1. No category

2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
Studying the Effect of Geometry and Type of Soil on the
Stability of Clover -Leaf Cofferdam
Dr. Raad Hoobi Irzooki, and Marwa Kaddori Majeed
The resisting forces and moments against the sliding and
overturning vary in magnitude from soil to soil depending on
the unit weight, the coefficient of friction of the soil, Young’s
Modulus of elasticity, poison’s ratio, and cohesion [1].
Cellular cofferdams are usually classified according to the
configuration and arrangements of the cells. As shown in Fig.
1, the three basic types of cellular cofferdams are:
Abstract- The stability of glover leaf coffer dam in two soil cases
(dry and wet) and two positions (neutral and longitudinal), with
nonlinear finite element analysis has been used in this study to
predict the load deflection behavior of cellular cell cofferdam under
lateral load. Series of laboratory tests have been carried out on one
(single) glover-leaf cofferdam cells of different width to height ratio
(b/H) (0.75, 0.85, 1.0) with three types of soil fill (sand passing sieve
No.4, river sand, subbase).
In finite element analysis, AN SYS(version 12.1) computer
program with combination of CivilFEM was used, eight-node solid
element SOLID 45 has been used to model filling with river sand
soil and the same element for steel sheet pile and foundation, by
using glue technique to model of steel sheet pile of cofferdam, the
load applied on the one third of the cell cofferdam height and full
Newton-Raphson method is used for the nonlinear solution algorithm
and the results are compared with experimental data.
The results expressed that the cells fill with wet subbase were
more stable against sliding and overturning at different (b/H) ratio
and the cells that put on the longitudinal direction was more stable
against sliding and overturning at different (b/H) ratio than the cells
with neutral direction. Also, the results obtained using the finite
element models show a good agreement with the experimental data.
The difference between the numerical ultimate loads and the
corresponding experimental ultimate loads is in the range between
(0-6.9)%.
A. Circular Cells
This type consists of a series of complete circular cells
connected by shorter arcs; these generally intercept the cells at
a point making an angle, (α), of 30 or 45 degrees with the
longitudinal axis of the cofferdam, as shown in Fig. 1-a.
B. Diaphragm Cells
These cells are comprised of a series of circular arcs
connected by 120 degree intersection pieces or cross walls
(diaphragm). The radius of the arc is often made equal to the
cell width so that these have equal tension in the arc and the
diaphragm, as shown in Fig. 1-b.
C. Cloverleaf Cells
This type of cell consists of four arc walls, within each of
the four quadrants, formed by two straight diaphragm walls
normal to each other, and intersecting at the center of the cell.
Adjacent cells are connected by short arc walls and are
proportioned so that the intersection of arcs and diaphragms
can form three angles of 120 degrees, as shown in Fig. 1-c.
Index Terms— ANN, ANSYS, Glover-leaf cofferdam, Stability.
I. INTRODUCTION
Cellular cofferdams are constructed of steel sheet piling
and used primarily as water-retaining structures. They depend
for stability on the interaction of the soil used to fill the cells
of these cofferdams and the steel sheet piling. Both materials
in combination provide a satisfactory means to develop a dry
work area in water-covered sites such as ocean or lakefront or
river area construction projects.
The purpose of the cofferdam is to retain a hydrostatic
head of water as well as the dynamic forces due to currents
and waves, ice forces, seismic loads and accidental loads or to
provide a lateral support to the mass of soil behind it.
However, the cofferdam is subjected to unbalanced lateral
forces acting at different heights. These unbalanced forces
will tend to produce a resultant moment which tends to
overturn the cofferdam or to produce a resultant force which
tends to slide the cofferdam on its base.
Dr. Raad Hoobi Irzooki, and Marwa Kaddori Majeed, are with
University - Civil Engineering Department, Iraq
http://dx.doi.org/10.15242/IAE.IAE0315413
(a) Circular cells
(b) Diaphragm cells
(c) Cloverleaf cells
Fig. 1. Cellular cofferdams; [TVA, (2003)]
Tikrit
60
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
The design analysis and stability of cofferdams,
especially circular type, was studied analytically and
experimentally by a number of researchers as follow:
Farrokh and Robert, (1972)[2] presented the design of
circular type cellular cofferdams formulated as a nonlinear
optimization model that takes explicit account of relevant
economic and technologic aspects. The objective is to
minimization of total expected cost. The geometric
programming approach provides important design insights by
yielding the proportions of the total cost to be assigned to the
cost components, fill material, sheet piling and flooding in an
optimal design.
