ARID ZONE JOURNAL OF ENGINEERING, TECHNOLOGY & ENVIRONMENT
AZOJETE, September 2019. Vol. 15(3) 692-701

Published by the Faculty of Engineering, University of Maiduguri, Maiduguri, Nigeria.
Print ISSN: 1596-2490, Electronic ISSN: 2545-5818

www.azojete.com.ng

692

ORIGINAL RESEARCH ARTICLE

PREDICTION MODELING OF 28-DAY CONCRETE COMPRESSIVE STRENGTH USING
ARTIFICIAL NEUTRAL NETWORK

O.A.U. Uche1 , M.T. Abdulwahab2*, A. Suleiman3 and Y. Ismail4
(1,4Department of Civil Engineering, Bayero University Kano, Kano, Nigeria

2Department of Civil Engineering, Kebbi State University of Science and Technology, Kebbi, Nigeria
3Department of Civil Engineering, Abubakar Tafawa Balewa University, Bauchi, Nigeria)

* Corresponding author’s email address: amtaytechnology@yahoo.com

ARTICLE INFORMATION

Submitted 22 November, 2018
Revised 7 March, 2019
Accepted 10 March, 2019

Keywords:
Prediction
compressive strength
Concrete
artificial neural network

ABSTRACT
Compressive strength of concrete at the age of 28 days is an important
parameter for the design of concrete structures and waiting for that
length of time to obtain the value can be tasky. This study developed
an alternative approach using Artificial Neutral Network (ANN) to
estimate or predict the compressive strength of concrete at 28th day
from early age results. In the study concrete cubes of mix ratio 1:2:4
were cast with different water-cement ratios (0.4, 0.5, 0.6 and 0.65) and
their seventh (7th) and twenty-eighth (28th) day strength were
measured in the laboratory. In all, 400 cubes of 150 x 150 x 150mm of
200 sets were subjected to compressive strength test using Avery
Denison Universal Testing Machine of 2000 kN load capacity at a
constant load application of 15kN/s. ANN model was then developed
using the time series tool of ANN in MATLAB 7.12.0 (R2011a) applying
back propagation algorithm. Out of the 200 sets of results, 110 sets
(55%) were used for the training of the network while 30 sets (15%)
were used to validate and 60 sets (30%) to test the network. The result
of the crushing test shows that the higher the compressive strength at
seventh (7th) day the higher it will be at twenty-eighth (28th) day. The
result of the ANN model shows a good correlation between the
seventh (7th) day compressive strength and the twenty-eighth (28th)
day compressive strength with training and validation correlation
coefficients of 0.99751 and 0.99736 respectively. It was also found that
the ANN model ��u � 쳌䁐�䀀㌳㌳�� � Ͷ䁐䀀Ͷⷸ� is quite efficient in
determining the twenty-eighth (28th) day compressive strength of
concreteas the predicted strength values match very well with those
obtained experimentally with a correlation coefficient of 0.99675.

© 2019 Faculty of Engineering, University of Maiduguri, Nigeria. All rights reserved.

1.0 Introduction
Concrete is one of the most widely used building materials in the World because of its versatile
application, satisfying performance in strength requirements and its ability to be moulded into a
variety of shapes and sizes (Kayode and Ilessama, 2015). It is a synthetic construction material
made by mixing of cement, fine aggregates, coarse aggregate and water in the correct
proportion, and each of these components contribute to the strength of the concrete as
reported by Gambhir (2004). The fine aggregates fill up the voids formed by the coarse

mailto:amtaytechnology@yahoo.com
http://www.azojete.com.ng


Arid Zone Journal of Engineering, Technology and Environment, September, 2019; Vol. 15(3):692-701. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com

693

aggregates; and cement fills up the voids of the fine aggregates as the lesser the voids the more
the strength of concrete (Mehta and Monteiro, 2006).

The strength of concrete comprises mainly of the compressive, split tensile and flexural. In the
construction industry, strength is a primary criterion in selecting concrete for a particular
application. Concrete used for construction gains a significant component of its strength during
the initial 3 to 4 weeks, but continues to do so over a long period of time after pouring (Chopra
et al., 2015). The characteristic strength of concrete is defined as the compressive strength of a
sample that has been cured for 28 days.

