~7 seconds.
The performance value reduced from 15.3 to 0.000358, indicating significant learning improvement. The
gradient decreased from 24.6 to 0.0345, confirming stable convergence of weight updates. The Mu parameter
reduced from 0.001 to 0.0001, indicating controlled adjustment during optimization. Validation checks reached
the limit of 6, triggering automatic stopping to avoid overfitting. Overall, Figs. 8 and 9 confirm stable learning
behavior, fast convergence, and good generalization capability of the ANN model.
Regression Analysis
The regression performance of the ANN model is presented in Fig. 10, which compares experimental and
predicted compressive strength values for training, validation, testing, and overall datasets. The obtained
correlation coefficients are: Training: R = 1.0000 , Validation: R = 0.99995 , Testing: R = 0.99999 ,Overall: R =
0.99999 The data points in Fig. 10 are closely aligned along the 45° line (y = x), indicating near-perfect
agreement between predicted and experimental values. The very high R-values confirm that the ANN model has
successfully captured the nonlinear interaction between input variables (fly ash %, RCA content, and curing age)
and compressive strength output[31].
www.rsisinternational.org
Page 3903
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
Fig 10: Regression Plot
Error Analysis
The error distribution is shown in Fig. 11, where residuals are highly concentrated around zero. The histogram
exhibits a narrow spread with near-symmetric distribution, indicating minimal bias in prediction. The majority
of errors fall within a small range, confirming high prediction stability. Minor deviations observed at higher error
values can be attributed to material heterogeneity in recycled aggregates and variability in fly ash reactivity.
Overall, Fig. 11 confirms strong prediction consistency and low model uncertainty.
Fig 11: Error Histogram
www.rsisinternational.org
Page 3904
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
Comparison Between Experimental and ANN Predicted Results
The comparison between experimental and ANN-predicted compressive strength values for all mixes (M1–M25)
at 3, 14, and 28 days is presented through graphical illustrations. The predicted values closely follow the
experimental results across all curing ages, indicating strong predictive capability of the developed ANN model.
The deviation between experimental and predicted values is minimal, confirming high model accuracy and
robustness.
Fig 12 : Parity plot for 28-day compressive strength
The ANN model successfully captured the influence of mix constituents such as recycled coarse aggregate (RCA)
and fly ash. The results indicate that increasing RCA content leads to a reduction in compressive strength due to
the weaker interfacial transition zone and higher porosity. In contrast, moderate replacement of fly ash enhances
later-age strength due to pozzolanic reactions and additional calcium silicate hydrate (C–S–H) formation. The
parity plot for 28-day compressive strength clearly illustrates a strong correlation between experimental and
ANN-predicted values. The data points are closely distributed along the ideal 45° line (y = x), indicating excellent
prediction accuracy and minimal error dispersion. This confirms the reliability of the ANN model in predicting
the compressive strength of sustainable concrete mixes.
Fig13: Subplot comparison of experimental and ANN predicted results
www.rsisinternational.org
Page 3905
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
The subplot comparison presents the variation of compressive strength at 3, 14, and 28 days for all mixes. In all
cases, the ANN-predicted values closely follow the experimental trends. The overlappingnature of the curves
demonstrates that the model effectively captures the curing-age-dependent strength development and mix
composition effects.
Figure14: Error distribution plot
The error plot represents the difference between ANN-predicted and experimental compressive strength values
for 3-day, 14-day, and 28-day curing periods. The errors are distributed around zero with minimal fluctuation,
indicating the absence of systematic bias. The low magnitude of errors confirms the high generalization
capability and stability of the ANN model.
Summary
The ANN model exhibited strong convergence behavior, high regression accuracy, and minimal prediction error.
Figures 5.10 to 5.12 collectively confirm stable learning performance, excellent correlation, and reliable
predictive capability. The model effectively captures nonlinear relationships between input variables (cement,
RCA, fly ash, and curing age) and compressive strength output.