Schroeder et al., (1977)[3] performed investigation on 12cell wharf at Terminal No.4 along the Willamette River in
Portland, Oregon. Individual cells are 19.74m in diameter,
spaced 25.74m center to center, a freestanding height of
20.1m, and connecting arcs which have a radius of 4.32m. He
was found that the maximum interlock force in the cell was
also near the dredge line and the lateral earth pressure values
recommended by Terzaghi are adequate for design[4].
Sorota and Kinner, (1981)[5] presented a description of
design of a steel sheet pile cellular cofferdam that was
required for construction of a graving dry dock. The
cofferdam was constructed approximately 168m offshore
within the Hood Canal of Washington State and was required
to retain 24m of water after basin dewatering. Items discussed
include the need for two pumped dewatering systems, the
need for vibratory probe compaction of the cell fills, site soil
conditions, dredging, and steel sheet pile corrosion protection.
Jahren, (1990)[6] stated that the sheet-pile cellular structure
is constructed by arranging straight web sheet piles in a
cylindrical shape and filling the enclosed volume with soil.
Further analysis indicates that the results are sensitive to
changes in several of the input parameters, but that the basic
findings are unchanged within the range of changes in the
sensitivity analysis.
AL-shamkey, (1992)[7] studied the analysis of stability of
circular cell in order to know the behavior of this cell under
the lateral load to back fill loads inside the cell. The
conclusion that obtained was the maximum hoop tension
sheet occurs in the same zone at the level (1/6 ) from cell
height.
Mohammod et al., (2001)[8] studied the behavior of
double sheet pile wall cofferdam on sandy soil subjected to
high level of water. Test results imply that: (i) the shear
deformation of the fill dominates the failure mechanism of the
cofferdam, (ii) as the width of the cofferdam increases, the
water height at failure increases and (iii) the sheet pile wall at
the downstream is subjected to higher stresses than the sheet
pile wall at the upstream.
Peng et al., (2007)[9] based on the hydraulic computation
theory, a hydraulic numerical model describing an overflow
cofferdam during multi - phase diversion was presented to
achieve the discharge capacities, stream wise water levels,
velocities of both overflow cofferdam and narrowed river.
The calculated values of this model are in good agreement
with the observed data.
Al-Rmmahi, (2009)[10] studied the design and
construction of cellular cofferdams through test models to
observe their stability. Series of laboratory tests had been
http://dx.doi.org/10.15242/IAE.IAE0315413
carried out on two diaphragm cells of different width to depth
ratios (0.75, 0.85, and 1). Then analysis of cellular cofferdam
by software which is known PLAXIS is used to compute
deformations, stresses, and strain in the body of cofferdam
and foundation. The functions represent the relation between
deformations and embedment depths that occurred after
applied loads. Many conclusions had been drawn from this
study. One of the main conclusions is that embedment depth
is greatly effect on the resistance and deformation of the cell.
AL-Khyatt, (2009)[11] studied the stability of cellular
cofferdams through testing the models of three cases of
isolated circular cofferdams with different cell width (b) to
cell height (H) ratio (b/H), with five types of soil fill. The
results of the tests indicate the following: the cells filled with
subbase were more stable against sliding at different (b/H)
ratios, the cells filled with sand passing No.8 were more
stable against overturning at different (b/H) ratios.
AL– Kelabbee, (2010)[12] presented a study about
nonlinear finite element analysis that has been used to predict
the load deflection behavior of circular cell cofferdam under
lateral load by using ANSYS (version 5.4). The full NewtonRaphson method was used for the nonlinear solution
algorithm.
Al-Kassar, (2011)[13] studied the effect of berm and
embedment depth on stability of cofferdam in wet soils. The
results of tests declared that the resistance of cellular retaining
structure (cofferdam) with wetting soils in wet foundations
gives greater resistance than in dry soils.
Al-janabi, (2012)[14] present study of nonlinear finite
element analysis predict the load deflection behavior of
cellular cell cofferdam under lateral load by using ANSYS
(version 12.1) computer program.
The main objective of this paper is studying the effect of
geometry and type of filling soil on the stability of glover-leaf
type cofferdams experimentally and analytically using
ANSYS (version 12.1).