The use of concrete for mass concreting in huge civil projects like bridges, dams, power plants,
are usually performed in several layers and the concrete strength of each layer must be achieved.
Therefore, one has to wait for 28 days to achieve a 28-day strength for each layer of the
concrete. This is time consuming and costly to perform as days will be needed to obtain
the 28-day strength for � layers. However, to hasten the construction progress we must be able
to predict the concrete strength based upon the early strength data. Therefore, rapid and
reliable prediction for the strength of concrete would be of great significance.
The prediction of concrete strength and behaviour, therefore, has been an active area of
research and a considerable number of studies have been carried out in it (Bamiyo et al., 2016;
Zain and Abd, 2009; Zelic et al., 2004; Akkurt et al., 2004; Hwang et al., 2004; Kheder et al., 2003;
Zain et al., 2002). It is well recognized that the prediction of concrete strength is important in the
modernized concrete construction and engineering judgment (Dias and Pooliyadda, 2001).
Thus, for the sake of saving time, decreasing the cost of production of concrete and wastage of
materials, the help of Artificial Neural Network (ANN) is taken to develop models, so that the
knowledge extracted from these models, can be utilized to predict the strength of concrete.

The Artificial Neural Network (ANN) is a form of Artificial Intelligence (AI) that attempts to mimic,
in a very simplistic way, the human cognition capability to solve engineering problems that have
defied solution using conventional computational techniques (Flood, 2008). A typical structure of
ANN consists of a number of processing elements (PEs), or nodes, that are usually arranged in
layers: an input layer, an output layer and one or more hidden layers as shown in Figure 1. The
ANN modelling philosophy is similar to a number of conventional statistical models in the sense
that both are attempting to capture the relationship between a historical set of model inputs
and corresponding outputs (Shain et al., 2001).

Figure 1: Typical Structure of ANN (Chopra et al., 2015)

Unlike conventional problem solving algorithms that are basically based upon statistical analysis
by which many linear and nonlinear regression equations have been constructed to model such
a prediction problem (Monjurul and Ahsanul, 2011), Artificial Neural Network can be trained to

file:///C:/Users/user/Downloads/azojete143/www.azojete.com.ng


Uche, et al: Prediction modeling of 28-day concrete compressive strength using artificial neutral network. AZOJETE,
15(3):692-701. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com 694

perform a particular task. This is done by presenting the system with a representative set of
examples describing the problem, namely pairs of input and output samples. The ANN will then
extrapolate the mapping between input and output data. After training, the ANN can be used to
recognize data that is similar to any of the examples shown during the training phase (Rafiq,
2001). The ANN can even recognize incomplete or noisy data and also have the ability to self-
organize, therefore enabling segmentation or coarse coding of data. This is an important
feature that is often used for prediction, diagnosis or control purposes.

In general, ANN have the ability to simulate the behaviour of systems with limited modelling
effort and provide speedy and reasonably accurate solutions in complex uncertainties and
subjective situations. Thus, this study aims at investigating the feasibility of predicting the 28th
day compressive strength of concrete with the aid of ANN using its strength obtained at the 7th
day.

2. Materials and Methods
The materials used for this study were sourced within Kano State, Nigeria and the experimental
procedures to arrive at the desired results were carefully followed. Also, the specification and
quality of the test materials were kept same throughout the test

2.1. Materials
The binder used in this study was Ordinary Portland Cement (OPC), 3X brand manufactured by
Dangote Cement Nigeria Plc. The cement was of grade 42.5 with a specific gravity of 3.14 as
determined in accordance with BS 812 (1985). Naturally occurring clean sand, free of
adulteration, obtained from river Challawa in Kano State was used as the fine aggregates. The
specific gravity of the fine aggregate was found to be 2.6 and it was then classified as Zone-II
after conducting sieve analysis in accordance to BS 882 (1992).

Crushed granite stones of 20mm nominal diameter with aggregates crushing value of 22% and
specific gravity of 2.7 were used as the coarse aggregates. It was obtained from a standard
quarry site of Abduljalil Hajaig and Sons Ltd in Kano State. The water used for mixing and curing
was portable tap water from the Civil Engineering Laboratory of Bayero University Kano and
satisfied ASTM C1602-12 specification of water for use in concrete mixtures. Artificial Neural
Network toolbox in MATLAB (R2011a) package was used in developing the network architecture
for the predictions in this study. The input variables for the proposed neural network consisted
of water cement ratio (w/c), cement content, fine aggregate content, coarse aggregate content
and 7th day compressive strength, while the output variable was the 28thday compressive
strength.