Overall, the developed ANN model is highly suitable for predicting and optimizing sustainable concrete mixes,
significantly reducing the need for extensive experimental trials and saving both time and cost in material design.
CONCLUSIONS
The present study investigated the compressive strength behavior of sustainable concrete incorporating fly ash
and recycled coarse aggregate (RCA), along with the development of an Artificial Neural Network (ANN)-based
prediction model. Based on the experimental investigation and analytical results, the following conclusions are
drawn:
The incorporation of recycled coarse aggregate resulted in a gradual reduction in compressive strength
due to the presence of adhered mortar, higher porosity, and weaker interfacial transition zones compared
to natural aggregates.
Partial replacement of cement with fly ash up to an optimum range (10–20%) improved later-age
compressive strength due to enhanced pozzolanic reactions and improved microstructural densification.
However, higher replacement levels beyond the optimum reduced strength because of reduced cement
content and slower hydration.
www.rsisinternational.org
Page 3906
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
All concrete mixes exhibited continuous strength development with curing age, with significant strength
gain observed between 3 and 14 days, followed by a gradual increase up to 28 days.
The combined use of fly ash and recycled aggregates demonstrated that sustainable concrete with
acceptable mechanical performance can be achieved when proper replacement levels are selected.
The ANN model developed using five input parameters (cement, fly ash, NCA, RCA, and curing age)
accurately predicted compressive strength with high reliability. The model achieved a coefficient of
determination (R²) of 0.9935 and a low root mean square error (RMSE) of 0.6465 MPa, indicating
excellent agreement between experimental and predicted results.
The study confirms that ANN-based modeling is an effective and efficient tool for predicting compressive
strength of sustainable concrete, significantly reducing the need for extensive experimental trials and
supporting optimized mix design.
REFERENCES
1. F. Khalid, A.-u.-R. Khan, and S. Fareed, “Estimation of parameters for constitutive modelling of recycled
aggregate concrete using artificial neural networks,” Engineering Computations, vol. 42, no. 4, pp. 1471–
1501, Apr. 2025. DOI: https://doi.org/10.1108/EC-07-2024-0679
2. I.-C. Yeh, “Modeling of strength of high-performance concrete using artificial neural networks,” Cement
and Concrete Research, vol. 28, no. 12, pp. 1797–1808, 1998. DOI: https://doi.org/10.1016/S0008-
8846(98)00165-3
3. I.-C. Yeh, “Analysis of strength of concrete using design of experiments and neural networks,” Journal of
Materials in Civil Engineering, vol. 18, no. 4, pp. 597–604, 2006. DOI:
https://doi.org/10.1061/(ASCE)0899-1561(2006)18:4(597)
4. B. K. Prasad, H. Eskandari, and B. V. V. Reddy, “Prediction of compressive strength of SCC and HPC with
high volume fly ash using ANN,” Construction and Building Materials, vol. 23, no. 1, pp. 117–128, 2009.
DOI: https://doi.org/10.1016/j.conbuildmat.2008.01.014
5. H. Naderpour, A. H. Rafiean, and P. Fakharian, “Compressive strength prediction of environmentally
friendly concrete using artificial neural networks,” Journal of Building Engineering, vol. 16, pp. 213–219,
2018. DOI: https://doi.org/10.1016/j.jobe.2018.01.007
6. A. Behnood and E. Golafshani, “Predicting the compressive strength of silica fume concrete using hybrid
artificial neural network with multi-objective grey wolves,” Journal of Cleaner Production, vol. 202, pp.
