You are on page 1of 8

OPTIMIZATION AND SENSITIVITY OF RETAINING STRUCTURES

By A§km Sarlba§l and Fuat Erbatur2

ABSTRACT: This paper is concerned with optimum design and sensitivity of retaining structures. The optimum
design formulation in terms of a constrained nonlinear programming problem, is given for reinforced concrete-
cantilever retaining walls. The objective function may be chosen as the cost or weight of the wall. The solution
is carried out by a specially prepared computer program (RETOPT). Illustrative problems are solved, and their
results are presented and discussed. The formulation allows for a detailed sensitivity analysis to be made for
selected design parameters, also depicted with numerical examples.

INTRODUCTION with the interaction of the retaining structure and the surround-
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

ing soil. The primary aim of the problem is to explore the


This paper presents the results of part of a comprehensive conditions supporting the backfill safely, with controlled dis-
investigation concerned with optimum design of retaining placements of the specific retaining structure and the backfill
structures. In this regard, formulation and solution methodol- itself. The general trend followed in the design of such struc-
ogy is presented for reinforced concrete-cantilever retaining tures is to ensure both that the structure and the soil surround-
walls (Fig. I). The optimum design problem is posed as a ing the structure do not fail and that the deformations that take
constrained nonlinear programming problem that is solved by place are within acceptable limits. These requirements are usu-
a specially prepared program. Two objective functions, ally met by avoiding the possibility of overturning and sliding
namely, the cost and weight of the wall are considered. Seven of the structure, by controlling the stresses within allowable
design variables representing the cross-sectional dimensions limits, and by safeguarding the stability of both the structure
and reinforcing steel are chosen. The programming problem is and the soil.
complemented by 24 behavioral and side constraints.
Optimum design of retaining walls has been the subject of REINFORCED CONCRETE-CANTILEVER RETAINING
a number of studies: Alshawi et al. (1988); Dembicki and Chi WALLS
(1989); Fang et al. (1980); Keskar and Adidam (1989); Pocht-
The reinforced concrete-cantilever retaining wall is the most
man et al. (1989); and Rhomberg and Street (1981). These common of the retaining structures. The wall has a simple
studies deal with various aspects of optimal design, including, geometry, consisting of a base slab and a stem. It is made in
optimal shapes. maximization of structural stability, minimi- the form of an inverted T or L. The main design principles,
zation of bending moments, and optimum location on sloping briefly mentioned in the previous section, are also valid for
hillsides. The main features of the present study are: it is com-
reinforced concrete-cantilever retaining walls. Specific to these
prehensive, assumptions are kept to a minimum, the computer
program is used with ease, optimal design is automatically
reached, the computation time is very short, and it is flexible
in formulation allowing for further modification and extension.
Two example problems are presented. The parameters of the
example problems are tabulated in Tables 3 and 7. The optimal
solutions include: objective function values, amount of steel
and concrete, magnitude of design variables, and magnitudes
of constraints. The results are given in Tables 4-6 and Tables
8-10.
The solution approach readily allows to study the sensitivity
of optimum design. The sensitivity of the optimal solution may
be searched with respect to a variety of design parameters,
which include those related to soil properties, loading, and H
Ws
material characteristics. The results of sensitivity analyses for
height and top thickness of the stem, surcharge load, backfill
slope. internal friction angle of retained soil. and the yield
I}
:w,. t
strength of reinforcing steel are presented (see Tables II and
12 and Figs. 2-7). 1: .
w2 :I
BEHAVIOR OF RETAINING STRUCTURES I
I
The study of retaining structures falls in the domain of the I

general soil-structure interaction problem. This is concerned


X4
'Res. Assoc.. MS in Civ. Engrg.• Dept. of Civ. Engrg.• Middle East
Tech. Univ., 06531 Ankara, Turkey. X
2Prof.• PhD in Civ. Engrg., Dept. of Civ. Engrg., Middle East Tech.
Univ.• 06531 Ankara. Turkey. x,-X 2 -X 3
Note. Discussion open until January I, 1997. To extend the closing
date one month, a written request must be filed with the ASCE Manager X2 x1 - X2
of Journals. The manuscript for this paper was submitted for review and
possible publication on August 10, 1995. This paper is part of the Journal Xl
of Geotechnical Engineering, Vol. 122, No.8. August. 1996. ©ASCE,
ISSN 0733-9410/96/0008-0649-0656/$4.00 + $.50 per page. Paper No. FIG. 1. Mathematical Model Used for Optimum Design of Re-
11369. Inforced Concrete Cantilever-Retaining Wall

