Engineering, 2010, 2, 559-572
doi:10.4236/eng.2010.28072 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Experimental Comparative and Numerical Predictive
Studies on Strength Evaluation of Cement Types: Effect of
Specimen Shape and Type of Sand
Hossam Hodhod1, Mostafa A. M. Abdeen2
1Department of Structural Engineering, Faculty of Engineering, Cairo University, Giza, Egypt
2Department of Engineering Mathematics & Physics, Faculty of Engineering, Cairo University, Giza, Egypt
Received May 6, 2010; revised July 23, 2010; accepted July 25, 2010
Abstract
Quality of cement is evaluated via group of tests. The most important, and close to understanding, is the
compressive strength test. Recently, Egyptian standards adopted the European standards EN-196 and
EN-197 for specifying and evaluating quality of cements. This was motivated by the large European invest-
ments in the local production of cement. The current study represents a comparative investigation, experi-
mental and numerical, of the effect of different parameters on evaluation of compressive strength. Main pa-
rameters are shape of specimens and type of sand used for producing tested mortars. Three sets of specimens
were made for ten types of cements. First set were 70.6 mm cubes molded according to old standards using
single sized sand. Second group were prisms molded from standard sand (CEN sand) according to the recent
standards. Third group were prisms molded from local sand sieved and regenerated to simulate same grading
of CEN sand. All specimens were cured according to relevant standards and tested at different ages (2,3,7,10
and 28 days). Results show that CEM-I Type of cement does not fulfill, in all of its grades, the strength re-
quirements of Ordinary Portland cement OPC specified in old standards. Also, the use of simulated CEN
sand from local source gives strengths lower than those obtained using standard certified CEN sand. A lim-
ited number of tests were made on concrete specimens from two most common CEM-I types to investigate
effect on concrete strength and results were also reported. Numerical investigation of the effect of specimen
shape and type of sand on evaluation of compressive strength of mortar specimens, presented in the current
study, applies one of the artificial intelligence techniques to simulate and predict the strength behavior at
different ages. The Artificial Neural Network (ANN) technique is introduced in the current study to simulate
the strength behavior using the available experimental data and predict the strength value at any age in the
range of the experiments or in the future. The results of the numerical study showed that the ANN method
with less effort was very efficiently capable of simulating the effect of specimen shape and type of sand on
the strength behavior of tested mortar with different cement types.
Keywords: Cement Type, Sand Type, Mortar Specimen, Strength, Modeling, Artificial Neural Network
1. Introduction
For decades, engineers used to apply cement based on
certain classification [1-3]. This classification refers to
its composition and consequently relevant properties.
Among these properties, strength was the main target of
using a specific type of cement. Ordinary Portland ce-
ment (OPC), sulphate resisting cement (SRC) and white
cement share almost same values for compressive
strength at different ages. One type: namely rapid hard-
ening cement had the higher early strength than others.
Recently, end of the year 2006, the Egyptian standards [4]
decided to adopt the European standards EN196 & EN
197 [5] for producing, specifying and testing almost all
types of cements. The new standard took the designation
ES4756 and included all types of cements but SRC. The
new standard included a drastic change in specifying
cement types, and appeared ambiguous in many aspects
since it added ranks and rate of hardening for the same
composition of cement. This raised many questions

