Hindawi

Advances in Materials Science and Engineering

Volume 2020, Article ID 9682740, 8 pages
https://doi.org/10.1155/2020/9682740

Research Article
Prediction of Later-Age Concrete Compressive Strength Using
Feedforward Neural Network

Thuy-Anh Nguyen, Hai-Bang Ly, Hai-Van Thi Mai , and Van Quan Tran
University of Transport Technology, Hanoi 100000, Vietnam

Correspondence should be addressed to Hai-Van i Mai; [email protected]

Received 26 June 2020; Revised 11 August 2020; Accepted 18 August 2020; Published 8 September 2020

Guest Editor: Hainian Wang

Copyright © 2020 uy-Anh Nguyen et al. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
Accurate prediction of the concrete compressive strength is an important task that helps to avoid costly and time-consuming
experiments. Notably, the determination of the later-age concrete compressive strength is more difficult due to the time required
to perform experiments. erefore, predicting the compressive strength of later-age concrete is crucial in specific applications. In
this investigation, an approach using a feedforward neural network (FNN) machine learning algorithm was proposed to predict
the compressive strength of later-age concrete. e proposed model was fully evaluated in terms of performance and prediction
capability over statistical results of 1000 simulations under a random sampling effect. e results showed that the proposed
algorithm was an excellent predictor and might be useful for engineers to avoid time-consuming experiments with the statistical
performance indicators, namely, the Pearson correlation coefficient (R), root-mean-squared error (RMSE), and mean squared
error (MAE) for the training and testing parts of 0.9861, 2.1501, 1.5650 and 0.9792, 2.8510, 2.1361, respectively. e results also
indicated that the FNN model was superior to classical machine learning algorithms such as random forest and Gaussian process
regression, as well as empirical formulations proposed in the literature.

1. Introduction earliest possible time after curing, especially in the first

hours, can be helpful for necessary adjustment.
In the field of civil engineering, concrete plays a significant In recent decades, the development of suitable tools to
role and is the most important construction material [1–7]. obtain the desired compressive strength has been the subject
Concrete is a kind of artificial stone material made by of numerous researches. Abrams’ law is the well-known
molding and solidifying an appropriate mixture of cement, concept in the field of civil engineering for calculating the
water, coarse, and fine aggregate. To further enhance the compressive strength of concrete with two inputs, namely,
properties of concrete, several chemical or mineral admix- cement and water. e results of Abrams’ law showed that an
tures such as fly ash, blast furnace slag, silica fume, meta- increase of compressive strength is inversely proportional to
kaolin, and superplasticizer could be added into the mixture. the water-to-cement ratio. Besides, the compressive strength
Compressive strength is the most important factor that of concrete also depends on the content and nature of the
reflects the quality of a concrete mix design [8]. e concrete mix design, the curing condition, and the age of specimens
compressive strength is usually characterized by estimating [10–12]. Apart from Abrams’ law, Hasan and Kabir [8] and
the values at 28 days. e compressive strength directly ACI 209-71 [13] have proposed empirical equations to
depends on the ingredients of concrete and their relative determine the compressive strength of concrete at any age
amount [9], making the procedure of selecting appropriate (Table 1). However, the empirical methods could not take
mixture components critical. erefore, knowing the po- into account all factors that affect the concrete compressive
tential compressive strength of later-age concrete at the strength, such as supplementary cementitious materials or
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 Advances in Materials Science and Engineering