JOURNAL OF GEOTECHNICAL ENGINEERING / AUGUST 1996/649

J. Geotech. Engrg. 1996.122:649-656.


structures, the following failure modes should be considered TABLE 2. Lower and Upper Bounds of Design Variables
(Keskar and Adidam 1989; Rhomberg and Street 1981): no Lower bounds Upper bounds
tension condition in the foundation, sliding stability, overturn- (1) (2)
ing stability (about the toe), failure in bearing capacity, and
failures related to the shear and moment capacities of the three X,,,,,. = OAH(l2/1l) X,_ = (0.7H)/0.9
main components of the wall, namely, the stem, toe, and heel.
X 2"".= [OAH(12111»)/3 X 2_ =[(0.7H)/0.9)/3
X 3"". = t X 3 _. = (H/0.9)110
X..... = [H(12111»)/12 X 4 _. = (H/0.9)/1O
FORMULATION OF OPTIMUM DESIGN PROBLEM X,,,,,. = 10,000 Pm..(t - 0.01 m) X,_. = 10,000 Pm~(X3_. - 0.01 m)
X..... = 10,000 Pmln(X4"". - 0.01 m) X"-. = 10,000 Pm~(X4_. - 0.01 m)
The general three phases considered in the optimum design =
X7"". 10,000 Pm..(X..... - 0.01 m) X7_. = 10,000 Pm~(X4_. - 0.01 m)
of any structure are: structural modeling, optimum design Note H = Height of stem. Pm.. = Minimum steel ratio. Pm~ = Maximum
modeling, and the optimization algorithm. Since the structure steel ratio. m = Clear concrete cover + half of diameter of reinforcing
is statically determinate, the cantilever wall does not present bars.
any problem as far as the structural analysis is considered. For
structural design details, the requirements of the American
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

Concrete Institute (ACI) (1990) are used. For the optimum pressed as functions of the design variables and correspond to
design modeling, one has to study the problem parameters in- the 10 behavior constraints, defined as inequalities
depth, so as to decide on design parameters, design variables, gj(X) S 0, j = I, ... , 10 (1)
constraints, and the objective function. These and the design-
optimization methodology are discussed in the following sec- where x = vector of design variables.
tions. The derived constraint expressions are found to be highly
nonlinear in the design variables (Sarlbll§ 1995). In addition,
Design Variables all the design variables have practical minimum and maximum
values (ACI 1990; Gaylord and Gaylord 1990; and Bowles
The design variables chosen for the formulation are related 1988), which are shown in Table 2. These side constraints are
to the cross-sectional dimensions of the wall and various re- also expressed as inequality constraints
inforcing steel areas (Fig. 1). Seven design variables are taken
into consideration. These include the following:
[(Xi)min/X;] - ISO, i = I, ... , 7 (2)
for minimum values and
X, = Total base width
X2 = Toe projection (3)
X3 = Stem thickness at the bottom for maximum values.
X4 = Thickness of base slab Thus, the total number of design constraints involved in the
X~ = Vertical steel area of the stem per unit length of the
optimum formulation is 24.
wall
X6 = Horizontal steel area of the toe per unit length of the Objective Functions
wall
X7 = Horizontal steel area of the heel per unit length of the Two objective functions, namely, the weight and cost, have
wall been chosen for flexibility of use and for comparison purposes.
In cost minimization the objective function is defined as
The first four design variables are related to the geometry
of the cross section, and the last three consider various steel TABLE 3. Input Parameters for Example 1
areas. The height of the stem and the stem thickness at the top
Input parameter Unit Symbol Value
are included in the design parameters. Design parameters are (1 ) (2) (3) (4)
preassigned at the beginning of the structural optimization pro-
cess. Other design parameters include some soil properties, Height of stem m H 3.0
Top thickness of stem m t 0.20
unit cost of materials, loading characteristics, and others (dis- Yield strength of reinforcing steel MPa I, 400
cussed later in relation to the sensitivity analysis). Compressive strength of concrete MPa j; 21
Wide beam shear strength of concrete MPa ~, 0.65
Constraints Concrete cover cm d, 7
Maximum steel percentage - Proal( 0.016
The main design philosophy covering the requirements, Minimum steel percentage - Pmin 0.00333
based on the behavior of reinforced concrete-cantilever retain- Shrinkage and temporary reinforcement
percent - PST 0.002
ing walls, is summarized in the preceding sections (Table 1). Diameter of bars cm <1>"", 1.2
These requirements represent the failure modes that are ex- Surcharge load kPa q 20
Backfill slope degree 13 10
TABLE 1. Failure Modes Internal friction angle of retained soil degree <1> 36
Internal friction angle of base soil degree <1>' 0
Inequality constraint Failure mode Unit weight of retained soil kN/m 3 "I, 17.5
(1 ) (2) Unit weight of base soil kN/m 3 'Y: 18.5
g,(X) Shear at bottom of the stem Unit weight of concrete kN/m 3 'Yc 23.5
Moment at bottom of stem Cohesion of base soil kPa c 125
g2(X)
g3(X) Overturning stability Design load factor - LF 1.7
g,<x) Sliding stability Depth of soil in front of wall m D 0.5
g,(X) No tension condition in foundation Cost of steel- $/kg C, 0040
Cost of concrete' $/m 3 C, 40
g.(X) Bearing capacity
g7(X) Toe shear Factor of safety for overturning stability - N, 1.5
g,<x) Toe moment Factor of safety against sliding - N, 1.5
g9(X) Heel shear Factor of safety for bearing capacity - SF 3.0
glO(X) Heel moment 'Local market rates.

