Sei sulla pagina 1di 6

BuiM. Sci. Vol. 8, pp. 33-38. Pergamon Press 1973.

Printed in Great Britain

I (16) I

I (A3) I

Minimum Cost Design of Concrete Beams


with a Reliability-Based Constraint
S. S. RAO*

The optimum cost design of under-reinforced concrete beams is considered by


treating all the design parameters as random variables. The design is considered
to be safe and adequate if the expected moment capacity of the beam exceeds
the design moment capacity by a certain number of standard deviations. The
optimum cost design problem is cast as a nonlinear mathematical programming
problem and is solved as a sequence of unconstrained minimization problems.
Numerical examples are presented to illustrate the effectiveness of the method.
The resulting computer program is used to examine the characteristics of
optimum designs as effected by changes in the cost coefficients and the design
parameters.

1. INTRODUCTION

babilistic. In this paper, the optimum cost design


of concrete beams is attempted by considering all
the design parameters as random variables.
In probability based design methods, usually a
design is considered to be safe and adequate if the
probability of failure of the structural member is
less than a specified small quantity, say, 10-6[5,6].
Since the quotation of such small probabilities may
have little significance, some alternative proposals
have also been made. Among these, one proposal
for strength specification of structural components
is based on the mean values and standard deviations
of the uncertain strength parameters. In particular,
it has been suggested[7] that the design strength be
set equal to the expected value of the strength
minus a constant number of standard deviations
of the strength. This paper presents a method for
incorporating such an approach into the optimum
design procedures. As a simple example, the
minimum cost ultimate strength design of rectangular under-reinforced concrete beams is considered in some detail. The cross-sectional dimensions of the beam, the area of reinforcing steel and
the strengths of concrete and steel are considered
as random design parameters. It is to be noted that
the approach is quite general and can be adopted
for the optimum design of any concrete member or
structure.
The design problem is formulated as a nonlinear
mathematical programming problem. This formulation requires the determination of the mean and
variance of the ultimate moment capacity of the
beam. An approximate method, which does
not require the actual probability distribution
functions of the strength parameters, is used in
evaluating the mean and standard deviation of the
beam moment capacity. The constrained optimiza-

IT HAS BEEN proposed for some time that a more


rational criterion for structural safety is the
reliability, or alternatively, the probability of
failure. Probability applications in structural
engineering recognize that both loads and strengths
have statistical frequency distributions that must
be considered in evaluating safety. Since concrete
exhibits more variability in strength than most of
the conventional structural materials, the statistical
nature of its strength must be considered in the
design of concrete structures.
In reference [l], Shaw presented a method of
predicting the probability of failure of concrete
members by treating the resistance minus loading
t R - L ) , as a random variable with a known probability distribution. Sexsmith[2] presented the
reliability analysis of beams with and without
compressive reinforcement and square tied columns.
By assuming a generalized beta probability law
for the understrength parameter ~t (Rtest/Rtheory),
the author derived the parameters of the distribution function by making use of the available experimental data through the Bayes rule. In the above
two references, much emphasis has been placed on
the reliability analysis and no design example was
considered to illustrate the application of the
reliability principles. In reference [3], Hill presented
a method for the automated optimum cost design
of continuous reinforced concrete girders. Recently,
GoNe and kapay[4] considered the optimal design
of prestressed concrete beams. In both the references [3] and [4], the design parameters were
considered to be deterministic rather than pro=

* Assistant Professor, Department of Mechanical Engin-

eering, Indian Institute of Technology, Kanpur, India.

33

S. S. Rao

34

tion problem is solved by the method of FiaccoMcCormick[8] as a sequence of unconstrained


minimization problems. Numerical results are
presented by varying all the parameters affecting
the design problem.

Cs = cost of steel reinforcing rod per unit volume,


and
x}u~ = upper bound on the variable xj.
The expected ultimate moment capacity of the beam
is given by[9]

2. F O R M U L A T I O N OF THE

My = A ~ f v ( d - 0 . 5 9 Asfr]
fcb }

OPTIMIZATION PROBLEM
A classical mathematical programming problem
is one in which a multivariate functionf(X)(where
X is a n-dimensional vector consisting of x~,
i = l, 2 . . . . n) is to be minimized subject to given

Equations (4) and (5) represent the geometrical or


side constraints, which impose limits on the size
of the design variables. The upper bound on the
area of steel xtg,) is taken as the balanced steel area
given by[9]

constraints gflX)> O, j = 1,2 . . . . m. The function

.fc

f ( X ) is called the objective function and its choice


is governed by the nature of the problem. In the
present problem, this function is taken to represent
the cost of the concrete beam per unit length.
As per the comments made in the introduction,
the beam will be designed such that the expected
moment capacity M v will be greater than or equal to
the specified design moment M~ plus a constant
number of standard deviations of the moment
capacity irma. Thus the reliability-based design
criterion becomes