Table 1: Summary of the empirical equations for determining the compressive strength of concrete.
Ref. Equation Symbol Time-dependent
Abrams’ law S � (A/BW/C ) S: compressive strength; A, B: constant; W/C: water/cement ratio No
Hasan and Kabir [8] Stn � (pDn /(Dn + q)) p, q: constant; n: day; Dn: number of days Yes
ACI 209-71 [13] Stn � (t/(a + bt))S28d a, b: constant; t: number of days; S28d: compressive strength at 28 days Yes

the mixture components. erefore, these methods cannot 2. Machine Learning Method
correctly estimate the compressive strength of concrete
using supplementary cementitious materials. Alternative 2.1. Feedforward Neural Network. Feedforward neural net-
approach, thus, needs to be investigated to predict the later- work (FNN) is the most commonly used neural network in
age compressive strength of concrete accurately. the literature [22]. It could be considered as one of the first
Recently, development and application of machine and most successful machine learning algorithms. FNN
learning techniques in the civil engineering sector have could also be called a multilayer perceptron (MLP) or in a
gained extensive attention [14]. Taking advantage of an simple manner as neural networks. e FNN structure
extensive experimental database, machine learning algo- contains four main compositions: the input layer, output
rithms exhibit the ability to simplify classical approaches, layer, one or more hidden layers, and the weights associated
such as experimental or numerical simulation. During the with each neuron in the hidden layer(s) [23]. In the input
last four decades, machine learning techniques have been layer, it contains a certain number of neurons, which cor-
proven to successfully solve many problems of civil engi- responds to the number of features in the dataset. e input
neering such as coastal engineering, construction engi- layer receives the information related to the inputs and
neering [15], earthquake engineering, geotechnical passes such information to the second layer, the hidden
engineering, environmental engineering, structural engi- layer(s). e hidden layer is located between the input and
neering [16, 17], and materials science engineering [18]. output layers, which contains many neurons to apply the
Moreover, taking into account numerous input factors is an transformation from the input layer to the output layer. Each
essential advantage of machine learning techniques. neuron in the hidden layer(s) is associated with a weight,
erefore, machine learning technique appears to be a representing the amplitude of a connection between two
potential approach to predict the later-age compressive neurons. e output layer is the target of the problem or the
strength of concrete. Among the machine learning algo- feature that we want to predict [24].
rithms that are currently being utilized, the feedforward e principle of the process of FNN can be summarized
neural network (FNN) is the most popular model due to the in 3 steps as follows: (i) multiplication of weights and inputs,
structural flexibility, excellent prediction performance, and where each feature in the input layer is multiplied by the
the availability of a significant amount of training algorithms weight associated; (ii) biases added: the product in the
[19]. With the ability to deal with complex functions, FNN is previous step is added, each with a bias; and (iii) activation
an excellent predictor in many studies, such as compressive step: an activation function is applied to map the features in
strength of concrete [20] or bearing capacity of steel tube the input layer to the output layer (Figure 1). Finally, the
columns [21]. desired output is obtained by summing up all the results of
In this paper, the FNN algorithm was developed and the previous steps (for each neuron). e difference, or the
used to predict the later-age compressive strength of con- error, of the predicted and actual outputs, is calculated and
crete. For this aim, a database containing 190 samples, called a loss function, which measures the algorithm
gathered from the available literature, was used to train and performance.
validate the FNN algorithm. e quality of the FNN model e effectiveness of FNN has been proven in many works
was assessed by common statistical measurements such as [25, 26], especially in material modeling [20, 27]. erefore,
root-mean-square error (RMSE), mean absolute error FNN with ten neurons is proposed to predict the later-age
(MAE), and the Pearson correlation coefficient (R). e compressive strength of concrete in this study. Detail in-
paper is structured as follows: the introduction section is formation on the FNN, as well as all equations governing the
presented in Section 1. In Section 2, the machine learning algorithm, could be found in the literature [28].
feedforward neural network (FNN) is introduced and
fundamental elements are focused. Section 3 presents the
database used for the development of FNN, as well as the 2.2. Performance Criteria. In this study, the performance of
statistical information of input and output variables. Section FNN is evaluated by common statistical measurements such
4 shows the results obtained by FNN and emphasizes the as the Pearson correlation coefficient (R), root-mean-square
superior prediction accuracy by comparing with classical error (RMSE), and mean absolute error (MAE). Typically,
machine learning models such as random forest and the values of R are in the [−1, 1] range, and better prediction
Gaussian process regression. Finally, Section 5 provides accuracy is obtained when the absolute value of R is close to
some conclusions and perspectives. 1. Oppositely, lower values of RMSE and MAE designate a
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Advances in Materials Science and Engineering 3

