Scientia Iranica A (2012) 19 (2), 242–248

Sharif University of Technology

Scientia Iranica
Transactions A: Civil Engineering
www.sciencedirect.com

Research note

Prediction of shear strength of reinforced concrete beams using

adaptive neuro-fuzzy inference system and artificial neural network
J. Amani ∗ , R. Moeini
School of Civil Engineering, Iran University of Science and Technology (IUST), Tehran, Iran
Alaodoleh Semnani Institute of Higher Education, Hajiabad, Garmsar, P.O. Box 35815-333, Iran

Received 13 July 2011; revised 25 December 2011; accepted 25 January 2012

KEYWORDS Abstract In this paper, the Artificial Neural Network (ANN) and the Adaptive Neuro-Fuzzy Inference
Reinforced concrete beam; System (ANFIS) are used to predict the shear strength of Reinforced Concrete (RC) beams, and the models
Shear strength; are compared with American Concrete Institute (ACI) and Iranian Concrete Institute (ICI) empirical codes.
Artificial neural network; The ANN model, with Multi-Layer Perceptron (MLP), using a Back-Propagation (BP) algorithm, is used
Adaptive neuro-fuzzy to predict the shear strength of RC beams. Six important parameters are selected as input parameters
inference system; including: concrete compressive strength, longitudinal reinforcement volume, shear span-to-depth ratio,
Iranian concrete institute transverse reinforcement, effective depth of the beam and beam width. The ANFIS model is also applied
code; to a database and results are compared with the ANN model and empirical codes. The first-order Sugeno
American concrete institute fuzzy is used because the consequent part of the Fuzzy Inference System (FIS) is linear and the parameters
code. can be estimated by a simple least squares error method. Comparison between the models and the
empirical formulas shows that the ANN model with the MLP/BP algorithm provides better prediction for
shear strength. In adition, ANN and ANFIS models are more accurate than ICI and ACI empirical codes in
prediction of RC beams shear strength.
© 2012 Sharif University of Technology. Production and hosting by Elsevier B.V.
Open access under CC BY-NC-ND license.

1. Introduction difficult to establish an overall model to provide accurate esti-

mation of shear strength. Hence, exact values of shear strength
Nowadays, the use of concrete structural members is in- are unknown. Several empirical formulas are proposed in the
creased. The strength of members in the design and behavior of literature and concrete codes for the prediction of RC beams
shear strength is an important issue in structural design. There resistance. The American Concrete Institute (ACI) code [1] has
been widely used in structural design. In addition, the Iranian
are several modes of failure in concrete structural members.
Concrete Instauration (ICI) code [2] is proposed for practical sit-
Due to the fragility of concrete structures, shear failure is one
uations in Iran. Each of the proposed empirical formulas in con-
of the most important and undesirable modes of failure. Hence,
crete codes yields good results just for a particular dataset. In
Reinforced Concrete (RC) members are used to resist shear fail-
the last decades, numerous works have been undertaken to im-
ure. Because of the complexity of shear mechanisms of rein- prove the ability of empirical formula to predict the shear be-
forced concrete beams and various influencing parameters, it is havior of concrete structural members [3–9].
In recent decades, Artificial Neural Networks (ANN) and
∗ Corresponding author. Tel.: +98 232 4533924, +98 914 1537221. Adaptive Neuro-Fuzzy Inference Systems (ANFIS) have been
E-mail addresses: [email protected] (J. Amani), [email protected] used to predict concrete properties. ANN is a powerful tool
(R. Moeini). for system modeling in a wide range of applications. However,
despite the excellent classification capacities of the latter, its
1026-3098 © 2012 Sharif University of Technology. Production and hosting by
development can be time-consuming and computer-intensive.
Elsevier B.V. Open access under CC BY-NC-ND license.
Peer review under responsibility of Sharif University of Technology. The most important advantage of the ANN model is that
the priority of the functional relationship among the various
doi:10.1016/j.scient.2012.02.009 variables is not required. The ANNs automatically build a
relationship for network architecture as experimental data
through a learning algorithm.
Kasperkiewicz et al. [10] applied ANN of the fuzzy type for
predicting the strength properties of High-Performance Con-
crete (HPC) mixes. Yeh [11] used augment-neuron networks to
 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248 243

model concrete strength. Guang and Zong [12] predicted the Table 1: ACI and ICI formulas for shear strength of RC beam.
compressive strength of concrete by ANN. Waszczyszyn and
Code Formula
Ziemianski [13] used ANN for the mechanics of structures and √
f′
materials. Bai et al. [14] used ANN to predict the workability of Vc = ( ρw VMuud ) bw d ≤ 0.3 fC′ bw d, Vu d