II. MATERIALS AND METHODS
A. Testing Instruments
The following testing instruments were selected and
constructed by authors in order reach the objectives of this
study.
1) The Steel Frame
A steel frame used to carry the soil box with dimension
(900*900mm), and 700mm height. At the middle of its width
fastened knee-braced frame. A knee-braced frame made of
two angle beams (50*50 mm) and 850mm length, are welded
vertically, at their bottom end, to the steel frame. The upper
side of the angles is connected by a steel beam of 200mm
length, 50mm width and 30mm thickness. A space of 150mm
is provided between the two angle beams to allow passing the
steel cable load, on each side of knee-braced frame. Finally,
two steel angle beams (50*50mm) of 470mm length, are
welded at 350mm height from knee-braced to support it.
In each angle beams, a slit-like opening with dimensions
of 650mm length and 25mm width is made at 50mm height
61
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
and 50mm height was fastened by screws above the two sides
of the box. There is one hole in the middle of the beam to
support a steel screwed shaft, 15mm diameter, and 800mm
length. The steel shaft was used to support four dial gages.
from the knee-braced base, this slit is use to fix the pulley
system.
1) The Loading System
The load is applied to the cell by steel cable loop 5mm in
diameter hold around the cell tightly from one end and
connected to the weight holder from the other.
The loading system consists of:
1- The dead weight holder comprises of two parts, the first is
a steel beam with dimension 300mm length, and 50mm
diameter.
2- Square steel plate of 100mm length and 10mm thickness.
The first part is welded vertically to the second part. A
different dead weights of were used as a loading unit.
4) The Dial Gages System
Three dial gages, of 0.01mm accuracy, were used to
measure the displacement of the models throughout the entire
testing program, these dial gages were mounted to the vertical
steel shaft. Fig. 2 shows the cell loading model which was
used in this study.
2) The Pulley System
The pulley system consists of a round steel shaft 30mm in
diameter and 200mm length, a pulley 50mm in diameter was
fixed in the middle of the shaft, and two brackets each one
surrounding ball 60mm and internal diameter 30mm, the two
brackets provided with two holes that was used to fix the
pulley set to the knee braced frame.
Fig. 2. Glover-leaf cell testing model
3) The Soil Box
B. Properties of Soil
A wooden container with inner dimensions (900*900mm),
and 200mm height, was used for modeling the cellular
cofferdams system. At a distance 150mm from the back end
of this box a steel angle beam of 920mm length, 50mm width
Type of soil
Subbase
Sand Passing on Sieve
No.4
River sand
Dry Density
(γd)
(kN/m3)
17.37
Three different types of soils were used in all tests, the
dry density, total unit weight and angle of internal friction for
wet and dry conditions of these soils are shown in Table I.
TABLE I
THE PROPERTIES OFTHE FILLING SOILS IN THE CELL
Total unit weight
Angle of internal
(γw)
friction (ød)
(kN/m3)
(Degree)
19.014
38
15.573
14.3
17.455
14.54
34
31.5
Symbol of soil
32
30
GM
SP
SM
and the second on one leaf (neutral or longitudinal
directions).
C. Experimental Testing Program
In all experimental tests, the soil filling the steel box was
compacted by vibrator to provide uniformly dense bed soil
for models which were used in the study [14]. Three types of
soils were used in the experimental tests; for each test the
same type of soil was used as cell fill and foundation, and
two cases of soils were used (dry and wet). Water content,
dry density, direct shear test were executed for all soil types,
the wetting unit weight was found. The second test tries to
find the angle of internal friction (Ø). The glover-leaf cell
was then filled with dry or wet soil at three layers and
compacted carefully. This cell was located on the middle of
foundation testing box, three different types of clover leaf
cells were tested, see Fig. 3, these types have bed width to
height ratio (b/H) equal to (1.0, 0.85, and 0.75). All types of
clover leaf cells were tested by applying loads at three
different location heights from the base, 100mm, 150mm,
Fig. 3. Glover-leaf cells
III. THEORETICAL MODELING
(ANSYS+CivilFEM), is a finite-element analysis
package, is used to simulate the response of a physical
system to structural loading, and thermal and electromagnetic
effects. (ANSYS+CivilFEM) uses the finite-element method
to solve the underlying governing equations and the
associated problem-specific boundary conditions. The
SOLID45 element has been used in the modeling of soil and
steel. This element is defined by eight nodes having three
degrees of freedom at each node: translations in the node’s x,
and at the top point of model, 300mm. Two different
conditions for applying loads, the first applied on two leafs
http://dx.doi.org/10.15242/IAE.IAE0315413
Angle of internal
friction (øw)
(Degree)
36
62
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
y, and z directions. The element has plasticity, creep,
swelling, stress stiffening, large deflection, and large strain
capabilities. Loads are described in Node and Element
Loads. Pressures may be input as surface loads on the
element faces. Positive pressures act into the element
temperatures and fluencies may be input as element body
loads at the nodes. The geometry and node locations for this
element are shown in Fig, 4, [15].