BFS

FA Later-age compressive
strength
W
LACS
SP

CA

FA

8 inputs
Figure 1: Illustration of the FNN algorithm used in this study, including 8 inputs.

higher performance of the given machine learning algo- Table 2: Summary of the statistical information of the database
rithm. e expressions of the indicators are as follows used in the present work.
[29, 30]:
Min Average Median Max SD
1 n C 102 294.95 288.5 540 103
MAE � 􏽘 u − ui 􏼁, BFS 0 66.9 0 305.3 82.2
n i�1 i FA 0 31.93 0 174.7 55.8
􏽶������������ W 121.8 190.19 192 228 25.8
􏽴 SP 0 3.93 0 32.2 6
1 n 2 CA 852.1 979.73 968 1134 64.6
RMSE � 􏽘 u − ui 􏼁 , (1)
n i�1 i FA 594 762.22 780.35 993 98.5
A 90 146.75 100 365 90
LACS 21.86 49.97 44.21 82.6 12.8
n 2
2 􏽘i�1 ui − ui 􏼁 SD � standard deviation.
R �1− 2,
􏽐ni�1 ui − u􏼁

where ui is the predicted output value, ui is the measured 4. Results and Discussion
target value, and n is the number of samples. 4.1. Performance of FNN. e samples of the training part
greatly affect the accuracy of the machine learning black box
3. Database Collection [36] so that the sampling technique is important in the
construction phase of the model. A certain number of
In this study, a database containing 190 samples was col- samples are randomly taken to be in the training part.
lected from the literature [31–35]. e input parameters of Besides, there is systematic sampling, stratified sampling, or
the concrete compressive strength database include the cluster sampling, belonging to the probability sampling
content of cement (denoted as C), the content of blast technique. In this study, simple random sampling was
furnace slag (denoted as BFS), the content of fly ash (denoted chosen as it was an effective way to select samples. e
as FA), water content (denoted as W), superplasticizer performance of FNN is evaluated throughout the sampling
(denoted as SP), aggregate content including coarse ag- technique in this section.
gregates (denoted as CA) and fine aggregates (denoted as A number of 1000 simulations were performed in
FA), and testing age (denoted as A). e considered output is shuffling the indexes of the database, and 70% of data were
the later-age compressive strength (LACS) of concrete. taken to construct the training part. e results of 1000
For the statistical information of the database, Table 2 simulations are given by the histograms of values R (Fig-
presents the min, average, max, median, and standard de- ure 2), RMSE (Figure 3), and MAE (Figure 4) for the training
viation values (denoted as SD) of all the parameters used in and testing parts. e values of R of the FNN algorithm
this study. showed a strong prediction capability of LACS as the
e database containing 190 samples is divided ran- minimum values of R were in the R � 0.7–1 range. us, a
domly into two parts, namely, the training part (70% of the strong correlation was observed between the predicted
total samples) and the testing part (30% of the remaining output given by FNN and the experimental results.
samples). e training part is used to construct the FNN e training dataset exhibited greater accuracy than the
model, whereas the testing subset was dedicated to the as- testing dataset, which was helpful in preventing overfitting
sessment of the accuracy of the model. problems. A high concentration of R close to 1 was observed,
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 Advances in Materials Science and Engineering

400 150

300
100

Frequency
Frequency

200

50
100

0
0 0 5 10 15
0.4 0.5 0.6 0.7 0.8 0.9 1
MAE
R
Training dataset
Training dataset
Testing dataset
Testing dataset
Figure 4: Histogram of MAE values over 1000 simulations for
Figure 2: Histogram of R values over 1000 simulations for the training and testing parts.
the training and testing parts.