7
C
+ 120
7 Mu
≤1
ACI (N)
concrete with cement replacement materials. Hadi [15] proved Av fy d α=90° Av fy d
Vs = S
(sin αs + cos αs ) −→ Vs = S
that ANN is more accurate and easy to implement compared
√
to conventional methods for concert structural design. Oreta Vc = vc bw d, vc = 0.2φC fc , φc = 0.6
ICI (kg) α=90°
and Kawashima [16] applied the neural network in the model- Vs =
Av fyd d
(sin αs + cos αs ) −→ Vs =
Av fyd d
S S
ing of the confined compressive strength and strain of circular
concrete columns. In addition, Lee [17] used ANN for concrete
strength prediction. Kim et al. [18] applied ANN for estimation the ACI and ICI formulas are presented in Table 1 calculating
of concrete strength with concrete mix parameters. Oreta [19] the concrete shear strength in the absence of axial force and
simulated size effect on the shear strength of RC beams using web reinforcement. For the presented formulas of Table 1, fC′
a neural network. Mansour et al. [20] used ANN to predict the and fc are concrete compressive strength in MPa and kg/cm2 ,
shear strength of RC beams. Cladera and Mari [21] used ANN in respectively; ρw is the longitudinal steel ratio given by As /bw d;
beams with stirrups for the shear design procedure of normal Vu is factored shear force; Mu is factored bending moment
and high strength reinforced concrete beams. Abdalla et al. [22] occurring simultaneously with Vu at the section considered; bw
simulated the shear of RC beams with ANN. Kim and Kim [23] is beam width (cm); d is effective depth of the beam (cm); Av
worked on the prediction of the relative crest settlement of is shear reinforcement area (cm2 ); fy is yield strength of shear
concrete-faced rock-fill dams analyzed using an artificial neural reinforcement (kg/cm2 ); S is stirrups spacing (cm); and αs is
network model. Caglar et al. [24] applied the neural network to stirrups inclination angle.
dynamic analysis of reinforced concrete buildings. Arslan [25]
predicted the tensional strength of RC beams by ANN and com- 2.2. ANN and ANFIS models
pared it with building codes. Erdem [26] used ANN to predict
the moment capacity of RC slabs in fire. Takagi and Sugeno [27] ANN and ANFIS models are also suggested to compare
developed Fuzzy Inference Systems (FIS) and applied them in with empirical formulas in the prediction of RC beams shear
modeling and controlling concepts. In addition, Jang introduced strength. An introduction and different application of the ANN
ANFIS in his book [28]. Topçu and Saridemir [29] predicted model [10–26] and the ANFIS model [27–30] are presented in
rubberized mortar properties using ANN and fuzzy logic. Bil- the literature.
gehan [30] used ANFIS and Neural Network (NN) models to For application of ANN and ANFIS models in any field,
estimate critical buckling load. different input parameters should be considered. In this paper,
In this paper, ANN and ANFIS are used in intelligent data six important parameters are selected as input parameters for
analysis in the prediction of the shear strength resistance of the prediction of the shear strength of RC beams including:
RC beams. The layout of the paper is as follows. In Section 2, concrete compressive strength (fc ), longitudinal reinforcement
empirical formulas of ACI and ICI concrete codes and ANN and volume (ρw ), shear span-to-depth ratio (S /d), transverse
ANFIS models for prediction of the shear strength of RC beam reinforcement (ρl ), effective depth of the beam (d) and beam
are discussed. Also, performances of the ANN and ANFIS models width (bw ).
are assessed in Section 3 and the results are compared with ICI Furthermore, in ANN and ANFIS models, two sets of dataset,
and ACI codes to demonstrate the efficiency and accuracy of named training and testing datasets, are used. The training
the models. Finally, some concluding remarks are addressed in dataset is used to train the network, whereas the testing dataset
Section 4. is selected to verify the accuracy of the trained models for the
prediction of the shear strength of RC beams.
2. Prediction of shear strength of reinforce concrete beams Also, choosing the appropriate number of hidden neurons
and number of hidden layers are major parameters in obtaining
an accurate ANN model. In addition, the best selection of
Shear failure is one of the most important concepts in con-
activation function has a considerable effect on the ability of
crete structural members. Shear failure is caused by shear
the model. The number of hidden layers and number of nodes
forces. RC members resist shear force using several mecha-
in hidden layers are usually determined via trial and error
nisms. The shear failure is tolerated in RC beams by providing
procedures or using suggested rules. For example, based on the
web reinforcement. The web reinforcement usually takes the
method suggested by Anderson and McNeill [31], the number
form of vertical stirrups or the 45° bars that surround the lon-
of upper bound processing nodes of the hidden layers can be
gitudinal bars along the faces of the beam. The required shear
calculated by dividing the number of input–output pairs of the
strength to be provided by the web reinforcement is usually training set by the total number of input and output nodes of
computed by formulas in concrete codes. The required area of the network, multiplied by a scaling factor between five and ten.
web reinforcement or spacing of the stirrups can be determined Larger scaling factors are used for relatively noisy data [32].
based on the amount of predicted shear strength [19]. In this The ANFIS is considered here as another method in
paper, different methods, such as empirical codes and ANN and combining the advantages of FIS and ANN. ANFIS is a class of
ANFIS models, are applied to predict the shear strength of RC adaptive network that is functionally equivalent to FIS. The
members. first-order Sugeno fuzzy is used because the consequent part of
the FIS is linear and the parameters can be estimated by a simple
2.1. Empirical codes least squares error method. ANFIS is an example of models
in which the shape parameters of the membership functions
The ACI and ICI empirical codes are different proposed of fuzzy premise variables, as well as the linear parameters
formulas that are used in the design of concrete structural of the consequent part of fuzzy rules in a Takagi–Sugeno FIS,
members. For a member subjected to shear and flexure only, are tuned using ANNs. ANFIS is a Sugeno type of FIS in which
244 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248