B. Material Properties
Parameters needed to define the material models can be
found in Table II. As shown in this table, there are two parts
of the material model for each element.
C. Cloverleaf Cofferdam in (ANSYS+CIVIL FEM)
Cloverleaf cofferdam was in depth of 300mm, and width
of (300,255 and 225 mm) respectively. The dimensions of the
foundation base (X=900mm, Y=900mm, Z=200mm), Fig. 5
shows the details of cofferdam geometry. The sand river soil
was used in the program model in dry and wet cases. The load
was applied at one third of height 100mm.
Fig. 4 SOLID45 element [ANSYS12.1, (2010)][14]
A. Nonlinear Solution
The finite element discrimination process yields a set of
simultaneous equations [14].
[K]{u}={Fa}
(1)
Where:
[K] = coefficient matrix. N.m
{u} = vector of unknown DOF (degree of freedom) values.
{Fa} = Vector of applied loads.
(ANSYS + CivilFEM) employs the "Newton - Raphson"
approach to solve nonlinear problems. In this approach, the
load is subdivided into a series of load increments. The load
increments can be applied over several load steps, illustrates
the use of Newton-Raphson equilibrium iterations in a single
DOF nonlinear analysis[14].
Parameters needed to define for the models material such
as (density, angle of internal friction, ect), there are two parts
of the material model for each element.
Material model number 1 consisting of soil filling the cell
and foundation and material model number 2 refers to the
steel sheet pile. The SOLID45 element was being used for
modeling the filling material of cofferdam and steel sheet
pile for the cofferdam.
Fig. 5-a: Clover leaf Cofferdam and foundation in longitudinal
direction
Fig. 5-b: Clover leaf Cofferdam and foundation in neutral
direction
TABLE II
MATERIAL PROPERTIES
Material Model
Number
1
2
Material Properties
Element Type
River sand soil
SOLID45
SOLID45
𝐸soil
𝜐soil
Young’s modulus (𝑁/𝑚2)
Poisson’s ratio
13*103
0.3
𝜌wet
Density of wet soil(𝑘g⁄m3)
1480
𝜌dry
Density of dry soil (𝑘g⁄m )
1457.69
𝐶d
𝐶w
𝜙d
Cohesion in dry case(𝑘𝑁⁄m2)
Cohesion in wet case(k𝑁⁄m2)
Angle of Friction in dry case
3
5
31.5
𝜙w
Angle of Friction in wet case
30
𝐸s
𝜐s
𝜌s
Steel
Young’s modulus (𝑁⁄m2)
Poisson’s ratio
Steel Density (𝑘g⁄m3)
201*109
0.3
7865
http://dx.doi.org/10.15242/IAE.IAE0315413
3
63
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
IV. RESULTS OF EXPERIMENTAL TESTS
Figs. 6 to 14 represent the relationship between the lateral
load and displacement for experimental tests on glover-leaf
cofferdam cells filling with different soil types and for
applying load at 100mm height from the base only. From
these figures it can be seen the following:1-The displacement is linearly proportional to applied load at
the beginning and after that, the curves show non linearly
displacement proportional with load, and then the loaddisplacement relationship become again linear with failure
occurrence.
2- For all tests, the displacement of cell which was filled with
wet soil is resist more than the cell with dry soil when
applying same load on these two cells. This is because the
weight of wet soil is more than the dry soil, in spite of the
friction between the wet soil and the foundation soil is less
than the friction between the dry soil and the base.
3- The cells filled with subbase was resist more than the two
other soils because the density of subbase is more than the
two other soil, so that the cell which fill with subbase will be
have more weight when the volume was constant.
4- The longitudinal direction of clover leaf cell expressed
more resistance than the neutral direction because the
increasing in the length side that versus to load.