H. HODHOD ET AL. 560
about the actual composition of cement and its properties.
Also, questions were raised about the properties of cem-
titious mixes for which cement is used and whether the
correlations between properties and type of cement will
remain the same or not. Besides, methods of testing ce-
ment to evaluate its compressive strength were changed
from using cubic specimens (70.6 mm side length) [6] to
part of prism (with 40 mm square cross section). More-
over, the standard dictated the use of specific type of
sand, which is not available locally, for making mortar
specimens. This sand must be procured from certified
suppliers and is called CEN sand. Such a condition
raised a question about the role of this sand in hydration
process and strength development too. All these ques-
tions motivate the need for research to clarify nature of
new cement types and declare their properties and effect
on properties of cementitious mixes. One local attempt [7]
was made and yielded that new standards are efficient.
However, the contradictions between test results of the
study and the known size effect rule urge the need for
more investigation.
Since the experimental work needs a lot of effort, time
and money, which is quite clear from the literature men-
tioned previously, the need for utilizing new methodolo-
gies and techniques to reduce this effort, save time and
money (and at the same time preserving high accuracy)
is urged. Artificial intelligence has proven its capability
in simulating and predicting the behavior of the different
physical phenomena in most of the engineering fields.
Artificial Neural Network (ANN) is one of the artificial
intelligence techniques that have been incorporated in
various scientific disciplines. Solomatine and Toorres [8]
presented a study of using ANN in the optimization loop
for the hydrodynamic modeling of reservoir operation in
Venezuela. Kheireldin [9] presented a study to model the
hydraulic characteristics of severe contractions in open
channels using ANN technique. The successful results of
his study showed the applicability of using the ANN ap-
proach in determining relationship between different
parameters with multiple input/output problems. Abdeen
[10] developed neural network model for predicting flow
characteristics in irregular open channels. The developed
model proved that ANN technique was capable with
small computational effort and high accuracy of predict-
ing flow depths and average flow velocities along the
channel reach when the geometrical properties of the
channel cross sections were measured or vice versa. Al-
lam [11] used the artificial intelligence technique to pre-
dict the effect of tunnel construction on nearby buildings
which is the main factor in choosing the tunnel route.
Allam, in her thesis, predicted the maximum and mini-
mum differential settlement necessary precautionary
measures. Park and Azmathullah et al. [12] presented a
study for estimating the scour characteristics downstream
of a ski-jump bucket using Neural Networks (NN). Ab-
deen [13] presented a study for the development of ANN
models to simulate flow behavior in open channel in-
fested by submerged aquatic weeds. Mohamed [14] pro-
posed an artificial neural network for the selection of
optimal lateral load-resisting system for multi-story steel
frames. Mohamed, in her master thesis, proposed the
neural network to reduce the computing time consumed
in the design iterations. Abdeen [15] utilized ANN tech-
nique for the development of various models to simulate
the impacts of different submerged weeds' densities, dif-
ferent flow discharges, and different distributaries opera-
tion scheduling on the water surface profile in an ex-
perimental main open channel that supplies water to dif-
ferent distributaries.
2. Problem Description
To study the effect of specimens shape and types of sand
used for producing tested mortars on evaluation of com-
pressive as well as flexural tensile strengths, experimen-
tal and numerical techniques will be presented in this
study. The experimental program and its results will be
described in detail in the following sections. After that,
numerical approach will be discussed to show the effi-
ciency of numerical techniques. The numerical models
presented in this study utilized Artificial Neural Network
technique (ANN) using the data of experiments and then
can predict the strength value in the range of the experi-
ment or in the future.
3. Experimental Program
The experimental program focuses on evaluating com-
pressive strength of mortar made from new cement types.
Ten types of cements with different grades and rate of
hardening were procured from local market in Egypt.
Compressive strength was evaluated for each type using
the cubic specimens (70.6 mm side length) and using the
testing of part of prism (40*40*160 mm). The last me-
thod was employed twice. First with local sand following
the same grade specified in ES4756 (and EN 196), and
second with certified CEN sand according to same stan-
dards. Specimens were tested at ages of 2, 3, 7, 10 and
28 days.
Concrete mixes with same proportions were cast from
different types of cements. Slump and compressive
strength were measured for each mix to investigate the
effect of type of cement on concrete properties. Com-
pressive strength was measured at 3, 7 and 28 days.
4. Materials and Specimens
Constituents for mortar and concrete mixes were:
4.1. Water
Tap water was used for mixing and curing of all speci-

H. HODHOD ET AL. 561
mens.
4.2. Cement
Ten types of cement were used. All were supplied in
bags carrying the symbols of both ES4756 and EN-197.
They were all produced locally in Egypt by different
Cement Companies. The ten types covered CEM I (ordi-
nary Portland cement) with different grades and rates of
hardening. The types also included white cement, sul-
phate resisting cement (SRC) and CEM II type cements.
Table 1 shows the investigated types of cement.
4.3. Sand
Two types of sand were used for the current study: CEN
sand that was imported from France in sealed transparent
bags (Figure 1), and local siliceous sand. Local siliceous
sand was used in its natural grading (Figure 2) for cast-
ing concrete. This sand was sieved to get the single size
sand required for testing mortar cubes according to old
ES (still effective as part of local code of practice ECP
203/2001 app.3. The local sand was also used to regener-
ate the CEN sand by collecting different sizes in the per-
centages specified in EN-196. The grading of this regen-
erated sand, and limits of CEN sand, are shown in Fig-
ure 3.
4.4. Gravel
Local siliceous gravel was used for casting concrete spe-
cimens. Gravel has a maximum nominal size of 20 mm.
4.5. Specimens (Cubes and Prisms)
Standard cubes with 70.6 mm side were used for evaluat-
ing compressive strength of mortar with mix proportions
of water: cement: sand = 0.4:1:3 by weight. Sand was
0.6–0.85 mm local sand. Constituents were mixed manu-
ally. Steel molds were used for casting. The other two
sets of specimens were prisms (40*40*160 mm) cast
from mixtures with proportions of water: cement: sand =
0.5:1:3 by weight.
For one set, standard CEN sand was used for casting.
For the other set, regenerated local sand with grading
Figure 1. Bags of CEN sand.
0
20
40
60
80
100
0.01 0.1110
Percentage Passing
Sieve Opening Size (mm)