(a) ICI code. (b) ACI code.

Figure 1: Comparison between observed and empirical codes.

the problem of fine tuning membership functions of premise

variables is carried out by a feed-forward neural network and
which combines the advantages of both a neural network and
FIS [33].

3. Results and discussions

In this section, an experimental database is used to

investigate the accuracy of ACI and ICI codes, and train and
test the ANN and ANFIS models in the prediction of the shear
strength of RC beams. The results of the ANN and ANFIS
models are compared with ACI and ICI emprical codes, based on
calculating the error values, such as Mean Absolute Error (MAE)
and Root Mean Square Error (RMSE). Also, the coefficient of
determination, and the R2 of the linear regression line between
the predicted values for models and codes are determined.
These three statistical parameters are used to compare the
performance of the various methods as follows:

 N Figure 2: Architecture of ANN model.
 (ti − yi )2

 i=1
RMSE = , (1) from codes are compared with measured experimental data.
N
Comparisons between observed (experimental) and predicted
N
shear strength are shown in Figure 1. It can be seen that the
(ti − yi )2

prediced shear strengths are more scattered from the observed
i=1
R2 = 1 −
N 
, (2) shear strengths.
 2 For construction of the ANN and ANFIS models, the 123
ti − ti
i=1 datasets are divided into two separate datasets randomly,
N
named training and testing datasets. 100 datasets are used
1  for training the network and the remaining 23 datasets are
MAE = |ti − yi | , (3)
N i=1 considered as testing datasets of the network. Figure 2 shows
the architecture of the final ANN model. Tables 2 and 3
where, ti and yi are the target and network output fot the ith show the range of different input–output parameters used for
output, ti is the average of target outputs, and N is the total training and testing datasets, respectively. Input parameters
number of events considered. The model that minimized the are selected, based on Bentz [34] and Bohigas [35] research.
error values (MAE and RMSE) and maximized R2 is selected as Various combinations of input parameters can be considered
the optimum network. in developing the ANN models, using experimental datasets,
To show the efficiency of the proposed methods, the leading to the construction of six types of architecture. At each
experimental datasets by Bentz [34] and Bohigas [35] are used step, one of the parameters is added to the network as the
and the shear strength of RC beams is calculated using all input parameter. It is shown that the network with all six input
proposed methods. Also, to evaluate the performance of the parameters has more accuracy and capability to predict the
ACI and ICI codes, values of the shear strength of RC beams shear strength of RC beams.
 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248 245

Table 2: Range of different input–output parameters for training data.