5- By making comparison between the experimental results
of the present study with the experimental results of (Alkassar, 2001)[13] which used circular cofferdam cell was
filled with saturated soils (subbase, sand passing sieve
No.4 and river sand) and (b/H) equal to (0.75 and 1), this
comparison show that the failure load required for glover-leaf
cofferdam is more than that load required for circular cell as
following:
- For neutral direction (130-190%) with (b/H)=0.75
- For neutral direction ( 40-80%) with (b/H)=1
- For longitudinal direction (160-250%) with b/H=0.75
- For longitudinal direction (50-100%) with b/H=1
From the above results it can be seen that the glover-leaf
cofferdams is more stable than other types.
Fig. 7. Horizontal displacement and lateral load relationship for
river sand (Load at 100mm height and, 𝐛/𝐇=0.75)
Fig. 8. Horizontal displacement and lateral load relationship for
subbase (Load at 100mm height and, 𝐛/𝐇=0.75)
Fig. 9. Horizontal displacement and lateral load relationship for
sand (Load at 100mm height and, 𝐛/𝐇=0.85)
Fig. 6. Horizontal displacement and lateral load relationship for
sand (Load at 100mm height and, 𝐛/𝐇=0.75)
http://dx.doi.org/10.15242/IAE.IAE0315413
64
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
Fig. 13. Horizontal displacement and lateral load relationship for
river sand (Load at 100mm height and, b/H=1)
Fig.10. Horizontal displacement and lateral load relationship
for river sand (Load at 100mm height and, 𝐛/𝐇=0.85)
Fig. 14. Horizontal displacement and lateral load relationship for
subbase (Load at 100mm height and, b/H=1)
V. RESULTS OF THEORETICAL ANALYSIS
Figs. 15 to 26 show the comparison between the
theoretical and experimental results of applied load with
lateral displacement for cloverleaf cofferdam cell using
different (b/H) ratios (0.75, 0.85 and 1), different cases of
soil (wet and dry), different cases of cell location
(longitudinal and neutral) and for case of applying load at
height 100mm from the base. These figures concluded a good
agreement between the results of the above two methods.
Fig. 11. Horizontal Displacement and lateral load relationship for
subbase (Load at 100mm height and, b/H=0.85)
Fig. 12: Horizontal displacement and lateral load relationship for
sand (Load at 100mm height and, b/H=1)
Fig.15. Comparison between ANSYS and experimental results for
dry river sand in neutral direction for b/H=0.75
http://dx.doi.org/10.15242/IAE.IAE0315413
65
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
Fig.16. Comparison between ANSYS and experimental results for
dry river sand in longitudinal direction for b/H=0.75
Fig.19. Comparison between ANSYS and experimental results for
dry river sand in neutral direction for b/H=0.85
Fig,17. Comparison between ANSYS and experimental results for
wet river sand in neutral direction for b/H=0.75
Fig. 20. Comparison between ANSYS and experimental results for
dry river sand in longitudinal direction for b/H=0.85
Fig,18. Comparison between ANSYS and experimental results for
wet river sand in longitudinal direction for b/H=0.75
http://dx.doi.org/10.15242/IAE.IAE0315413
Fig,21. Comparison between ANSYS and experimental results for
wet river sand in neutral direction for b/H=0.85
66
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
Fig, 22. Comparison between ANSYS and experimental results for
wet river sand in longitudinal direction for b/H=0.85
Fig.25. Comparison between ANSYS and experimental results for
wet river sand in neutral direction for b/H=1
Fig.23. Comparison between ANSYS and experimental results for
dry river sand in neutral direction for b/H=1
Fig. 26. Comparison between ANSYS and experimental results for
wet river sand in longitudinal direction for b/H=1
VI. ARTIFICIAL NEURAL NETWORK (ANN) MODEL
Artificial Neural Network (ANN) is powerful solution to
many complex modeling problems. Many studies have
demonstrated that the (ANN) models are very successful in
hydraulic maters [16, 17, 18]. ANN is an information
processing system that is inspired by the biological nervous
system, such as brain.
The human brain is composed of large number of
interconnected processing elements (neurons). Due to
structure in which the neurons arranged and operate, human
are able to quickly recognize patterns and process data. An
(ANN) is a simplified mathematical representation of
biological neural network. It has the ability to learn from
examples, recognize a pattern in the data, adapt solution over
time, and process information[17]. There are many different
types of artificial neural networks in terms of structure and
mode of operation. In this study, one of the most popular
Fig. 24. Comparison between ANSYS and experimental results
for dry river sand in longitudinal direction for b/H=1
http://dx.doi.org/10.15242/IAE.IAE0315413
67
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
neural networks is examined, the widely used multilayer
perceptron (MLP) network.