Figure 2. Grading of local sand used in concrete mixes.
Table 1. Investigated types of cement.
Strength Evaluation (on Mortar)
Flexure Compn Flexure Compn
No Type of Cement Manufacturer Cube Compn.Local sand(*) CEN Sand
1 CEM I-52.5N SINAI     
2 CEM I-42.5N
(SRC)(**) El-MASRIYA   
3 CEM I-42.5N (SRC) HELWAN     
4 CEM I-42.5N El-MASRIYA     
5 CEM I-42.5R HELWAN     
6 CEM I-42.5N
(White) HELWAN   
7 CEM I-32.5R ELKAUMIYA (NCC)     
8 CEM I-32.5R (SRC) ASSIUT (CEMEX)     
9 CEM II-B-S 32.5N HELWAN   
10 CEM II-B-L -32.5N ELKAUMIYA (NCC)   
(*) Sand having a grading similar to CEN sand.
(**) This cement will be denoted in figures as SRC-1.

H. HODHOD ET AL. 562
0
20
40
60
80
100
0.01 0.1110
Sieve Opening Size - mm
Percentage Passing
CEN Upper Limit
CEN Lower Limit
Regenerated Sand
Figure 3. Grading limits of CEN sand and grading of locally
regenerated sand.
similar to CEN was used. Constituents were mixed me-
chanically using 5 liter mixer. Steel molds were used for
casting. For all sets, specimens were compacted using
vibrator and left covered with impervious sheet for 24
hours. Then specimens were demolded and immersed in
water till day of testing. Concrete cubes (with 150 mm
side length) were cast to evaluate concrete strength. Mix
proportions were water: cement: sand: gravel = 0.6:1:
1.5:3.0. Constituents were mixed mechanically using 140
liter tilting type mixer. Dry materials were mixed first for
about one minute. Then, water was added gradually and
mixing continued till uniform mix was obtained. Con-
crete was cast in steel molds and compacted using a vi-
brating table. Specimens were covered with plastic
sheets for 24 hours. Then molds were removed and
specimens were wet cured till age of testing.
5. Test Results
The test results are explained in the following sections.
5.1. Cement Setting Time
Initial and final setting times measured for different
types of cement are shown in Figure 4. One can see that
the initial setting time ranges from 70 to 120 min. Final
setting time ranges from 140 to 240 min. Generally, final
setting time is almost double the initial setting time. The
least setting time was recorded for CEM I 52.5N and the
longest setting time was recorded for CEM II B-S-32.5N.
Setting time increases as cement grade decreases, and
SRC shows less setting time for same grade. Rate of set-
ting (expressed by N or R after grade) does not seem to
affect setting time results. Recorded values of initial and
final setting times comply with limits of ES 4756 and
EN-197.
5.2. Mortar Compressive Strength
Compressive strength measured for all specimens and
Figure 4. Setting time of different types of cement.
types of cements are plotted versus time in Figure 5.
One can see that cube specimens specified in old stan-
dards give strength lower than part of prisms specified in
the current standards. Large size of cubes helps reducing
its strength as the grading of the single sized sand does.
However, the low w/c ratio is supposed to help increas-
ing the strength of cubes. This indicates that the effect of
size and confinement of prism specimens and the grading
and type of CEN sand could compensate for the increase
of w/c ratio of the specimens.
There is a difference between results obtained for
CEN sand and regenerated sand composed by adding the
proper percentage of each size from local sand. CEN
sand always gives higher strength. This implies that it is
not only sand grading that contributes to the strength.
Shape of particles and probably some chemical charac-
teristics of sand may also contribute to this increase of
strength. These last two points need more research for
clarification.
It must be said that the strength of prisms does not ful-
fill the requirements of cement grade for all types. The
strength of prisms at 28 days reaches a percentage from
43 to 70% of corresponding grade of cement. Although
the compaction was not made using a jolting table (as
specified by current standard test method), this is not
expected to yield such big difference for all types of ce-
ment.
It is note-worthy that similar strength values were ob-
tained for all rapid setting cements regardless of their
grade. However, the normal setting CEM I 52.5N gave
the highest strength among all other cement mortars.
Strength factor, which is the ratio of 28 day strength to
the strength at a specific age, is plotted in Figures 6-8
for different specimens and different types of cement.
The small strength values for CEN sand prisms denote
the rapid strength development of strength with this sand.
However, for same sand, it seems that strength develop-
ment does not follow the indication R & N for Type of
cement, where smaller factors are observed for normal
setting cements. Similar trend was found for other types
of sand.