150 Table 3: Statistical results over 1000 simulations using FNN in

this study.
Min Average Max SD
Training
100 R 0.8179 0.9391 0.9677 0.0233
RMSE 3.1481 4.1988 7.1676 0.7116
Frequency

MAE 2.4860 3.0472 3.8312 0.3110

Testing
R 0.4379 0.9350 0.9781 0.0698
50 RMSE 2.7580 4.3342 11.5334 1.5287
MAE 2.2113 3.2552 7.5907 1.0247
SD � standard deviation.

0
0 5 10 15
RMSE
80
Training dataset
Testing dataset 70
Compressive strength (MPa)

Figure 3: Histogram of RMSE values over 1000 simulations for the

training and testing parts. 60

and R � 0.94 was the most frequent result obtained for the 50
testing dataset.
Similarly, the RMSE and MAE values showed comple- 40
mentary information on the observations. e most frequent
values of RMSE and MAE were 2.5 and 2.1 for the training 30
dataset, whereas those of the testing dataset were 5 and 3.9,
respectively. In all the cases, no extreme values were ob-
20
served for the training dataset, showing that the prediction 0 50 100 150
performance of FNN was satisfactory. Table 3 shows the Sample index
values of R, RMSE, and MAE for 1000 simulations per-
Target
formed in this study, along with the standard deviation
Output
values associated with each case. e average values of R
were 0.9391 and 0.9350 for the training and testing datasets, Figure 5: Prediction results of FNN for the training dataset.
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Advances in Materials Science and Engineering 5

90 6
Error mean = 1.3724
80
5 Error st.D. = 2.5211

70
Compressive strength (MPa)

4
60

Frequency
3
50

40 2

30 1

20
0 10 20 30 40 50 60 0
Sample index –10 –5 0 5
Error
Target
Output
Figure 8: Histogram of error for the testing dataset and a fit
corresponds to the distribution of error.
Figure 6: Prediction results of FNN for the testing dataset.

14 70
Error mean = 0.08317
12 Error st.D. = 2.1566

10 60
Predicted data
Frequency

8
50
6

4 40

30 Ytrain = 0.97Xtrain + 1.3

0
–5 0 5 10
Error Ytest = 0.98Xtest – 0.4

Figure 7: Histogram of error for the training dataset and a fit 30 40 50 60 70

corresponds to the distribution of error. Actual data
Training data Testing data
respectively. Overall, FNN is a promising algorithm to Training (R = 0.9861) Testing (R = 0.9792)
predict the LACS of concrete. Diagonal
Figure 9: Regression graph for the training and testing datasets
using the FNN model.
4.2. FNN Typical Configuration Result. e results of a
typical FNN configuration are presented in this section.
Figure 5 shows the comparison between LACS results of the and testing datasets, respectively. It can be seen that the
FNN model and the experimental data for the training accuracy of the training part was slightly superior to the
dataset. e comparison shows that the predicted results testing part, but both results showed excellent accuracy.
were very close to the experimental results. Figure 6 shows Finally, the relationships of the actual and predicted data
that the FNN algorithm can successfully predict the ex- were given in the form of regression plots in Figure 9. e
perimental compressive strength for the testing dataset. values of R were 0.9861 and 0.9792 for the training and
e histograms of error with respect to the training and testing datasets, respectively. Two equations related to such
testing parts are displayed in Figures 7 and 8, respectively. relationships were given in the graph, where
e mean errors were 0.08317 for the training part and “Y � 0.97X + 1.3” was the best fit line for the training set and
1.3724 for the testing part. e standard deviations of error “Y � 0.98X − 0.4” was the best fit linear regression for the
values were computed as 2.1566 and 2.5211 for the training testing set.
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 Advances in Materials Science and Engineering

80 80

70 70

60
60

Predicted data
Predicted data

50
50

40
40

30 Ytrain = 0.8Xtrain + 9.6 Ytrain = 0.87Xtrain + 6

Ytest = 0.66Xtest + 16 30 Ytest = 0.84Xtest + 8.9

30 40 50 60 70 80 30 40 50 60 70 80
Actual data Actual data
Training data Testing data Training data Testing data
Training (R = 0.9654) Testing (R = 0.9006) Training (R = 0.9301) Testing (R = 0.9343)
Diagonal Diagonal
(a) (b)

Figure 10: Regression graphs for the training and testing datasets for (a) RF algorithm and (b) GPR algorithm.