Artificial Neural Network is a layered network of artificial
neurons. The neurons or nodes are generally arranged in
parallel to form layers. The first layer, which receives the
inputs, is called input layer and the last layer is called output
layer. The rest are hidden layers whose depend on the
problem to be solved.
The input layer which takes the input values from the
outside. All the nodes of the input layer from the inputs of the
neural network. The nodes of the output layer send the
output values to the user’s external environment[19]. The
hidden layers are the processing center of network system.
The weights are adjusted in an iterative manner to achieve
the expected output values. A typical artificial neural is
shown in Fig. 27.
The number of nodes in input and output layer are fixed
according to the number of dependent and independent
variable in the training data, while the selection of an optimal
number of nodes in the hidden layer depend on the specific
problem. If the number of neurons is small in hidden layer,
the network may not learn the process correctly. On the other
hand if the number is too high, the training will take a long
time and the over fitting of the training data may produce
[17].
From the above table it can be seen that the most important
variable in the stability of glover-leaf cofferdams, in each
direction, is the density of filling soil (γ), while the bed
width of the cell (b) has a little importance.
VIII. CONCLUSIONS
In the present research, the effect of geometry and soil
type on the stability of glover-leaf cofferdams was
investigated experimentally and theoretically. This study
concludes to the following:
1- The clover leaf cells filled with wet subbase were more
stable against sliding and overturning at different (b/H)
ratio.
2- For all tests, the displacement of clover leaf cell which
was filled with wet soil expressed more resistance than
the cell with dry soil when applying same load on these
two cells.
3- The cells that put on the longitudinal direction were more
stable against sliding and overturning at different (b/H)
ratios, because the increase of the opposite load side of
the loading cell.
4- In general, the results obtained using the finite element
models represented by the load applied at one third of the
cell cofferdam height for river sand deflection curves
show good results with the experimental. The difference
between the numerical ultimate loads and the
corresponding experimental ultimate loads is in the range
between (0-6.9)%.
5- The glover-leaf cofferdam has more resistance and it’s
more stable than other types against applying load.
6- By using Artificial Neural Network (ANN) program, the
importance of each effecting variables determined in
neutral and longitudinal direction, the results show that
the density of soil is the bigger effect variable on the
failure loading with effect ratio (35.3%) in neutral
direction and (36%) in longitudinal direction. The
effectiveness of
, loading height(y) and the cell
position (B) in neutral and longitudinal direction was
(10.2%, 33.3%, 21.2%), (18.6%, 31.9%, 13.5%)
respectively.
Fig. 27. Typical Artificial Neural Network
VII. RESULTS OF ANN
The SPSS.17 software application allows the selection
of data division into training set. The (ANN) model
comprised of four neurons in the input layer the input data
namely (b, ɣ (Gama), (b/H) and y), the output value was the
failure force (F). The most suitable data division found here
to be 68.5% in neutral direction (37 runs) for training and
31.5% (17 runs) for testing. Also, the most suitable data
division found for longitudinal direction is 68.5% (37 runs)
for training and 31.5% (17 runs) for testing. Table III shows
the relative importance for each input variable in neutral and
longitudinal directions.
REFERENCES
[1]
[2]
[3]
[4]
TABLE III
THE RALATIVE IMPORTANCE OF VARIABLES
Input
Variable
b
γ (Gama)
b/H
y
Neutral Direction
Importance
0.130
0.426
0.148
0.296
Normalized
Importance
30.4%
100%
34.7%
69.5%
[5]
Longitudinal Direction
Importance
0.135
0.360
0.186
0.319
http://dx.doi.org/10.15242/IAE.IAE0315413
Normalized
Importance
[6]
37.5%
100%
51.7%
88.8%
[7]
[8]
68
K. M. Nemati,"Temporary structure cofferdam", Department of
Construction Management, University of Washington, 2007.
Farrokh Neghabat and Robert M. Stark, "A Cofferdam Design
Optimization", Mathematical programming 3, North-Holland
publishing company , 1972, pp.263-275.