H. HODHOD ET AL. 563

Figure 5. Compressive strength of mortar specimens produced under different conditions, for different types of cement.

H. HODHOD ET AL. 564
Figure 6. Strength factor for mortar prisms produced from CEN sand (left: rapid setting cements, right: normal setting ce-
ments).
Figure 7. Strength factor for mortar prisms produced from regenerated sand (left: rapid setting cements, right: normal set-
ting cements).
Figure 8. Strength factor for mortar cubes (left: rapid setting cements, right: normal setting cements).
5.3. Mortar Tensile Strength
Flexural tensile strength was measured for prism speci-
mens since it is the first step in producing compression
specimens. Measured flexural strength for all types of
cements and for different sands are plotted in Figure 9.
One can observe the effect of CEN sand in increasing
strength of mortar. This effect confirms the above men-
tioned need for investigation of particle shape and chem-
ical reactivity of CEN sand.
One more finding can be found when tensile flexural
strength is plotted versus compressive strength at differ-
ent ages, as in Figure 10. It can be seen that there exists a
significant increase of tensile strength between 7 and 28
days. This could be observed for both types of sand. This
implies that the correlation between flexural strength and
compressive strength is significantly different at early
and later ages.
5.4. Concrete Slump
Concrete slump measuring results are shown in Figure
11 for all cement types. One can identify 3 main ranges
of slump: 0-40 mm, 40-80 mm, and 80-120 mm. First
low range of slump was recorded for rapid setting ce-
ments and 52.5 grade cement. Highest slump range was

H. HODHOD ET AL. 565
Figure 9. Flexural strength of mortar prisms produced different sand types, for different types of cement.

H. HODHOD ET AL. 566
Figure 10. Flexural strength vs. compressive strength (left: cen sand, right: regenerated sand).
Figure 11. Slump values for different types of cement.
observed for CEM II cements. The medium grade was
observed for the rest normal setting CEM I type of ce-
ment. Since the water consumption is related to rates of
hydration and heat evolution. One can conclude that
grade 52.5 has high rate of hydration. It is noteworthy
that CEM I 52.5R does not exist in local Egyptian mar-
ket. The high slump of CEM II cement mixes can be
correlated to low clinker content.
5.5. Concrete Compressive Strength
Measured values of concrete compressive strength are
plotted versus age, for all types of cement, in Figure 12.
Generally, one can see in Figure 12 that the effect of
cement grade can be distinguished in the limits where top
curve belongs to grade 52. N and bottom curve belongs
to grade 32.5 N. Curves for higher grades of cement are
shown in Figure 13 Left. One can see that, up to 7 days
all 42.5 grade cements show almost same strength. How-
ever, at later ages (28 days) the rapid setting type shows
higher strength than the normal setting ones. One can
also see that SRC cement show slightly higher strength
than similar 42.5 N cements. Curves for low grade ce-
ments are shown in Figure 13 Right. The effect of setting
rate can be identified between 32.5N and 32.5R cements.
Still SRC cement shows higher strength at 28 days.
Figure 14 shows strength development for rapid setting
and normal setting cements, respectively. For rapid set-
ting cements, there is no difference between early age
0
5
10
15
20
25
30
0102030
Age - days
Concrete Compressive Strength (MPa)
CEM I 52.5N
CEM I 42.5N-SRC
CEM I 42.5N
CEM I 42.5R
CEM I 42.5N white
CEM I 32.5R
CEM I 32.5R (SRC)
CEM II-B-S32.5N
CEM II B-L32.5N
Figure 12. Measured concrete compressive strength for
different types of cement.
strength of different grades of cement. At 28 days, SRC
of 32.5 grade yields same strength as 42.5 grade. For
normal setting cements, there is a clear distinction be-
tween strength of different grades at all ages. Strength
ratio at 28 days is almost proportional to grade of cement.
One can still observe that SRC show higher values of
strength than other CEM I cements of same grade.
6. Numerical Model Structure
Neural networks are models of biological neural struc-
tures. Briefly, the starting point for most networks is a
model neuron as shown in Figure 15. This neuron is con-
nected to multiple inputs and produces a single output.
Each input is modified by a weighting value (w). The
neuron will combine these weighted inputs with refer-
ence to a threshold value and an activation function, will
determine its output. This behavior follows closely the
real neurons work of the human’s brain. In the network
structure, the input layer is considered a distributor of the
signals from the external world while hidden layers are
considered to be feature detectors of such signals. On the
other hand, the output layer is considered as a collector
of the features detected and the producer of the response.
6.1. Neural Network Operation
It is quite important for the reader to understand how the