Table 4: Performance analysis of FNN, RF, and GPR algorithms. appears as a promising numerical tool to predict the later-
age compressive strength of concrete.
FNN RF GPR
Training dataset
R 0.9861 0.9654 0.9301 5. Conclusion
RMSE 2.1501 3.6900 4.6411
MAE 1.5650 2.7728 3.1891 In this study, the FNN algorithm was introduced to predict
Testing dataset
the later-age compressive strength of concrete containing
R 0.9792 0.9006 0.9343 cement, fly ash, blast furnace slag, water, and aggregates. A
RMSE 2.8510 6.2663 4.9469 number of 190 experimental results were collected to
MAE 2.1361 5.0560 3.7502 construct the FNN model. A number of 70% of data was
randomly chosen for the training phase of FNN, and 30% of
data was chosen for the validation phase of the constructed
e FNN algorithm showed an excellent performance in FNN model. To fully evaluate the performance of FNN, a
predicting the LACS of concrete. It is worth noticing that the number of 1000 simulations were performed using a random
values of R could reach 0.9792 with the use of a simple FNN sampling technique, and the evaluation of FNN model was
structure FNN (8-10-1), representing one hidden layer and carried out by common statistical measurements.
ten neurons in such hidden layer. e results showed that the FNN was a promising al-
For comparison purpose, the performance of FNN al- gorithm to predict the LACS of concrete, with the average
gorithm was shown superior to other benchmark machine values of 0.9350, 4.3342, and 3.2552 for R, RMSE, and MAE,
learning algorithms, such as random forest (RF) and respectively, for the testing part. e most accurate FNN
Gaussian process regression (GPR). e detail and devel- model exhibited an excellent prediction performance with
opment of RF and GPR algorithms are described in [37, 38], R � 0.9792, RMSE � 2.8510, and MAE � 2.1361. ese results
respectively. e regression graphs for the training and were shown superior to other classical machine learning
testing datasets of RF and GPR algorithms are shown in algorithms, namely, RF and GPR. e results could help in
Figure 10. A summarized performance of FNN, RF, and GPR constructing a reliable soft computing tool to predict ac-
algorithms is shown in Table 4. curately and quickly the later-age compressive strength of
Overall, the FNN algorithm exhibited the best prediction concrete. Once the tool is built, the forecast process would
performance with respect to RMSE, MAE, and R, compared take a millisecond to estimate the compressive strength of
with RF and GPR algorithms. erefore, the FNN algorithm the given mix design.
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Advances in Materials Science and Engineering 7