W. L. Schroeder, D. K. Marker, and T. Khuayjarempanishk,
"Performance of a cellular wharf", J. Geotech, Eng. Div., ASCE,
Vol.103, No. GT3, pp. 153-168, 1977.
K. Terzaghi, R. B. Ralf and G. Mesri, Soil Mechanics in
Engineering Practice, 3rd ed., John Wiley & Sons. Inc., 1996, ch. 6,
p. 349.
Max. D. Sorota and Edward B. Kinner, "Cellular cofferdam for
trident dry dock: Design", J. Geotech. Eng. Div., ASCE, Vol. 107,
No. GT12, pp. 1643-1655, 1981.
Charles T. Jahren "Reliability comparison for sheet-pile cellular",
ASCE, Structures Vol. 4, No. 4, pp. 216-235, November 1990.
AL-shamkey, "Stability of cellular cofferdams", M.Sc.. Thesis,
College of Engineering, University of Baghdad, 1992.
R. Amin Khan Mohammod, T. Jiro, F. Hiroki and K. Osamu,
"Behavior of double sheet pile wall cofferdam on sand observed in
2015 International Conference on Food Nutrition and Civil Engineering (ICFNCE’2015) March 14-15, 2015 Dubai (UAE)
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
centrifuge tests ", IJPMG- International Journal of Physical
Modeling in Geotechnics, Tokyo ,Japan, 2001.
Y. Peng, W. Chen and C. JI, "Hydraulic calculation for overflow
cofferdam in staged diversion ", North China Electric Power
University, Ministry of Education, Beijing, China, 2007.
S. H. Al-Rmmahi, "Effect of width to depth ratio on stability of
cellular cofferdams", M.Sc. thesis, College of Engineering,
University of Babylon, 2009.
H. S. Al-Khyatt, "Design sheet pile cellular structures cofferdams
and retaining structures", M.Sc. thesis, College of Engineering,
University of Babylon, 2009.
Z. D. Al-kelabbee, "Finite element analysis of cellular circular
bulkheads". M.Sc. thesis, College of Engineering, University of
Babylon, 2010.
H. M. Al-kassar, "Experimental study of the stability of the cellular
retaining structures in wetted soils", M.Sc. thesis, College of
Engineering, University of Kufa, 2011.
P. B. Kelly, "Design and evaluation of a foundation model testing
device", .M.Sc. thesis, Oregon State University, U.S.A, 1969.
ANSYS 12.1, “ANSYS 12.1 Finite Element Analysis System Help,”
SAS IP, Inc., 2010.
H. K. CIGIZOGLU, "Estimation, forecasting and extrapolation of
river flows by artificial neural networks", J. Hydrol. Sci., 48(3),
2003, pp. 349-360.
O. Kisi, "Daily river flow forecasting using artificial neural
networks and auto-regressive models", J. Eng. Sci., 2005, pp. 9-20.
R. Teschl and W. L Randue. A neural network model for short term
river flow prediction", Nat. Hazards Earth Syst. Sci., 6, 2006, pp.
629-635.
M. S. Erdurmaz, "Neural network prediction of tsunami parameters
in the Aegean and Marmara seas. M.Sc. thesis. Middle East
Technical University, 2004.
Dr. Raad Hoobi Irzooki, birth in Iraq, Baghdad,
October-1963, Assistant Professor, B.Sc. in Water
Resources Engineering from Baghdad University,
Iraq in 1985. M.Sc. in Water Resources
Engineering – Hydraulic Structures from Baghdad
University in 1991. Ph.D. in Building and
Construction
Engineering-Water
Resources
Engineering from University of Technology, Iraq
in 1998. Areas of Expertise: Hydraulic Structures and Seepage Through
Earth Dams.
He is instructor in the College of Engineering – Tikrit University, Iraq
from 1992 till now. He is a head of Civil Engineering in Tikrit University
from 2001 to 2003 and the director of consulting engineering bureau from
2007 to 2011.
Dr. Irzooki issued supervised more than 20 Ph.d , M.Sc. and Higher
Diploma students and published 15 papers in local and international
journals. E-mail: dr.raadhoobi@yahoo.com
Marwa.k Majeed from Iraq, Samarra, birth in June
1990, B.Sc. in Civil Engineering 2012 from Tikrit
University, Iraq, She is a master Student in Tikrit
university too.
She is work as a lecturer in Engineering College in
Samarra University.
http://dx.doi.org/10.15242/IAE.IAE0315413
69