H. HODHOD ET AL. 567
0
5
10
15
20
25
30
0102030
Age - days
Concrete Compressive Strength (MPa)
CEM I 52.5N
CEM I 42.5N-SRC
CEM I 42.5N
CEM I 42.5R
CEM I 42.5N white
0
5
10
15
20
25
30
0102030
Age - days
Concrete Compressive Strength (MPa)
CEM I 32.5R
CEM I 32.5R-SRC
CEM II-B-S32.5N
CEM II B-L32.5N
Figure 13. Measured concrete compressive strength (left: high grades of cement, right: low grades of cement).
0
5
10
15
20
25
30
0102030
Age - days
Concrete Compressive Strength (MPa)
CEM I 42.5R
CEM I 32.5R
CEM I 32.5R-SRC
0
5
10
15
20
25
30
0102030
Age - days
Concrete Compressive Strength (MPa)
CEM I 52.5N
CEM I 42.5N-SRC
CEM I 42.5N
CEM I 42.5N white
CEM II-B-S32.5N
CEM II B-L32.5N
Figure 14. Measured concrete compressive strength (left: rapid setting cements, right: normal setting cements).
Figure 15. Typical picture of a model neuron that exists in
every neural network.
neural network operates to simulate different physical
problems. The output of ach neuron is a function of its
inputs (Xi). In more details, the output (Yj) of the jth neu-
ron in any layer is described by two sets of equations as
follows:


j
iij
UXw (1)
and


j
th jj
YFUt (2)
For every neuron, j, in a layer, each of the i inputs, Xi,
to that layer is multiplied by a previously established
weight, wij. These are all summed together, resulting in
the internal value of this operation, Uj. This value is then
biased by a previously established threshold value, tj, and
sent through an activation function, Fth. This activation
function can take several forms such as Step, Linear,
Sigmoid, Hyperbolic, and Gaussian functions. The Hy-
perbolic function, used in this study, is shaped exactly as
the Sigmoid one with the same mathematical representa-
tion, as in Equation (3), but it ranges from –1 to +1 rather
than from 0 to 1 as in the Sigmoid one.

1
1
x
fx e  (3)
The resulting output, Yj, is an input to the next layer or
it is a response of the neural network if it is the last layer.
In applying the Neural Network technique, in this study,
Neuralyst Software, Shin [16], was used.
6.2. Neural Network Training
The next step in neural network procedure is the training
operation. The main purpose of this operation is to tune
up the network to what it should produce as a response.
From the difference between the desired response and
the actual response, the error is determined and a portion
of it is back propagated through the network. At each
neuron in the network, the error is used to adjust the
weights and the threshold value of this neuron. Conse-
quently, the error in the network will be less for the same

H. HODHOD ET AL. 568
inputs at the next iteration. This corrective procedure is
applied continuously and repetitively for each set of in-
puts and corresponding set of outputs. This procedure
will decrease the individual or total error in the responses
to reach a desired tolerance.
Once the network reduces the total error to the satis-
factory limit, the training process may stop. The error
propagation in the network starts at the output layer with
the following equations:


'
ijijj i
wwLReX (4)
and,




1
j
jjjj
eY YdY 
(5)
where, wij is the corrected weight, w’
ij is the previous
weight value, LR is the learning rate, ej is the error term,
Xi is the ith input value, Yj is the ouput, and dj is the de-
sired output.
7. Simulation Models
To fully investigate numerically the effect of specimen
shape and type of sand on the strength behavior of tested
mortar with different cement types, seven models are
considered in this study. Two models for sitting time
(initial and final), model for cube compression strength,
two models for prism compression strength (CEN and
regenerated sand) and two models for flexural strength
(CEN and regenerated sand).
7.1. Neural network Design
To develop a neural network models to simulate the ef-
fect of specimen shape and type of sand on the strength
behavior of tested mortar, first input and output variables
have to be determined. What we have in the current
study, to be considered as an input variable, is the types
of cement used in the mortar specimen. So from the
name of cement type we have to create a certain numeric
characteristic values could be used as input variables in
the present models as shown in Table 2.
Table 3 is designed to summarize all neural network
key input variables and output for all the seven models
presented in the current study. Some abbreviations used
in Table 3 due to space limitation as follows:
Str.: Strength
Compn.: Compression
Flex.: Flexural
Table 2. Characteristic values for types of cement.
No Type of Cement I or II 32.5 or 42.5
or 52.5 N or RA or BS or LSRC or White Manufacturer
1 CEM I-52.5N 1 52.5 19 100 0 60 0
2 CEM I-42.5N
(SRC) (**) 1 42.5 19 100 0 67 1
3 CEM I-42.5N
(SRC) 1 42.5 19 100 0 67 2
4 CEM I-42.5N 1 42.5 19 100 0 60 0
5 CEM I-42.5R 1 42.5 23 100 0 60 0
6 CEM I-42.5N
(White) 1 42.5 19 100 0 50 0
7 CEM I-32.5R 1 32.5 23 100 0 60 0
8 CEM I-32.5R
(SRC) 1 32.5 23 100 0 67 0
9 CEM II-B-S-32.5N2 32.5 19 60 24 50 0
10 CEM II-B-L-32.5N2 32.5 19 60 12 50 0
Table 3. Key input variables and output for all ANN models.
Input Variables
Model I
or II
32.5 or
42.5 or
52.5
N or
R
A or
B
S or
L
SRC or
White Manufacturer Days Output
Initial Sitting Time        Initial Time
Final Sitting Time        Final Time
Cube Str.         Compn. Str.
Prism Str.
(CEN sand)     Compn. Str.
Prism Str.
(Regenerated)         Compn. Str.
Prism Flex. Str.
(CEN)     Flex. Ten. Str.
Prism Flex. Str.
(Regenerated)         Flex. Ten. Str.