54–64, 2018. DOI: https://doi.org/10.1016/j.jclepro.2018.08.065
7. Y. Deng, T. Shi, and Y. Zhang, “Prediction of compressive strength of recycled aggregate concrete using
artificial neural networks,” Construction and Building Materials, vol. 40, pp. 1200–1206, 2013. DOI:
https://doi.org/10.1016/j.conbuildmat.2012.11.042
8. J. Abellan-Garcia, J. Fernández-Gómez, M. Iqbal Khan, Y. M. Abbas, and C. Pacheco-Bustos, “ANN
approach to evaluate the effects of supplementary cementitious materials on the compressive strength of
recycled aggregate concrete,” Construction and Building Materials, vol. 402, p. 132992, Oct. 2023. DOI:
https://doi.org/10.1016/j.conbuildmat.2023.132992
9. M. J. Moradi, M. Khaleghi, J. Salimi, V. Farhangi, and A. M. Ramezanianpour, “Predicting the
compressive strength of concrete containing metakaolin with different properties using ANN,”
Measurement, vol. 183, p. 109790, 2021. DOI: https://doi.org/10.1016/j.measurement.2021.109790
10. W. Liu, H. Mo, and C. K. Lee, “Compressive strength prediction for SCM-based concretes using multiple
machine learning models and datasets approach,” Results in Engineering, vol. 30, p. 110262, 2026. DOI:
https://doi.org/10.1016/j.rineng.2026.110262
11. G. Pazouki, N. Saeed, Z. Tao, and W.-H. Kang, “A predictive ML framework for 28-day compressive
strength using early-age–driven models for concrete with sustainable supplementary cementitious
materials,” Materials Today Communications, vol. 52, p. 115052, 2026. DOI:
https://doi.org/10.1016/j.mtcomm.2026.115052
www.rsisinternational.org
Page 3907
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
12. J. Xiao, J. Li, and C. Zhang, “Mechanical properties of recycled aggregate concrete under uniaxial loading,”
Cement and Concrete Research, vol. 35, no. 6, pp. 1187–1194, 2005. DOI:
https://doi.org/10.1016/j.cemconres.2004.09.020
13. S. C. Kou and C. S. Poon, “Properties of self-compacting concrete prepared with coarse and fine recycled
concrete aggregates,” Cement and Concrete Composites, vol. 31, no. 9, pp. 622–627, 2009. DOI:
https://doi.org/10.1016/j.cemconcomp.2009.06.005
14. Y. Limbachiya, T. Leelawat, and R. K. Dhir, “Use of recycled concrete aggregate in high-strength concrete,”
Materials and Structures, vol. 33, no. 9, pp. 574–580, 2000. DOI: https://doi.org/10.1007/BF02480538
15. V. Corinaldesi and G. Moriconi, “Behavior of cementitious mortars containing different kinds of recycled
aggregate,” Construction and Building Materials, vol. 23, no. 1, pp. 289–294, 2009. DOI:
https://doi.org/10.1016/j.conbuildmat.2007.12.006
16. V. W. Y. Tam, X. F. Gao, and C. M. Tam, “Microstructural analysis of recycled aggregate concrete produced
from two-stage mixing approach,” Cement and Concrete Research, vol. 35, no. 6, pp. 1195–1203, 2005.
DOI: https://doi.org/10.1016/j.cemconres.2004.10.025
17. P. Dinakar, K. G. Babu, and M. Santhanam, “Durability properties of high-volume fly ash self-compacting
concretes,” Cement and Concrete Composites, vol. 30, no. 10, pp. 880–886, 2008. DOI:
https://doi.org/10.1016/j.cemconcomp.2008.08.005
18. P. Chindaprasirt, C. Jaturapitakkul, and T. Sinsiri, “Effect of fly ash fineness on compressive strength and
pore size of blended cement paste,” Cement and Concrete Composites, vol. 27, no. 4, pp. 425–428, 2005.