Data Availability manufactured sand concrete,” Applied Sciences, vol. 9, no. 18,
p. 3841, 2019.
e processed data are available from the corresponding [15] L. M. Le, H.-B. Ly, B. T. Pham et al., “Hybrid artificial in-
author upon request. telligence approaches for predicting buckling damage of steel
columns under axial compression,” Materials, vol. 12, no. 10,
p. 1670, 2019.
Conflicts of Interest [16] H.-B. Ly, L. M. Le, H. T. Duong et al., “Hybrid artificial
e authors declare that they have no conflicts of interest. intelligence approaches for predicting critical buckling load of
structural members under compression considering the in-
fluence of initial geometric imperfections,” Applied Sciences,
References vol. 9, no. 11, p. 2258, 2019.
[17] H. Q. Nguyen, H.-B. Ly, V. Q. Tran, T.-A. Nguyen, T.-T. Le,
[1] V. Q. Tran, A. Soive, S. Bonnet, and A. Khelidj, “A numerical
and B. T. Pham, “Optimization of artificial intelligence system
model including thermodynamic equilibrium, kinetic control
by evolutionary algorithm for prediction of axial capacity of
and surface complexation in order to explain cation type effect
rectangular concrete filled steel tubes under compression,”
on chloride binding capability of concrete,” Construction and
Materials, vol. 13, no. 5, p. 1205, 2020.
Building Materials, vol. 191, pp. 608–618, 2018.
[18] D. Dao, H.-B. Ly, S. Trinh, T.-T. Le, and B. Pham, “Artificial
[2] V. Q. Tran, A. Soive, and V. Baroghel-Bouny, “Modelisation
of chloride reactive transport in concrete including ther- intelligence approaches for prediction of compressive strength
modynamic equilibrium, kinetic control and surface com- of geopolymer concrete,” Materials, vol. 12, no. 6, p. 983, 2019.
plexation,” Cement and Concrete Research, vol. 110, pp. 70–85, [19] D. J. Montana and L. Davis, “Training feedforward neural
2018. networks using genetic algorithms,” Proceedings of the 11th
[3] V. Q. Tran, H. L. Nguyen, V. D. Dao et al., “Temperature International Joint Conference on Artificial Intelligence -,
effects on chloride binding capacity of cementitious mate- vol. 1, no. 89, pp. 762–767, 1989.
rials,” Magazine of Concrete Research, pp. 1–14, 2019. [20] H.-G. Ni and J.-Z. Wang, “Prediction of compressive strength
[4] A. Soive, V.-Q. Tran, and M. Gasc-Barbier, “e advantages of of concrete by neural networks,” Cement and Concrete Re-
using a geochemical transport model including thermody- search, vol. 30, no. 8, pp. 1245–1250, 2000.
namic equilibrium, kinetic control and surface complexation [21] Q. H. Nguyen, H.-B. Ly, V. Q. Tran et al., “A novel hybrid
to simulate the durability of concretes exposed to chlorides model based on a feedforward neural network and one step
and sulphates,” European Journal of Environmental and Civil secant algorithm for prediction of load-bearing capacity of
Engineering, vol. 23, no. 5, pp. 552–563, 2019. rectangular concrete-filled steel tube columns,” Molecules,
[5] A. Soive, V. Q. Tran, and V. Baroghel-Bouny, “Requirements vol. 25, no. 15, p. 3486, 2020.
and possible simplifications for multi-ionic transport mod- [22] H. Adeli, “Neural networks in civil engineering: 1989–2000,”
els—case of concrete subjected to wetting-drying cycles in Computer-Aided Civil and Infrastructure Engineering, vol. 16,
marine environment,” Construction and Building Materials, no. 2, pp. 126–142, 2001.
vol. 164, pp. 799–808, 2018. [23] A. Mukherjee and S. Nag Biswas, “Artificial neural networks
[6] A. Soive and V. Q. Tran, “External sulfate attack of cemen- in prediction of mechanical behavior of concrete at high
titious materials: new insights gained through numerical temperature,” Nuclear Engineering and Design, vol. 178, no. 1,
modeling including dissolution/precipitation kinetics and pp. 1–11, 1997.
surface complexation,” Cement and Concrete Composites, [24] P. Muthupriya, K. Subramanian, and B. G. Vishnuram,
vol. 83, pp. 263–272, 2017. “Prediction of compressive strength and durability of high
[7] V. Q. Tran, Contribution à la compréhension des mécanismes performance concrete by artificial neural networks,” Inter-
de dépassivation des armatures d’un béton exposé à l’eau de national Journal of Optimization in Civil Engineering, vol. 1,
mer: théorie et modélisation thermochimique, Ph.D. thesis, pp. 189–209, 2011.
École Centrale de Nantes, Nantes, France, 2016. [25] A. Khalaf, Z. Naser, and F. Idan, “Predicting the ultimate
[8] M. Hasan and A. Kabir, “Prediction of compressive strength strength of circular concrete filled steel tubular columns by
of concrete from early age test result,” in Proceedings of the 4th using artificial neural networks,” pp. 9–16, 2018.
Annual Paper Meet and 1st Civil Engineering Congress, Dhaka, [26] A. M. Al-Khaleefi, M. J. Terro, A. P. Alex, and Y. Wang,
Bangladesh, December 2011. “Prediction of fire resistance of concrete filled tubular steel
[9] S. V. Deodhar, Civil Engineering Materials, Khanna Pub- columns using neural networks,” Fire Safety Journal, vol. 37,
lishers, Delhi, India, 6th edition, 2009. no. 4, pp. 339–352, 2002.
[10] L. Zongjin, Advanced Concrete Technology, John Wiley & [27] Y. N. Chan, P. Jin, M. Anson, and J. S. Wang, “Fire resistance
Sons, Inc, Hoboken, NJ, USA, 1st edition, 2011. of concrete: prediction using artificial neural networks,”
[11] M. S. Shetty, Concrete Technology: 8eory and Practice, Multi- Magazine of Concrete Research, vol. 50, no. 4, pp. 353–358,
Color Illustration Edition and S. Chand, New Delhi, India, 1998.
2005. [28] M. Sazli, “A brief review of feed-forward neural networks,”
[12] M. O. Reilly, D. Darwin, J. Sperry, and J. Browning, “Variation Communications, Faculty of Science, University of Ankara,
of concrete strength , permeability , and porosity due to vol. 50, pp. 11–17, , Ankara, Turkey, 2006.
specimen,” e Kansas Department of Transportation, [29] H.-B. Ly, T.-T. Le, L. M. Le et al., “Development of hybrid
Topeka, KS, USA, 2017. machine learning models for predicting the critical buckling
[13] J. Meader, D. Janssen, and M. Eberhard, Structural Design load of I-shaped cellular beams,” Applied Sciences, vol. 9,
Parameters of Current WSDOT Mixtures, National Trans- no. 24, p. 5458, 2019.
portation Library, Washington, DC, USA, 2013. [30] H.-B. Ly, L. M. Le, L. V. Phi et al., “Development of an AI
[14] H.-B. Ly, B. Pham, Dao, L. Le, Le, and Le, “Improvement of model to measure traffic air pollution from multisensor and
ANFIS model for prediction of compressive strength of weather data,” Sensors, vol. 19, no. 22, p. 4941, 2019.
 5928, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/9682740 by National Institute Of Technology (Nit) Patna, Wiley Online Library on [27/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 Advances in Materials Science and Engineering