H. HODHOD ET AL. 569
Ten.: Tensile
Several neural network architectures are designed and
tested for all simulation models investigated in this study
to finally determine the best network models to simulate,
very accurately, the effect of specimen shape and type of
sand on the strength behavior of tested mortar with dif-
ferent cement types based on minimizing the Root Mean
Square Error (RMS-Error). Figure 16 shows a schematic
diagram for a generic neural network. The training pro-
cedure for the developed ANN models, in the current
study, uses the experimental data presented in the previ-
ous sections of the current study. After the ANN models
are settled for all cases, prediction procedure takes place
to predict the compression as well as tensile strengths at
different age-days rather than those days measured in the
experiment (internal and after 28 days).
Table 4 shows the final neural network models for the
seven simulation models and their associate number of
neurons. The input and output layers represent the key
input and output variables described previously for each
simulation model.
The parameters of the various network models devel-
oped in the current study for the different simulation
models are presented in Table 5. These parameters can
be described with their tasks as follows:
Learning Rate (LR): determines the magnitude of the
correction term applied to adjust each neuron’s weights
during training process = 1 in the current study.
Momentum (M): determines the “life time” of a cor-
rection term as the training process takes place = 0.9 in
the current study.
Training Tolerance (TRT): defines the percentage
error allowed in comparing the neural network output to
the target value to be scored as “Right” during the train-
ing process = 0.001 in the current study.
Testing Tolerance (TST): it is similar to Training
Tolerance, but it is applied to the neural network outputs
and the target values only for the test data = 0.003 in the
current study.
Input Noise (IN): provides a slight random variation
to each input value for every training epoch = 0 in the
current study.
Figure 16. General schematic diagram of a simple generic neural network.
Table 4. The developed neural network models.
No. of Neurons in each Layer
Simulation
Model No. of Layers Input Layer First Hidden Second
Hidden Third Hidden Output Layer
Initial Sitting
Time 4 7 5 3 - 1
Final Sitting
Time 4 7 5 3 - 1
Cube Str. 5 8 6 4 2 1
Prism Str.
(CEN sand) 5 4 4 3 2 1
Prism Str.
(Regenerated) 5 8 6 4 2 1
Prism Flex. Str.
(CEN) 5 4 4 3 2 1
Prism Flex. Str.
(Regenerated) 5 8 6 4 2 1
Input # 1
Input # 2
Output # 1
Output # 2
Hidden layer
3 neurons
Hidden layer
3 neurons