2.2. Method
A grade 25 concrete was produced using water/cement ratio of 0.4, 0.5, 0.6 and 0.65 as shown in
Table 1. Fifty (50) concrete cubes were cast for each mix proportion using 150mm x 150mm x
150mm metal moulds. A total of 200 sets of concrete cubes were cast, de-moulded after
24hours and then immersed in water for curing ages of 7 and 28 days respectively.

http://www.azojete.com.ng


Arid Zone Journal of Engineering, Technology and Environment, September, 2019; Vol. 15(3):692-701. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com

695

Table 1: Concrete Mix Proportion
Cement (kg/m3) Fine Aggregate

(kg/m3)
Coarse Aggregate
(kg/m3)

Water (kg/m3) Water Cement
Ratio (W/C)

360.4 720.8 1441.7 144.2 0.4
348.8 697.6 1395.2 174.4 0.5
337.6 675.2 1350.4 202.6 0.6
332.2 664.5 1329.0 216.0 0.65

2.2.1 Compressive Strength
The compressive strengths of the cubes were carried out using the mix proportions in Table 1.
Mixing was done manually and cast in steel cube molds of 150mm and cured in water for 7 and
28 days. A total of 400 cubes amounting to 200 sets were tested at the end of each curing
regime. The samples were crushed in accordance with BS EN12390-3 (2009) using the Avery
Denison Compressive Testing machine of 2000 kN load capacity at constant loading rate of
15kN/s and the average loading rate taken. The compressive strength was taken as the
maximum compressive load the cube can carry per unit areas and was calculated as follows:

A
FstrengtheCompressiv  (1)

2.2.2 Network Development, Training, Validation and Testing
ANN tool box in MATLAB (R2011a) was used in developing the network architecture. Input data
in the form of 6-column matrix containing the input parameters: water content (kg/m3), cement
content (kg/m3), fine aggregates (kg/m3), coarse aggregates (kg/m3) and 7th day compressive
strength (N/mm2) were fed into the workspace. These automatically made the input neurons in
the input layer to be six (6). The training algorithm adopted for the study in the feed forward
network is the Levenberg – Macquardt’s and after several random trials, the number of neurons
in the hidden layer for optimal convergence was found to be three (3). Since the output
parameter is one (1) (28th day compressive strength), the number of neurons in the output layer
was one (1). The final network architecture for the study and the workflow showing the step-by-
step ANN development processes are as shown in Figures 2 and 3 respectively.

Figure 2: Final Artificial Neural Network Architecture

file:///C:/Users/user/Downloads/azojete143/www.azojete.com.ng


Uche, et al: Prediction modeling of 28-day concrete compressive strength using artificial neutral network. AZOJETE,
15(3):692-701. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com 696

Figure 3: Flow Chart for the Artificial Neural Network Development

The imported data were randomized automatically by the system and divided into three (3); 55%
for training, 15% for validation and 30% for testing. Thus, among the 200 data sets for this study,
110 randomly collected data were used in the training stage to construct the ANN model, 30 for
validation and the remaining 60 data set were used in testing the network to estimate the model
performance. The network was trained and retrained until it understood the relationship
between the input variables and the 28th day compressive strength (output variable).

Back-propagation, as one of the most well-known training algorithms for the multilayer
perception (MLP), which is a gradient descent technique that minimizes the error for a particular
training pattern in which it adjusts the weights by small amount at a time, was used. However, it
was recommended by El-Chabib and Nehdi (2005) that a sigmoid function be employed as
transfer function for the hidden layers because it was non-linear, differentiable, continuous, and
it varies monotonically between 0 and 1 (log-sigmoid) or -1 and 1 (tan-sigmoid). Momentum
rate and learning rate values were determined and the model was trained through iterations. The
trained model was tested with the input values and the results found were close to experiment
results. The values of parameters used in this study were as follows:
Number of input layer units = 6
Number of hidden layers = 3
Number of output layer unit = 1
Learning cycle = 24

http://www.azojete.com.ng


Arid Zone Journal of Engineering, Technology and Environment, September, 2019; Vol. 15(3):692-701. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com

697

3. Results and Discussion
This section presents the results of the laboratory tests for concrete cubes strengths in order to
form the database of sets of data for use in the artificial neural network modelling. It also
describes the findings on the suitability of ANN in the prediction of 28th day concrete
compressive strength.

3.1 Compressive Strength of Concrete Cubes
The prepared samples of 150mm by 150mm by 150mm with four different water cement ratios
(0.4, 0.5, 0.6, 0.65) were tested at 7th and 28th days using compression machine and their
strength results were obtained as shown in Table 2.