650/ JOURNAL OF GEOTECHNICAL ENGINEERING / AUGUST 1996

J. Geotech. Engrg. 1996.122:649-656.


TABLE 4. Optimum Values of Design Variables for Example 1 TABLE 7. Input Parameters for Example 2
Optimum Optimum Input parameter Unit Symbol Value
values values (1 ) (2) (3) (4)
Design Lower Upper minimum minimum Height of stem m H 4.5
variables Unit bounds bounds cost weight Top thickness of stem m t 0.25
(1 ) (2) (3) (4) (5) (6) Yield strength of reinforcing steel MPa h 400
XI m 1.309 2.333 1.578 1.574 Compressive strength of concrete MPa f; 21
X2 m 0.436 0.778 0.436 0.441 Wide beam shear strength of concrete MPa {}e 0.65
X, m 0.200 0.333 0.258 0.200 Concrete cover cm de 7
X4 m 0.273 0.333 0.273 0.273 Maximum steel percentage - Pmax 0.016
X, cm2/m 4.129 41.173 12.574 21.072 Minimum steel percentage - Pmin 0.00333
cm2/m 6.551 41.173 6.551 6.551 Shrinkage and temporary reinforcement
X.
X7 cm2/m 6.551 41.173 6.551 6.681 percent - PST 0.002
Diameter of bars cm <1>..... 1.4
Surcharge load kPa q 30
Backfill slope degree 13 15
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

TABLE 5. Values of Behavioral Constraints at Optimum Val-


ues of Design Variables for Example 1 Internal friction angle of retained soil degree <1> 36
Internal friction angle of base soil degree <1>' 34
Minimum Minimum Unit weight of retained soil kN/m' 'I, 17.5
cost weight Unit weight of base soil kN/m' 'I: 18.5
Unit value value Unit weight of concrete kN/m' 'Ie 23.5
Constraint Symbol
Cohesion of base soil kPa c 100
(1 ) (2) (3) (4) (5)
Design load factor - LF 1.7
Shear capacity of stem gl(X) kN 55.893 17.985 Depth of soil in front of wall m D 0.75
Moment capacity of stem g2(X) kN'm 0 0 Cost of steel $/kg C, 0.40
Overturning stability g,(X) kN'm 73.333 74.001 Cost of concrete $/m' Ce 40
Sliding stability g.(X) kN 69.823 69.212 Factor of safety for overturning stability - No 1.5
No tension in foundation g,(X) m 0 0 Factor of safety against sliding - N, 1.5
Bearing capacity g.(X) kPa 104.663 102.204 Factor of safety for bearing capacity - SF 3.0
Shear capacity of toe g,(X) kN 51.980 49.76
Moment capacity of toe g.(X) kN'm 27.165 26.455
Shear capacity of heel g9(X) kN 60.808 61.734 The program constructs an unconstrained minimization prob-
Moment capacity of heel glO(X) kN'm 2.102 0 lem by using a selected penalty-function approach. Four dif-
ferent types of penalty functions are provided in the program:
(4)
one pass-external function, the Fiacco-McCormick combined
external and internal function, Powell's function, and
where C, = unit cost of steel; Cc = unit cost of concrete (Cc Schuldt's function (Siddal 1992).
has been selected to account for fonnwork, placing of con- The fonnulation and solution algorithm is tested on several
crete, vibration, and so on, including the labor charges); Wst = example problems. Two representative cases are reported. The
weight of steel per unit length of the wall; and Vc = volume details of these examples are given in the input data (see Ta-
of concrete per unit length of the wall. bles 3 and 7). In the example problems Sl units are used. In
For weight optimization the objective function definition is each case, both weight and cost minimization is considered.
Optimum design results are shown in Tables 4-6 and Tables
f(x) = W" + l00V/yc (5)
8-10. The optimum values of the design variables are tabu-
where 'Yc = unit weight of concrete, and 100 is used for con- lated together with suggested (ACI 1990; Gaylord and Gaylord
sistency of units. 1990; and Bowles 1988), upper and lower limits for easy in-
terpretation (Tables 4 and 8). The values of the behavioral
SOLUTION AND RESULTS constraints at the optimum and for each case are given in Ta-
bles 5 and 9, respectively. Finally, the optimum values of the
The optimum design problem is now posed as: find x, sat- objective functions, the weight of steel, and the volume of
isfying concrete in these designs are tabulated in Tables 6 and 10,
j = 1, ... , 10 (00) respectively.
ANALYSIS
[(X;)mJnlx/] - 1 :S 0, i = 1, ... ,7 (6b)
Example 1
[xi(Xt)max] - 1 :S 0, k = I, ... , 7 (6c) As shown in Table 4, of the first four design variables that
describe the shape of the optimum wall, only the thickness of
minimizing f(x) = C, WS1 + Cc Vc for cost or, minimizing f(x)
= Wst + 100Vc'Yc for weight. TABLE 8. Optimum Values of Design Variables for Example 2
This is a constrained nonlinear programming problem. For
the numerical solution, a special program (RETOPT), the de- Optimum Optimum
tails of which are given in Sanba~ (1995), has been prepared. values values
Design Lower Upper minimum minimum
TABLE 6. Optimum Values of Objective Functions, Weight of variables Unit bounds bounds cost weight
Steel, and Volume of Concrete for Example 1 (1 ) (2) (3) (4) (5) (6)
X, m 1.964 3.500 2.254 2.238
Weight of Volume of
X2 m 0.655 1.167 0.655 0.655
Optimum steel concrete
X, m 0.250 0.500 0.417 0.300
Objective function Unit value (kg/m) (m 3 /m)
~ m 0.409 0.500 0.409 0.409
(1 ) (2) (3) (4) (5)
X, cm'/m 5.761 67.680 23.475 41.626
Minimum cost $/m 82.474 60.410 1.118 X. cm'/m 11.059 67.680 11.059 11.059
Minimum weight kglm 2,498.774 79.788 1.029 X7 cm2/m 11.059 67.680 11.059 11.059