H. HODHOD ET AL. 570
Table 5. Parameters used in the developed neural network models.
Simulation
Parameter
Initial
Sitting
Time
Final
Sitting
Time
Cube
Str.
Compn.
Str. (CEN)
Compn. Str. (Re-
generated)
Flex. Str.
(CEN)
Flex. Str. (Re-
generated)
Training Ep-
ochs 1146 4985 672361 301098 179853 315475 505672
MPRE 0.067 0.034 1.175 0.174 0.281 1.512 0.321
RMS-Error 0.0005 0.0005 0.0008 0.0004 0.0003 0.0016 0.0002
Function Gain (FG): allows a change in the scaling
or width of the selected function = 1 in the current study.
Scaling Margin (SM): adds additional headroom, as a
percentage of range, to the rescaling computations used
by Neuralyst Software, Shin (1994), in preparing data for
the neural network or interpreting data from the neural
network = 0.1 in the current study.
Training Epochs: number of trails to achieve the pre-
sent accuracy.
Percentage Relative Error (PRR): percentage rela-
tive error between the numerical results and actual meas-
ured value and is computed according to Equation (6) as
follows:
PRE = (Absolute Value (ANN_PR –AMV)/AMV)*
100
(6)
where:
ANN_PR: Predicted results using the developed ANN
model
AMV: Actual Measured Value
MPRE: Maximum percentage relative error during the
model results for the training step.
8. Results and Discussions
Numerical results using ANN technique will be pre-
sented in this section for all the seven models. Due to
space limitation in the present paper the numerical re-
sults of one type of cement (CEM I 52.5N) will be pre-
sented to show the simulation and prediction powers of
ANN technique for compressive as well as tensile
strengths.
8.1. Sitting Time
For the sitting time models (initial and final), Table 6
presents the ANN results with experimental ones. One
can see from this table that ANN technique can simulate
very efficient the experiment results for different types of
cements for mortar specimens.
8.2. Mortar Compressive Strength
Three ANN models are developed to simulate and pre-
dict the effect of specimen shape and type of sand on
evaluating the compressive strength of mortar specimens
for all the types of cement presented in the current study
at different ages. Figures 17 and 18 show the ANN
results and experimental ones for compressive strength
(cube and prism specimens) for one type of cement at the
ages of experiment (2,3,7,10,28 days) and then predict
the behavior at 14 days (internally) and after 28 days up
to 42 days (externally). From these figures, it is very
clear that ANN technique succeeded very well to simu-
late and predict the compressive strength behavior at
different ages for different specimens.
8.3. Mortar Tensile Strength
Another two ANN models are developed to simulate and
predict the flexural tensile strength for prism specimen
for two types of sand (CEN and Regenerated) at different
ages. Figure 19 presents the numerical results and ex-
perimental ones at the ages of experiments (2,3,7,10,28
days). One can observe that ANN technique can simulate
the tensile behavior and then predict the strength at ages
different than the ages of experiment (before and after 28
days) very successfully.
9. Conclusions
Based on the experimental investigation conducted in the
course of the current research, the following can be con-
cluded:
1) There is an inverse proportion between setting time
and cement grade, and a direct proportion between grade
and water requirement for standard consistency.
2) Applying old cement standards, for testing and
evaluating mortar compressive strength of cement mor-
tar, results in rejection of new cement types. Using of
CEMI52.5N
0.0
5.0
10.0
15.0
20.0
25.0
0 1020304050
Age - days
Compressive Strength- Mp
a
Expe rime nt
ANN Training
ANN Prediction
Figure 17. Cube compressive strength.

H. HODHOD ET AL. 571
Table 6. Sitting time models results.
Simulation Model No. 1 No. 2 No. 3 No. 4 No. 5 No. 6 No. 7 No. 8 No. 9 No. 10
Exp. 70.0 85.0 75.0 100.0 90.0 80.0 110.0 80.0 120.0 100.0
Initial ANN 69.9 84.9 75.0 100.0 89.9 79.9 109.9 80.0 119.9 100.0
Exp. 140.0 140.0 180.0 210.0 200.0 170.0 225.0 185.0 240.0 200.0
Final ANN 140.0 140.0 180.0 210.0 200.0 170.0 225.0 185.0 240.0 200.0
CEMI52.5N
0
5
10
15
20
25
30
0 1020304050
Age - days
Compressive Strength- Mpa
Ex periment
ANN Training
ANN Prediction
CEMI52.5N
0.0
5.0
10.0
15.0
20.0
25.0
30.0
01020304050
Age - days
Compressive Strength- Mpa
Experime nt
ANN Training
ANN Prediction
Figure 18. Prism compressive strength (left: cen sand, right: regenerated sand).
CEMI52.5N
0
4
8
12
16
01020304050
Age - days
Flexural Strength - M
P
Expe riment
ANN Training
ANN Prediction
CEMI52.5N
0
4
8
12
16
0 1020304050
Age - days
Flexural Strength - M
P
Expe riment
ANN Training
ANN Prediction
Figure 19. Prism flexural strength (left: cen sand, right: regenerated sand).
jolting table for compaction is essential for obtaining
successful test results according to new standards (EN
196 and ES 4756).
3) Standard CEN sand cannot be regenerated locally
based only on its grading. Further investigation is re-
quired to get its other properties like particle shape and
chemical reactivity. There is some evidence on having
early strength development when CEN sand is used in
mortar.
4) Sulphate resisting cements show higher strength
than CEM I cements of same grade, for both mortar and
concrete mixtures.
5) There is some evidence that locally available ce-
ments do not follow the rate of strength development
denoted on packs.
6) For normal setting cements (N coded) there is a
clear distinction between concrete strength obtained for
specific cement grade. However, this could not be seen
for rapid setting types (R coded).
7) Correlation between flexural tensile and compres-
sive strength of mortar differs significantly between early
and later ages.
Based on the results of implementing the ANN tech-
nique in this study, the following can be concluded:
1) The developed ANN models presented in this study
are very successful in simulating the effect of specimen
shape and type of sand on the behavior of mortar speci-
mens (initial and sitting times, compressive strength,
flexural tensile strength) for different types of cement.
2) The presented ANN models are very efficiently ca-
pable of predicting the strength behavior at different ages
rather than the ages of the experimental results (in the