Parameter Training data (100 data set)

Min Average Max

Beam width (mm) 110 239.34 457

Beam depth (mm) 198 387.36 925
Concrete compressive strength (MPa) 25.7 64.455 125
Input parameters
Longitudinal reinforcement (%) 0.5 2.8171 5.8
Shear reinforcement (%) 0.33 0.8535 3.57
Shear span-to-depth 2.49 3.0346 5

Output parameter Shear strength (KN) 63.3 259.506 788

Table 3: Range of different input–output parameters for testing data.

Parameter Testing data (23 data set)

Min Average Max

Beam width (mm) 110 224.261 457

Beam depth (mm) 198 427.739 925
Concrete compressive strength (MPa) 21 58.8217 125
Input parameters
Longitudinal reinforcement (%) 0.75 2.5282 4.46
Shear reinforcement (%) 0.33 0.8847 3.43
Shear span-to-depth 2.5 2.9504 4
Output parameter Shear strength (KN) 93.6 242.939 721

For construction of ANN models, there is no theoretical

reason ever to use more than two hidden layers. For many
practical problems, there is no reason to use any more than
one hidden layer. Those problems that require two hidden
layers are only rarely encountered in real-life situations [36].
The number of hidden layers is considered equal to one to
improve the efficiency and compatibility of the model in
predicting and testing the accuracy of the new experimental
datasets faster and more easily, with a lower number of weights
and connections in the neurons. For selecting the number of
neurons in the hidden layer, the 19 models are constructed
using 2–20 neurons, respectively, and for each model, the
RMSE value is calculated. Figure 3 shows an optimum model is
obtained when 11 neurons are used in the hidden layer. Hence,
an ANN model with 6 input parameters, and 1 hidden layer with
11 neurons, based on the MLP/BP algorithm, is constructed.
To avoid over-training of the model, the network is trained
by initially performing with lower iterations. The number of
iterations is increased gradually (with an equal step), while the
Figure 3: Selection of number neurons in hidden layer.
process is stopped when the performance starts to drop. In this
way, the optimal ANN model is identified in 758 iterations.
Comparison results of the ANN model with experimental
results are shown in Figure 4. Figure 5 shows the comparison Table 4: Comparison results obtain using different models and codes.
between the prediction of the shear strength of the training and Method R2 RMSE MAE
testing data, as given by ICI, ACI codes and the ANN model. It is
ANN model (test) 0.9203 0.04340 0.03243
clear that the performance of the ANN MLP/BP model is much
ANFIS model (test) 0.9095 0.46253 0.03666
better than ACI and ICI codes. ICI code 0.81604 0.12305 0.07566
For the ANFIS model, the same as the ANN model, 100 data ACI code 0.7440 0.14516 0.08675
sets are used as training, and 23 data sets as testing data. After
training, all 100 data sets are used for testing the model to verify
the accuracy of the predicted values of shear strength. ANFIS
To show the superiority of proposed models, the RMSE, MAE
detects a function for mapping the input variables to the output
and R2 values for ANN and ANFIS models and ICI and ACI codes
layer. These predicted values are compared with observed data
to show the performance of the ANFIS model for the prediction are presented in Table 4. Comparison of the results indicates
of RC beam shear strength. Figure 6 shows the comparison that the ANN model with the MLP/BP algorithm has a higher
between observed and predicted values of the ANFIS model. value of R-square and a lower value of RMSE and MAE, so that
As can be seen from this figure, ANFIS has performed well in its performance is more accurate. Also, the performance of the
predicting the shear strength. Figure 7 shows the comparison ANN model is better than the ANFIS model and the two models
between the prediction of the shear strength of the training and perform a much better prediction, compared with ACI and ICI
testing data, as given by ICI, ACI codes and the ANFIS model. empirical codes.
246 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248

(a) Training set. (b) Testing set.

Figure 4: Plot of observed and predicted shear strength for ANN model.

(a) Training set. (b) Testing set.

Figure 5: Comparison between ICI, ACI codes and ANN model.

(a) Training set. (b) Testing set.

Figure 6: Plot of observed and predicted shear strength for ANFIS model.
 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248 247

(a) Training set. (b) Testing set.

Figure 7: Comparison between ICI, ACI codes and ANFIS model.