Table 2: Results of Compressive Strength at Various Concrete Mix Constituent
Concrete Mix Constituent (kg/m3) Compressive Strength (N/mm2)
W/C ratio Cement Coarse Aggregate Fine Aggregate 7th Day 28th Day
0.4 360.4 1441.7 720.8 24.9 38.0
0.5 348.8 1395.2 697.6 20.1 30.5
0.6 337.6 1350.6 675.2 16.0 23.7
0.65 332.2 1329.0 664.5 14.3 22.1

Figure 4: Compressive Strength at Various Water-Cement Ratio

Figure 4 shows the relationship between compressive strength (N/mm2) at 7th and 28th day with
water-cement ratio, this shows an inverse relationship, where the compressive strength
decreased with increase in water-cement ratio, hence the compressive strength increased with
increase in cement content and decreased with increase in water content. Also from Table 2 it
can be deduce that the compressive strength decreased with decrease in fine and coarse
aggregate content, hence a direct relationship exist between compressive strength and
aggregates content.

file:///C:/Users/user/Downloads/azojete143/www.azojete.com.ng


Uche, et al: Prediction modeling of 28-day concrete compressive strength using artificial neutral network. AZOJETE,
15(3):692-701. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com 698

Figure 5: Relationship between 7th and 28th Day Compressive Strength

Figure 5 shows an approximately linear relationship between the 7th day and the 28th day
concrete strength, with the 28th day increasing with increase in the 7thday strength. The
relationship between the 28thday and 7thday concrete strength is defined by Equation (2).

(2)
where:
f28 = 28th day compressive strength (N/mm2)
f7 = 7th day compressive strength (N/mm2)

Table 3: Comparison between Experimental and Mathematical Model Strengths at 28th Day
Experimental result (N/mm2)

f�u � 쳌䁐�䀀㌳㌳f� � Ͷ䁐䀀Ͷⷸ�
38 38.52
30.5 31.15
23.7 24.86
22.1 22.25

The comparison between the experimental 28th day strength and the 28th day strength
obtained using the mathematical model as in Equation (2) is presented in Table 3. It was
observed that the model result is slightly higher than the experimental results.

3.2 Outcomes of the ANN
The results of the network testing after training and validation shows that the predicted 28th day
compressive strength of concrete are very close to those measured in the laboratory. This is an
indication that the network has learned the relationship between the input and the output
values during the training. The comparison between the measured and predicted compressive
strength at 28th day are shown in Figures 6 and 7.

http://www.azojete.com.ng


Arid Zone Journal of Engineering, Technology and Environment, September, 2019; Vol. 15(3):692-701. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com

699

Figure 6: Relationship between Measured and Predicted 28th Day Compressive Strength

Figure 7: Comparison between Measured and Predicted 28th Day Compressive Strength

In general, the R value for the training data set was larger than that for the testing set, that is,
the neural network made better prediction for the training data sets than the testing data set.
The combination of transfer function composed of tan-sigmoid and linear function gives a good
result. Figure 8 shows the relationship between output targets and predicted values obtained
through the training and testing process. The model shows very good correlation for both the
training (R = 0.99751), validation (R = 0.99736) and testing data (R = 0.99482) and the general
correlation of R = 0.99675.

file:///C:/Users/user/Downloads/azojete143/www.azojete.com.ng


Uche, et al: Prediction modeling of 28-day concrete compressive strength using artificial neutral network. AZOJETE,
15(3):692-701. ISSN 1596-2490; e-ISSN 2545-5818, www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com 700

Figure 8: Correlation Between Measured and ANN Predicted Result

4. Conclusion
From the study, the following conclusion and observations were drawn.
The 28thday compressive strength was found to be directly proportional to 7th day compressive
strength. It has been demonstrated in this study that the ANN model ��u � 쳌䁐�䀀㌳㌳�� � Ͷ䁐䀀Ͷⷸ�
was quite efficient in determining the 28th day compressive strength of concrete asit was found
that the measured compressive strength and predicted strength are very close with a correlation
of 0.99675. Mathematical model results are slightly higher than the experimentally results.