JOURNAL OF GEOTECHNICAL ENGINEERING I AUGUST 1996/651

J. Geotech. Engrg. 1996.122:649-656.


the stem at the bottom, X 3 , differs slightly in the two optimi- shown in Table 6. With respect to the weight of steel and
zation models (others being almost the same). As for the de- volume of concrete for each optimization model (Table 6), the
sign variables representing steel areas, while X6 resumes the former one takes greater value for the minimum weight model,
lower limit for each objective function, X7 takes the lower limit and the latter takes greater value for the minimum cost model.
for cost optimization and is a slightly larger value than the
lower limit for weight optimization. The only significant dif- Example 2
ference between the results for the two objective functions is
experienced in X" the vertical steel area of the stem per unit Table 8 shows the optimum values of the design variables.
length of the wall. This is 12.574 and 21.072 for cost- and The variables, X 2 , X.., X 6 , and X7 , resume the lower limits for
weight-optimization models, respectively. As for the con- both objective functions. The others fall within the lower lim-
straints, in Table 5, the second and fifth constraints corre- its and the arithmetic average of the lower and upper limits,
sponding to moment capacity of the stem and no tension in as in the first example. And, X3 violates this pattern by taking
the foundation and the tenth constraint corresponding to mo- a value jn between the arithmetic average and the upper limit.
ment capacity of the heel are all active in both models, ig- As in the first example, a significant difference is seen in X"
noring the small slack in the tenth constraint for cost optimi- which is 23.475 for the minimum cost model and 41.626 for
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

zation. The optimum values for total cost and total weight are the minimum weight model. Design variable X3 shows a sim-
350
TABLE 9. Values of Behavioral Constraints at Optimum Val-
ues of Design Variables for Example 2
Minimum Minimum 300
cost weight
Constraint Symbol Unit value value
(1 ) (2) (3) (4) (5) I 250
Shear capacity of stem
Moment capacity of stem
g,(X)
g,(X)
kN
kN'm
75.661
0
0
0 !
Overturning stability
Sliding stability
No tension in foundation
g3(X)
g4(X)
gs(X)
kN'm
kN
m
228.051 229.756
142.246 143.491
0 0
J
:E
200
_ _ toO.2Om
_ _ toO.25m

Bearing capacity g.(X) kPa 1,762.754 1,756.297 i _ _ toO.3Om


Shear capacity of toe
Moment capacity of toe
g7(X)
g.(X)
kN
kN'm
42.701
67.200
36.989
68.555 I...
::0
150
Shear capacity of heel g9(X) kN 76.315 78.083
Moment capacity of heel glO(X) kN'm 13.834 1.542 t
~ 100
'0
TABLE 10. Optimum Values of Objective Functions, Weight of
Steel, and Volume of Concrete for Example 2 !
50
Weight of Volume of
Optimum steel concrete
Objective function Unit value (kg/m) (m'/m)
(1 ) (2) (3) (4) (5) 0
Minimum cost $/m 189.546 155.486 2.423 3 3.5 .. ".5 5 5.5 6
Minimum weight kglm 5,280.961 218.319 2.154 Height of the Stem (Homl