H. HODHOD ET AL. 572
range of the experiment or in the future).
10. Acknowledgements
The Authors would like to express their gratitude to-
wards Prof. Dr. Farouk El-Hakim of 15th May institute
for Civil and Arch. Engineering, and undergraduate stu-
dents (4th year– civil) for the help they provided during
the experimental part of this research.
11. References
[1] A. M. Neville, “Properties of Concrete,” John Wiley &
Sons, New York, 1997.
[2] S. H. Kosmatka, et al., “Design and Control of Concrete
Mixtures,” 5th Edition, Cement Association of Canada,
Ottawa, 2002.
[3] Egyptian Standards (ES) 372, “Standards of Ordinary and
Rapid Hardening Portland Cements,” Egyptian General
Authority for Standards, Cairo, 1991.
[4] Egyptian Standards (ES) 5476, “Standards of Cements,”
Egyptian General Authority for Standards, Cairo, 2007
[5] EN 197, “CEMENT: Part 1: Composition, Specifications
and Conformity Criteria for Common Cements,” 2004.
[6] Egyptian Code of Practice 203, “Basics of Design and
regulations of Construction of Reinforced Concrete
Structures: Appendix III, Guide for Testing of Concrete
Materials,” Egyptian Ministry of Housing, Egypt, 2001.
[7] K. M. Yosri, “Properties of New Cements Produced in
Egypt as per ES 4756/2005,” HBRC Journal, Vol. 3, No.
3, 2007, pp. 23-33.
[8] D. Solomatine, and L. Toorres, “Neural Network Ap-
proximation of a Hydrodynamic Model in Optimizing
Reservoir Operation,” Proceedings of the 2nd Interna-
tional Conference on Hydroinformatics, Zurich, 1996, pp.
201-206.
[9] K. A. Kheireldin, “Neural Network Application for Mod-
eling Hydraulic Characteristics of Severe Contraction,”
Proceedings of the 3rd Internetional Conference, Hydro-
informatics, Copenhagen, 24-26 August 1998, pp. 41-48.
[10] M. A. M. Abdeen, “Neural Network Model for predicting
Flow Characteristics in Irregular Open Channel,” Scien-
tific Journal, Faculty of Engineering-Alexandria Univer-
sity, Vol. 40, No. 4, 2001, pp. 539-546.
[11] B. S. M. Allam, “Artificial Intelligence Based Predictions
of Precautionary Measures for Building Adjacent to Tun-
nel Rout during Tunneling Process,” Ph.D. Thesis, Fac-
ulty of Engineering, Cairo University, Giza, 2005.
[12] H. Azmathullah, M. C. Deo and P. B. Deolalikar, “Neural
Networks for Estimation of Scour Downstream of a
Ski-Jump Bucket,” Journal of Hydrologic Engineering,
ASCE, Vol. 131, No. 10, 2005, pp. 898-908.
[13] M. A. M. Abdeen, “Development of Artificial Neural
Network Model for Simulating the Flow Behavior in
Open Channel Infested by Submerged Aquatic Weeds,”
Journal of Mechanical Science and Technology, KSME
International Journal, Vol. 20, No. 10, 2006, pp. 1691-
1879.
[14] M. A. M. Mohamed, “Selection of Optimum Lateral
Load-Resisting System Using Artificial Neural Net-
works,” M.Sc. Thesis, Faculty of Engineering, Cairo
University, Giza, 2006.
[15] M. A. M. Abdeen, “Predicting the Impact of Vegetations
in Open Channels with Different Distributaries Opera-
tions on Water Surface Profile using Artificial Neural
Networks,” Journal of Mechanical Science and Technol-
ogy, KSME International Journal, Vol. 22, 2008, pp.
1830-1842.
[16] Y. Shin, “NeuralystTM User’s Guide—Neural Network
Technology for Microsoft Excel,” Cheshire Engineering
Corporation Publisher, 1994.