References
Akkurt, S., Tayfur, G. and Can, S. 2004. Fuzzy Logic Model for the Prediction of Cement
Compressive Strength. Cement Concrete Research, Vol. 34, No. 8, PP. 1429 – 1433.
ASTM C1602-12. Standard Specification for Mixing Water used in the ProductionHydraulic
Cement Concrete, ASTM International, West Conshohocken, PA, 2012, www.astm.org
Bamiyo, S.P., Uche, O.A.U. and Abdulwahab, M.T. 2016. Load Deflection Behaviour of Two-Way
RC Slab using Artificial Neural Network Approach. Proceedings of the 2nd National Engineering
Conferenceon Bridging the Gap between Academia and Industry in Nigeria, Bayero University
Kano, PP. 144 – 149, ISBN: 978-978-8203-24-7.

http://www.astm.org
http://www.azojete.com.ng


Arid Zone Journal of Engineering, Technology and Environment, September, 2019; Vol. 15(3):692-701. ISSN 1596-2490;
e-ISSN 2545-5818; www.azojete.com.ng

Corresponding author’s e-mail address: amtaytechnology@yahoo.com

701

BS EN 12390-3 2009. Testing Hardened Concrete—Part 3: Compressive Strength of Test
Specimens. Management Centre, Brussels.
Chopra, P., Sharma, R.K. and Kumar M. 2015. Artificial Neural Networks for the Prediction of
Compressive Strength of Concrete. International Journal of Engineering and Applied Sciences
(IJEAS), Vol. 13, Issue 3, PP. 187 – 204.
Dias, W. P. S. and Pooliyadda, S. P. 2001.Neural networks for predicting properties of Concretes
with admixtures. Construction and Building Materials, Vol. 15, pp. 371- 379.
El-Chabib, H. and Nehdi, M. 2005. Neural Network Modelling of Properties of Cement-Based
Materials Demystified. Advances in Cement Research, Vol. 17, No. 3, PP. 91 – 102.
Flood, I. 2008. Towards the Next Generation of Artificial Neural Networks for Civil Engineering.
Advanced Engineering Informatics, Vol. 22, No. 1, PP. 4 – 14.
Gambhir, M. L. 2004. Concrete Technology, Tata McGraw – Hill Publishing Company Limited,
New Delhi, PP. 352 - 448
Hwang, K., Noguchi, T. and Tomosawa, F. 2004. Prediction Model Compressive Strength
Development of Fly-Ash Concrete. Cement Concrete Research, Vol. 34, No. 12, PP. 2269 – 2276.
Kayode, O. and Ilesanmi, O. 2015.Self-Compacting Concrete: Strength Evaluation of Corn Cob
Ash in a Blended Portland Cement. American Scientific Research Journal for Engineering,
Technology, and Sciences Vol. 13, No. 1, PP. 123-13.
Kheder, G.F., Al-Gabban, A.M. and Abid, S.M. 2003. Mathematical Model for the Prediction of
Cement Compressive Strength at the Ages of 7 and 28 Days Within 24hours. Materials and
Structure, Vol. 36, No. 10, PP. 693 – 701.
Mehta, P.K. and Monteiro, P.J.M. 2006. “Concrete, microstructure, Properties and Materials”. 3rd
Edition, McGraw-Hill, USA.
Monjurul, H.M. and Ahsanul, K. 2011. Prediction of Compressive Strength of Concrete from Early
Age Test Result. Proceedings of 4th Annual Paper Meet and 1st Civil Engineering Congress,
Dhaka, Bangladesh, PP. 1 – 7.
Rafiq, M.Y., Bugmann, I. and Easterbrook, D.J. 2001. Neural Network Design for Engineering
Applications. Computers and Structures, Vol. 79, PP.1541 – 152.
Shain, M.A., Mark, B.J. and Holger, R.M. 2001. Artificial Neural Network-Based Settlement
Prediction Formula for Shallow Foundations on Granular Soils.School of Civil and Environmental
Engineering, the University of Adelaide.
Zain, M.F.M. and Abd, S.M. 2009. Multiple Regressions Model for Compressive Strength
Prediction of High Performance Concrete. Journal of Applied Sciences, Vol. 9, No. 1, PP. 155 –
160.
Zain, M.F.M., Mahmud, H.B., Ilham, A. and Faizal, M. 2002. Prediction of Splitting Tensile Strength
of High-Performance Concrete. Cement Concrete Research, Vol. 32, No. 8, PP. 1251 – 1258.
Zelic, J., Rusic, D. and Krstulovic, R. 2004. A Mathematical Model for Prediction of Compressive
Strength in Cement-Silica Fume Blends. Cement Concrete Research, Vol. 34, No. 12, PP. 2319 –
2328

file:///C:/Users/user/Downloads/azojete143/www.azojete.com.ng