DOI: https://doi.org/10.1016/j.cemconcomp.2004.07.003
19. E. Güneyisi, M. Gesoğlu, and H. Özbay, “Strength and drying shrinkage properties of self-compacting
concretes incorporating multi-system blended mineral admixtures,” Construction and Building Materials,
vol. 24, no. 10, pp. 1878–1887, 2010. DOI: https://doi.org/10.1016/j.conbuildmat.2010.04.015
20. G. Xu and X. Shi, “Characteristics and applications of fly ash as a sustainable construction material: A
state-of-the-art review,” Resources, Conservation and Recycling, vol. 136, pp. 95–109, 2018. DOI:
https://doi.org/10.1016/j.resconrec.2018.04.010
21. J. Miklečić and V. Jirouš-Rajković, “Effectiveness of finishes in protecting wood from liquid water and
water vapor,” Journal of Building Engineering, vol. 43, p. 102621, Nov. 2021. DOI:
https://doi.org/10.1016/j.jobe.2021.102621
22. N. Cristelo, V. M. C. F. Cunha, A. Topa Gomes, N. Araújo, T. Miranda, and M. de Lurdes Lopes, “Influence
of fibre reinforcement on the post-cracking behaviour of a cement-stabilised sandy-clay subjected to
indirect tensile stress,” Construction and Building Materials, vol. 138, pp. 163–173, May 2017. DOI:
https://doi.org/10.1016/j.conbuildmat.2017.02.010
23. S. X. Ruan, J. Zhang, and X. Wang, “Predicting failure process of precast deck joints using physics-guided
LSTM model,” Structures, vol. 59, p. 105732, Jan. 2024. DOI:
https://doi.org/10.1016/j.istruc.2023.105732
24. J. Wang, L. He, X. Wu, and L. Zhang, “Influence of nanoparticles on age-associated mechanical properties
of cement asphalt mortar,” Construction and Building Materials, vol. 301, p. 124033, Sep. 2021. DOI:
https://doi.org/10.1016/j.conbuildmat.2021.124033
25. J. N. Akhtar, R. A. Khan, R. A. Khan, M. N. Akhtar, and B. S. Thomas, “A comparative study of strength
and durability characteristics of concrete and mortar admixture by bacterial calcite precipitation: A review,”
Materials Today: Proceedings, 2023. DOI: https://doi.org/10.1016/j.matpr.2023.03.490
26. S. Spadea, I. Farina, A. Carrafiello, and F. Fraternali, “Recycled nylon fibers as cement mortar
reinforcement,” Construction and Building Materials, vol. 80, pp. 200–209, Apr. 2015. DOI:
https://doi.org/10.1016/j.conbuildmat.2015.01.075
27. M. Nikoo, F. Torabian Moghadam, and Ł. Sadowski, “Prediction of concrete compressive strength by
evolutionary artificial neural networks,” Advances in Materials Science and Engineering, vol. 2015,
Article ID 849126, 2015. DOI: https://doi.org/10.1155/2015/849126
28. M. G. Asteris and K. G. Kolovos, “Self-compacting concrete strength prediction using surrogate models,”
Neural Computing and Applications, vol. 31, pp. 409–424, 2019. DOI: https://doi.org/10.1007/s00521-
017-3007-7
29. A. Mardani-Aghabaglou, M. Tuyan, and K. Ramyar, “Mechanical and durability performance of concrete
incorporating fine recycled concrete aggregates,” Construction and Building Materials, vol. 64, pp. 477–
488, 2014. DOI: https://doi.org/10.1016/j.conbuildmat.2014.04.017
www.rsisinternational.org
Page 3908
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue VI, June 2026
30. C. S. Poon, Z. H. Shui, and L. Lam, “Compressive behavior of fiber reinforced high-performance concrete
subjected to elevated temperatures,” Cement and Concrete Research, vol. 34, no. 12, pp. 2215–2222, 2004.
DOI: https://doi.org/10.1016/j.cemconres.2004.02.011
31. K. Ganesan, K. Rajagopal, and K. Thangavel, “Evaluation of bagasse ash as supplementary cementitious
material,” Cement and Concrete Composites, vol. 29, no. 6, pp. 515–524, 2007. DOI:
https://doi.org/10.1016/j.cemconcomp.2007.03.001