FIG. 2. Effect of Height of Stem on Optimum Value of Cost·


Mlnlmlzatlon-objectlve Function for Various Top Thicknesses
TABLE 11. Input Parameters for Various H-tComblnatlons
10000
Input parameter Unit Symbol Value
(1 ) (2) (3) (4)
9000
Yield strength of reinforcing steel MPa h 400
Compressive strength of concrete MPa f; 21 ~
:!!. 8000
Wide beam shear strength of concrete MPa {)e 0.65 2!
Concrete cover em 7
Maximum steel percentage -
de
P""", 0.016 f 7000

-
Minimum steel percentage
Shrinkage and temporary reinforcement
percent -
Pmln 0.00333

0.002
I.5
6llOO
PST ~
Diameter of bars em <l>b" 1.4 .a 5000
Surcharge load kPa q 25
Backfill slope
Internal friction angle of retained soil
degree
degree
I)
<I>
10
36
j «lOll

i
Internal friction angle of base soil degree <1>/ 0 3000
Unit weight of retained soil kN/m 3 "Y, 17.5
3 J;'
Unit weight of base soil kN/m "Y; 18.5 0 _ _ t=a.2Om
2000
Unit weight of concrete kN/m3 "Ye 23.5 '0 ___ toO.25m
Cohesion of base soil kPa c 125 J 1000
Design load factor - LF 1.7 ~ _ _ t=a.3Om
Depth of soil in front of wall m D 0.75
Cost of steel" $/kg C, 0.40 0
Cost of concrete $/m 3 Ce 40 3.0 3.5 ".0 ".5 5.0 5.5 6.0
Factor of safety for overturning stability - No 1.5 Height of the stem lKoml
Factor of safety against sliding - N, 1.5
FIG. 3. Effect of Height of Stem on Optimum Value of Welght-
Factor of safety for bearing capacity - SF 3.0
Minimization-Objective Function for Various Top Thicknesses

652/ JOURNAL OF GEOTECHNICAL ENGINEERING I AUGUST 1996

J. Geotech. Engrg. 1996.122:649-656.


Har tendency. It is 0.417 and 0.300 for minimum cost and weight model compared to the minimum cost model. The con-
minimum weight models, respectively. The second, fifth, and trary is true for the concrete volume, which is higher in the
tenth constraints are active for each optimization model (Table minimum cost model. This is also seen in the first example.
9). Another constraint that is active is constraint number one,
however only for the minimum weight model. Table 10 gives SENSITIVITY ANALYSIS
the optimum value for total cost and total weight, together with
the weight of steel and the volume of concrete for each ob- The design parameters considered in the present study cover
jective function. The steel weight is higher in the minimum a wide range of parameters that are related to loading, ge-

TABLE 12. Optimum Values of Design Variables In Minimum Cost- and Minimum Weight-Optimization Models for Various H-tCom-
blnatlons
Optimum Value for Minimum Cost Optimum Value for Minimum Weight
H
(m)
(1 )
t= 0.20 m
(2)
I t= 0.25 m
(3)
I t= 0.30 m
(4)
t= 0.20 m i t = 0.25 m i t = 0.30 m
(5) (6) (7)
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

(a) = XI (m)
3.0 1.607 1.607 1.609 1.595 1.602 1.609
3.5 1.851 1.851 1.853 1.841 1.846 1.852
4.0 2.095 2.096 2.097 2.087 2.090 2.094
4.5 2.340 2.340 2.341 2.332 2.334 2.336
5.0 2.584 2.583 2.583 2.576 2.577 2.578
5.5 2.827 2.826 2.825 2.818 2.819 2.820
6.0 3.070 3.068 3.067 3.061 3.061 3.062
(b X2 (m)
3.0 0.481 0.477 0.443 0.436 0.436 0.436
3.5 0.522 0.517 0.514 0.509 0.509 0.509
4.0 0.582 0.582 0.582 0.582 0.602 0.603
4.5 0.655 0.655 0.655 0.658 0.684 0.700
5.0 0.736 0.730 0.727 0.735 0.762 0.789
5.5 0.826 0.819 0.814 0.812 0.839 0.866
6.0 0.921 0.914 0.908 0.889 0.916 0.943
(c) X, (m)
3.0 0.272 0.272 0.300 0.200 0.250 0.300
3.5 0.313 0.312 0.312 0.212 0.250 0.300
4.0 0.355 0.354 0.354 0.244 0.250 0.300
4.5 0.399 0.399 0.399 0.278 0.278 0.300
5.0 0.447 0.446 0.445 0.316 0.316 0.316
5.5 0.495 0.495 0.494 0.356 0.356 0.356
6.0 0.546 0.545 0.545 0.400 0.400 0.400
(d) X. (m)
3.0 0.273 0.273 0.273 0.273 0.273 0.273
3.5 0.318 0.318 0.318 0.318 0.318 0.318
4.0 0.364 0.364 0.364 0.364 0.364 0.364
4.5 0.409 0.409 0.409 0.409 0.409 0.409
5.0 0.455 0.455 0.455 0.455 0.455 0.455
5.5 0.500 0.500 0.500 0.500 0.500 0.500
6.0 0.545 0.545 0.545 0.545
- 0.545 0.545