4. Conclusions [14] Bai, J., Wild, S., Ware, J.A. and Sabir, B.B. ‘‘Using neural networks to predict
workability of concrete incorporating Meta kaolin and fly ash’’, Adv. Eng.
Softw., 34, pp. 663–669 (2003).
The main goal of the present study is to design and develop [15] Hadi, M. ‘‘Neural networks applications in concrete structures’’, J. Comput.
ANN and ANFIS models for estimating RC beam shear strength. Struct., 81, pp. 373–381 (2003).
The performances of the models were evaluated and the re- [16] Oreta, A.W.C. and Kawashima, K. ‘‘Neural network modeling of concrete
compressive strength and strain of circular concrete columns’’, J. Struct.
sults were compared with empirical ICI and ACI concrete codes. Eng., ASCE, 29(4), pp. 554–561 (2003).
Several empirical formulas were used for computing the shear [17] Lee, S.C. ‘‘Prediction of concrete strength using artificial neural networks’’,
strength of RC beams in concrete codes. The models were ap- Eng. Struct., 25, pp. 849–857 (2003).
plied in the prediction of the shear resistance strength of RC [18] Kim, J.I., Kim, D.K., Feng, M.Q. and Yazdani, F. ‘‘Application of neural
networks for estimation of concrete strength’’, J. Mater. Civ. Eng., 16(3),
beams. Results show that the ANN model with the MLP/BP al- pp. 257–264 (2004).
gorithm provided a better prediction of shear strength than the [19] Oreta, A.W.C. ‘‘Simulating size effect on shear strength of RC beams
ANFIS model. Also, ANN and ANFIS models are more accurate without stirrups using neural networks’’, Eng. Struct., 25(5), pp. 681–691
(2004).
than empirical ACI and ICI codes. In addition, the predictions [20] Mansour, M.Y., Dicleli, M., Lee, J.Y. and Zhang, J. ‘‘Predicting the shear
of ANN model were distributed around experimental results, strength of reinforced concrete beams using artificial neural networks’’,
while ICI and ACI were more scattered from experimental re- Eng. Struct., 26, pp. 781–799 (2004).
sults, indicating that they predominately under-estimate the [21] Cladera, A. and Mari, A.R. ‘‘Shear design procedure for reinforced normal
and high-strength concrete beams using artificial neural networks. Part
shear stength. II: beams with stirrups’’, Eng. Struct., 26(7), pp. 927–936 (2004).
[22] Abdalla, J.A., Elsanosi, A. and Abdelwahab, A. ‘‘Modeling and simulation of
shear resistance of R/C beams using artificial neural network’’, J. Franklin
References Inst., 344(5), pp. 741–756 (2007).
[23] Kim, Y.S. and Kim, B.T. ‘‘Prediction of relative crest settlement of concrete-
[1] Buildings code requirements for structural concrete (ACI 318-08), Amer- faced rockfill dams analyzed using an artificial neural network model’’,
ican Concrete Institute (ACI), Detroit (MI) (2008). Comput. Geotech., 35, pp. 313–322 (2008).
[2] Management and Planning Organization of I.R. Iran Iranian concrete code [24] Caglar, N., Elmas, M., Yaman, Z.D. and Saribiyik, M. ‘‘Neural network in
of practice, ABA, Publication No. 120, Iran (2001). 3-dimensional dynamic analysis of reinforced concrete buildings’’, Constr.
[3] Kani, G.N.J. ‘‘The riddle of shear failure and its solution’’, ACI J., 61(4), Build. Mater., 22, pp. 788–800 (2008).
pp. 441–466 (1964). [25] Arslan, M.H. ‘‘Predicting of tensional strength of RC beams by using
[4] Kani, G.N.J. ‘‘How safe are our large RC beams’’, ACI J., 64(3), pp. 128–141 different artificial neural network algorithms and building codes’’, Adv.
(1967). Eng. Softw., 41(7–8), pp. 946–955 (2010).
[5] Krefeld, W.J. and Thurston, C.W. ‘‘Contribution of longitudinal steel to [26] Erdem, H. ‘‘Prediction of the moment capacity of reinforced concrete
shear resistance of reinforced concrete beams’’, ACI J., 63(3), pp. 325–344 slabs in fire using artificial neural networks’’, Adv. Eng. Softw., 41(2),
(1966). pp. 270–276 (2010).
[6] Krefeld, W.J. and Thurston, C.W. ‘‘Studies of the shear and diagonal tension [27] Takagi, M.T. and Sugeno, M. ‘‘Fuzzy identification of systems and its
strength of simply supported RC beams’’, ACI J., 63(4), pp. 451–476 (1966). applications to modeling and control’’, IEEE Trans. Syst. Man Cybern., 15(1),
[7] Zsutty, T. ‘‘Shear strength prediction for separate categories of simple pp. 116–132 (1985).
beam tests’’, ACI J., 68(2), pp. 138–143 (1971). [28] Jang, J.S.R. ‘‘ANFIS: adaptive network-based fuzzy inference systems’’,
[8] Zsutty, T. ‘‘Beam shear strength prediction by analysis of existing data’’, IEEE Trans. Syst. Man Cybern., 23(3), pp. 665–685 (1993).
ACI J., 65(11), pp. 942–951 (1968). [29] Topçu, İ.B. and Saridemir, M. ‘‘Prediction of rubberized mortar properties
[9] Mphonde, A.G. and Frantz, G.C. ‘‘Shear tests of high and low strength using artificial neural network and fuzzy logic’’, J. Mater. Process. Technol.,
concrete beams without stirrups’’, ACI J., 81(4), pp. 350–357 (1984). 199(1–3), pp. 108–118 (2008).
[10] Kasperkiewicz, J., Racz, J. and Dubrawski, A. ‘‘HPC strength prediction us- [30] Bilgehan, M. ‘‘Comparison of ANFIS and NN models—with a study in critical
ing artificial neural networks’’, J. Comput. Civil Eng., ASCE, 9(4), pp. 279–284 buckling load estimation’’, Appl. Soft Comput., 11(4), pp. 3779–3791 (2011).
(1995). [31] Anderson, D. and McNeill, G. Artificial Neural Networks Technology, Rome
[11] Yeh, I.C. ‘‘Modeling concrete strength with augment-neuron networks’’, J. Laboratory, NY, Rome, NY (1992).
Mater. Civ. Eng., 10(4), pp. 263–268 (1998). [32] Rasa, E., Ketabchi, H. and Afshar, M.H. ‘‘Predicting density and compressive
[12] Guang, N.H. and Zong, W.J. ‘‘Prediction of compressive strength of concrete strength of concrete cement paste containing silica fume using artificial
by neural network’’, Cem. Concr. Res., 30, pp. 1245–1250 (2000). neural network’’, Sci. Iran., Trans. A: Civ. Eng., 16(1), pp. 33–42 (2009).
[13] Waszczyszyn, Z. and Ziemianski, L. ‘‘Neural networks in mechanics of [33] Kazeminezhad, M.H., Etemad-Shahidi, A. and Mousavi, S.J. ‘‘Application
structures and materials—new results and prospects of applications’’, J. of fuzzy inference systems in prediction of wave parameters’’, Ocean Eng.,
Comput. Struct., 79, pp. 2261–2276 (2001). 32(14), pp. 1709–1725 (2005).
248 J. Amani, R. Moeini / Scientia Iranica, Transactions A: Civil Engineering 19 (2012) 242–248