[31] I.-C. Yeh, “Modeling concrete strength with augment-neuron

networks,” Journal of Materials in Civil Engineering, vol. 10,
no. 4, pp. 263–268, 1998.
[32] I.-C. Yeh, “Design of high-performance concrete mixture
using neural networks and nonlinear programming,” Journal
of Computing in Civil Engineering, vol. 13, no. 1, pp. 36–42,
1999.
[33] I.-C. Yeh, “A mix proportioning methodology for fly ash and
slag concrete using artificial neural networks,” Chung Hua
Journal of Science and Engineering, vol. 1, pp. 77–84, 2003.
[34] 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.
[35] I.-C. Yeh, “Modeling of strength of high-performance con-
crete using artificial neural networks,” Cement and Concrete
Research, vol. 28, no. 12, pp. 1797–1808, 1998.
[36] H.-B. Ly, E. Monteiro, T.-T. Le et al., “Prediction and sen-
sitivity analysis of bubble dissolution time in 3D selective laser
sintering using ensemble decision trees,” Materials, vol. 12,
no. 9, p. 1544, 2019.
[37] T. A. Pham, H.-B. Ly, V. Q. Tran, L. V. Giap, H.-L. T. Vu, and
H.-A. T. Duong, “Prediction of pile axial bearing capacity
using artificial neural network and random forest,” Applied
Sciences, vol. 10, no. 5, p. 1871, 2020.
[38] D. V. Dao, H. Adeli, H.-B. Ly et al., “A sensitivity and ro-
bustness analysis of GPR and ANN for high-performance
concrete compressive strength prediction using a monte carlo
simulation,” Sustainability, vol. 12, no. 3, p. 830, 2020.