3.0 13.293 13.212 11.404 25.341 15.392 11.404


3.5 16.011 16.032 16.051 37.382 23.857 17.108
4.0 18.905 18.942 18.916 39.673 36.778 24.835
4.5 21.852 21.889 21.861 42.319 42.319 35.502
5.0 24.829 24.856 24.924 45.123 45.123 48.009
5.5 27.977 28.007 28.035 48.009 48.009 48.009
6.0 31.224 31.257 31.288 50.942 50.942 50.942

3.0 6.518 6.518 6.518 6.518 6.518 6.518


3.5 8.031 8.031 8.031 8.031 8.031 8.031
4.0 9.545 9.545 9.545 9.545 9.545 9.545
4.5 11.059 11.059 11.059 11.059 11.059 11.059
5.0 12.572 12.572 12.572 12.572 12.572 12.572
5.5 14.086 14.086 14.086 14.086 14.086 14.086
6.0 15.600 15.600 15.600 15.600 15.600 15.600

3.0 6.518 6.518 6.518 7.499 7.039 6.577


3.5 8.031 8.031 8.031 9.078 8.689 8.188
4.0 9.545 9.545 9.545 10.669 10.374 9.809
4.5 11.059 11.059 11.059 12.358 12.041 11.588
5.0 12.572 12.572 12.572 14.136 13.784 13.435
5.5 14.086 14.086 14.086 16.016 15.630 15.247
6.0 15.600 15.600 15.600 17.991 17.570 17.152

JOURNAL OF GEOTECHNICAL ENGINEERING / AUGUST 1996/653

J. Geotech. Engrg. 1996.122:649-656.


215 5800 iables for the rest of the design parameters are briefly illus-
·§
.51 trated.
:l!!
..., 205 5600 The input parameters of the wall considered for the sensi-
:; tivity analysis for height and top thickness are given in Table
~
195 5400 11. Both objective functions are considered, and their optimum
a value variations with respect to changes in height of the stem
.. "" 185 5200 for various top thickness values are given in Figs. 2 and 3.
J) Table 12 shows the optimum values of the design variables
~ a 175 5000 for both the minimum cost- and minimum weight-optimization
,se II ... models. In the range of the stem height considered (3-6 m),
165 4800
~f.
';1
Figs. 2 and 3 reveal that for higher values of height, the op-
!~ 155 4600
timum cost and weight become more sensitive to variations in
the stem height. This is more apparent for the cost function.
i
~
145 4400 For example, when t = 0.20 m and H changes from 3.0 to 6.0
m, the cost of the wall increases 3.73 times, whereas the
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

...
'C 135 4200 weight increases 3.38 times. These rates show a declining
trend as t increases from 0.20 to 0.30 m. Smaller top thickness
;;
> 125 4000 . values of the stem produce more favorable optimum solutions
for both objective functions.
0 10 20 30 40 50 With regards to the design variables (Table 12), the opti-
Surcharge Load (q-kPa) -+-Min.Cost ($1m) mum values of the first four are sensitive to changes in H for
_ _ Min'wcight (kg/m) both of the minimization models. For a given H, the first four
design variables are not much affected by increases of t, from
FIG. 4. Effect of Surcharge Load on Optimum Values of ObJec- t = 0.20 to 0.30 m, apart from the third design variable (stem
tive Functions for Minimum Cost- and Minimum Weight-Optimi- thickness at the bottom), the lower bound of which is t. For
zation Models both minimization models, the optimum values of the last
three design variables corresponding to reinforcing steel areas
200 5500 show sensitivity to changes in H, but not in general to shifts
.il.!l in t. Only the vertical steel reinforcing area in the weight-
)l
.... 5400 minimization model is influenced by changes in t.
iii 190 The input parameters assumed for the study of sensitivity
i
~
5300 for surcharge loads are mostly the same as for the sensitivity
analysis for Hand t (Table 11). Though here, H = 4.5 m, t =
1.- 0.25 m, "Is = 18 kN/m3 , and D = 1.0 m. The optimum values
~~ 180 5200
for the objective functions are shown in Fig. 4 for the sur-
.. a charge load, varying from 0 to 50 kPa. According to Fig. 4,
a~ 5100
the cost-minimization model is more sensitive to variations in
if. 170 surcharge load compared to the weight-minimization model.
!~
..
5000
In fact, as q changes from 0 to 50 kPa, the optimum cost
increases 1.32 times, and the optimum weight increases by
~ 4900
1.20.
"E
Q
160
As for the sensitivity of the design variables, significant sen-
os
oj
.. 4800 sitivity is observed in X I -X3 (total base width, toe projection,
> and stem thickness at the bottom) and Xs (vertical steel area
150 4700 of the stem). These variables increase as the surcharge load
0 5 10 15 20 25 30
__ Min. Cost (gm)
BlII:kflll Slope (degree) __ Min. Weight (kg/m) 220 - , - - - - - - - - - - - - . , . - 6000