[34] Bentz, E.C. ‘‘Sectional analysis of reinforced concrete members’’, He is author/coauthor of 5 research articles that have been published in
Ph.D. Thesis, Department of Civil Engineering, University of Toronto international journals and 9 refereed conference papers. His research interests
(2000). include: numerical modeling and refinement procedures focused on meshless
[35] Bohigas, A.C. ‘‘Shear design of reinforced high-strength concrete beams’’, and finite element methods and their applications in elasticity problems,
Ph.D. Thesis, Department of Construction Engineering, Polytechnic and crack growth and fracture mechanics in concrete structures. He also has
University of Catalonia, Barcelona (2002). experience in the application of artificial neural networks, adaptive neuro-
[36] Masters, Timonly, Practical Neural Network Recipes in C + +, Morgan fuzzy inference systems and the cellular automata approach in structural and
Kaufmann (1993). hydraulic problems.

Ramtin Moeini was born in Isfahan, Iran, in 1981. He received his B.S. degree
Jafar Amani was born in Ardabil, Iran, in 1985. He received his B.S. degree from the Civil Engineering Department at Isfahan University of Technology
from the Civil Engineering Department at the University of Mohaghegh Ardabili (IUT) in 2002 and his M.S. degree, honors, from the Civil Engineering
(UMA), Iran, in 2007, and his M.S. degree, honors, in Structural and Hydraulic Department of Iran University of Science and Technology (IUST) in 2007, where
Engineering, in 2009, from Iran University of Science and Technology (IUST), he is currently a Ph.D. student of Civil Engineering. He is also a member of the
where he is currently a Ph.D. degree student of Civil Engineering. He is also academic staff at Alaodoleh Semnani Institute of Higher Education (ASIHE). He is
a member of the academic staff at Alaodoleh Semnani Institute of Higher author/coauthor of 9 research articles in international journals and 7 conference
Education (ASIHE). papers.