FIG. 5. Effect of Backfill Slope on Optimum Values of ObJec- 5800


tive Functions for Mlnlmum-Cost- and Minimum Welght-optlml-
zatlon Models 5600

5400
ometry, soil properties, code specifications, unit cost, and other
characteristics of construction materials. Sensitivity of the op- 5200
timum solution to changes in these parameters is an important
issue as far as practical design is concerned. The analysis re- 5000
sults include the sensitivities of the optimum weight and op-
timum cost as objective functions and the optimum values of 4800
the seven design variables. As a representative of such anal-
yses, results concerned with the sensitivity of optimum solu- 100 -j----t-----t---+---+ 4600
tions with respect to height and top thickness of stem, sur- 28 30 32 34 36
charge load, backfill slope, internal friction angle of retained Internal Friction Angle or Retained Soil (degree).- -..
soil, and the yield strength of reinforcing steel are reported. -+-Min.Cost (gm)
Sensitivities of the objective functions are explained for all _ _ Min'wcight (kg/m)
design parameters considered. However, due to space restric-
tions, from the sensitivities of the design variables, only those FIG. 6. Effect of Internal Friction Angle of Retained 5011 on Op-
related to changes in height and the top thickness of the stem timum Values of Objective Functions for Minimum Cost- and
are reproduced (Table 12). The sensitivities of the design var- Minimum Weight-Optimization Models

6541 JOURNAL OF GEOTECHNICAL ENGINEERING 1 AUGUST 1996

J. Geotech. Engrg. 1996.122:649-656.


200 variables show some degree of sensitivity, though not very
significant.
E
~
-+- rc=21 MPa For the sensitivity analysis related to the internal friction
.Ec __ rc=24MPa
angle of retained soil, the same parameters are taken as in the
:i... 195 .....-rc-28MPa case of the surcharge load analysis. However, two additional
-~
c
0 parameters, q = 15 kPa and J3 = 10° are added. The optimum
solutions for weight and cost are shown in Fig. 6. Both models
are sensitive to changes in the internal friction angle of re-
c 190
~-
II. E tained soil. The minimum weight model is more sensitive in
~~ the range considered when compared to the minimum cost
'filii
III 0
model.
:soU As far as the design variables are concerned, in the cost-
0 185
....0 minimization model, X4 (thickness of base slab), X6 (horizontal
III
~
steel area of the toe), and X7 (horizontal steel area of the heel)
are insensitive to changes in the internal friction angle, and
~
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

E 180 others decrease as the angle increases in the defined range. In


~
the weight-minimization model, X4 and X6 are insensitive to
.E
a.
0
changes in the internal friction angle, and the others decrease
with an increase in the angle, with the exception of X5 (vertical
175 -1-------+--------; steel arc of the stem). And, X5 shows a decrease in cost min-
imization and an increase in weight minimization with an in-
275 345 400 crease in the angle.
f,-Yield Strength of Reinforcing Steel (MPa) The input parameters for studying sensitivity with respect
FIG. 7. Effect of fyand f'; on Optimum Value of Cost-Mlnlmlza- to yield strength of reinforcing steel are the same as those for
tion-Objectlve Function the internal friction angle, with <l> = 36° added to the list. The
optimum weights and costs are shown in Figs. 7 and 8. Both
the cost and weight are reduced as the reinforcing steel
__ fc=21 Mr. strength is increased. This is more significant for the cost-
5140 - , - - - - - - - - - - - - - - - ,
___ fc=24 Mr. optimization model. For example, with respect to the com-
:c
'" __ fc=28Mr. pressive strength of concrete of 21 MPa, using fy = 400 MPa
~ 5120
instead of h = 275 MPa reduces the cost by 10%.
E
= For both models, the design variables that are notably sen-
E
'c sitive to increases in the yield strength are those corresponding
:i to the reinforcing areas-the most sensitive being the vertical
... 5100
.2 reinforcement area.
c
.S!
ti_
~.g, 5080 CONCLUSIONS
~:!. This paper discusses optimum design of reinforced con-
tl crete-cantilever retaining walls as a part of a comprehensive
~ 5060 investigation of retaining structures. The problem is posed as
o
'0 a nonlinear programming problem. In this formulation, two
! objective functions, the cost and weight of the wall, are con-
~ 5040 sidered. The constraints are the design requirements that are
E the behavioral constraints and side constraints defined as lower
~

.S and upper limits of the design variables. The optimum solution


g 5020 -f--------+---------i is obtained using a specially prepared computer program
275 345 400 (RETOPT). 1\\10 numerical examples are given, and the results
f,-Yield Strength of Reinforcing Steel IMPe' are discussed. The proposed method automatically yields an
FIG. 8. Effect of fyand f'; on Optimum Value of Welght-Mlnlml- optimum design in terms of seven design variables covering
zation-Objectlve Function certain dimensions of the cross section and reinforcing areas.
The formulation is quite flexible and open to further modifi-
cations and extensions. Additionally, sensitivity of the opti-
increases, in both of the optimization models. The optimum mum solutions with respect to certain design parameters can
value of X7 (horizontal steel area of the heel) also increases be performed, also illustrated with numerical examples.
with an increase in the surchage load, yet, after 20 kPa and
only in the weight-minimization model. ACKNOWLEDGMENTS
The input parameters considered for sensitivity, with respect
to backfill slope, are similar to those used for the study of the This paper is part of an MS thesis entitled, "Optimum Design of Re-
inforced Concrete Cantilever Retaining Walls by Non-linear Program-
effect of the surcharge load with an added parameter q = 15 ming," submitted to the Department of Civil Engineering, Middle East
kPa. The range of backfill slope conceived is from 0° to 30°. Technical University, Ankara, Turkey. The thesis was prepared by the
The optimum weights and costs are shown in Fig. 5. Objective first author under the supervision of the second author. The contributions
function values gradually decrease as the backfill slope in- of J. E. Bowles and Prof. A. V. Keskar during the preparation of the
creases from 0° to 20°, and then they start increasing again, thesis are gratefUlly acknowledged.
approaching towards a slope of 30°.
Reflecting on the optimum design variables, X4 (thickness APPEND~. REFERENCES
of base slab), X6 (horizontal steel area of the toe), and X7 Alshawi, F. A. N., Mohammed, A. I., and Farid, B. J. (1988). "Optimum
(horizontal steel area of the heel) are insensitive to changes in design of tied-back retaining walls." The Struct. Engr., London, En-
backfill slopes in both minimization models. The other design gland, 66(6), 97-105.

JOURNAL OF GEOTECHNICAL ENGINEERING / AUGUST 1996/655

J. Geotech. Engrg. 1996.122:649-656.


"Building code requirement for reinforced concrete, with design appli- Keskar, A. V., and Adidam, S. R. (1989). "Minimum cost design of a
cations." (1990). ACI 318 -89, American Concrete Institute (ACI), De- cantilever retaining wall." Indian Concrete J., Bombay, India, 401-
troit, Mich. 405.
Bowles, J. E. (1988). Foundation analysis and design, 4th Ed., McGraw- Pochtrnan, Y. M., Zhmuro, O. V., and Landa, M. S. (1989). "Design of
Hill Book Co., Inc., Singapore. an optimal retaining wall with anchorage." Soil Mech. and Found.
Dembicki, E., and Chi, T. (1989). "System analysis in calculation of Engrg., 25(5), 508-510.
cantilever retaining walls." Int. J. Numer. and Analytical Methods in Rhomberg, E. J., and Street, W. M. (1981). "Optimal design of retaining
Geomech., 13, 599-610. walls." J. Struct. Div., ASCE, 107(5).
Fang, H. Y., Atsuta, T., and Chen, W. F. (1980). "Optimum design of Sanba§, A. (1995). "Optimum design of reinforced concrete cantilever
retaining walls on sloping hillside." 6th Southeast Asian Con! on Soil retaining walls by non-linear programming, MS thesis, Dept. of Civ.
Engrg., Taipei, 391-406. Engrg., Middle East Tech. Univ., Ankara, Turkey.
Gaylord, E. H. Jr., and Gaylord, C. N. (1990). Structural engineering Siddall, J. N. (1982). Optimal engineering design. principles and appli-
handbook, 3rd Ed., McGraw-Hill Book Co., Inc., New York, N.Y. cations. Marcel Dekker, Inc., New York, N.Y.
Downloaded from ascelibrary.org by University of Brighton on 07/09/14. Copyright ASCE. For personal use only; all rights reserved.

656/ JOURNAL OF GEOTECHNICAL ENGINEERING / AUGUST 1996

J. Geotech. Engrg. 1996.122:649-656.

You might also like