My> M~,+k trM~,

(1)

where k is a constant.
The objective function and design criterion being
known, the minimum cost beam design problem
can be cast as a mathematical programming problem once we choose the design variables. In the
present problem, the depth of the beam d, the
compressive strength of concrete fc, the yield
strength of steel fy, the area of reinforcing steel A t,
and the width of the beam b are taken as the five
design variables x~ (i = 1, 2 . . . . 5).
The optimization problem can now be expressed
as:
minimize

f(X) = 2(b+d).Ci+b.d.C~+A,.C ~

(2)

subject to

Mt~- Mr* --kaMc > 0

(3)

~j'")-xj =>O, (j = 1,2, . . . 4)

(4)

xk

(5)

> 0, (k = 1,2 . . . . 5)

where
C s = cost of form-work per unit area of concrete
surface,
Cc = cost of ready-mix concrete per unit volume,

(6)

x~ul = 0 " 7 5 x 0 ' 8 5 x k l . f ~ .

(870

b.d. \87"0 +/,,]

(7)

with
= t0'85

kl

for fc < 4.0 ksi

~0.85-O'05(f.-4.0) forf.>4.0ksi

(8)

The optimization problem formulated above is a


nonlinear mathematical programming problem
since the objective function, equation (2), the inequality of equation (3), and the last inequality
of equation (4) are nonlinear.
3. M E A N A N D S T A N D A R D DEVIATION
OF M O M E N T CAPACITY
The ultimate resisting moment of a beam
controlled by the failure of steel is given by equation
(6). It can be seen that M v is a function of the five
design variables x;. In a practical design, all these
variables are uncertain to a certain degree. Only the
grade of steel and concrete, the drawing dimensions
and some knowledge of construction inspection
and control are known. These uncertainties have
a considerable effect on the expected value and
standard deviations of the beam moment capacity.
If the exact distribution functions of the random
variables xi are known, the exact statistical distribution of M v can also be determined.
In practice, o n e can only hope to estimate the
means and coefficients of variation of the variables
effecting the moment capacity. As a result, only
the mean and the coefficient of variation of M v
can be determined with any confidence. A perturbation scheme can be adopted to derive the approximate expressions for the mean and standard
deviation of the moment capacity as follows.
Let the independent random variables xi,
i = l, 2 . . . . n be distributed around their mean

Minimum Cost Design o f Concrete Beams with a Reliability-Based Constraint

values xl, x2 . . . . ~, with standard deviations at,


tr 2. . . . a, respectively. Let M v = p (xl, x2 . . . . x,)
be an arbitrary function of these variables. The
Taylor's series expansion of p about the mean
values of the variables x~ is given by
p =

....

....

ap

i= l

(9)

dxi

+ higher order terms


Thus, if terms up to the first order only are considered, the arbitrary function p can be approximated by a linear function of the variables x~.
In this case, it can be easily shown that the mean of
p is given by
E[p] = p = p(21 . . . . ~.)

Op

constraint. The combination of these two terms is


called the ~b-function:
---,
---,
gp(X, rk) = f ( X ) + r k

~
1
~,
j=l gj(.~)

(I 3)

The vector of design variables corresponding to


the minimum value of the objective function is
found by carrying out a sequence of minimizations
of the @function for a decreasing sequence of
values r k (rk+l<rk). For any fixed rk, the unconstrained minimum of the ~b-function is found by the
Davidon-Fletcher-Powell method[lO]. In this
method, the successive steps in different directions
are taken as
----->

.__>

Xi+ 1 = X i + r*S~

(10)

(14)

where

and the standard deviation ~p by


v a r [ p ] = % =2

35

,. ..

. ,72,

(11)

Xi+l = design vector corresponding to the minimum of @function along the ith search

i=1

direction S;,

In the present case, since the function p = Mu can


be differentiated with respect to the variables x~,
equations (10) and (11) have been used in the design
optimization problem without the need for knowing
the distribution functions of the individual variables.
From equations (6) and (I 1), the standard deviation
of the beam moment capacity can be derived as

tr.u = [(frd_ l.18 A~f2~ .A~ . V


+ A s d - l'18"-s~Y| . f ;2. V32 +A2 f 2 . d 2. V~

A-(0"59)2(As'G)4

+(0.59)2 ( A j , p

The initial feasible points necessary for the optimization procedure have been found by a process of
trial and error.

Several examples have been considered in order to


illustrate the effectiveness of the method described.
In these examples, the cost coefficients are taken
as follows:
C f = $0.88/ft 2

V2

(b.f~)2 "

___>

X i = starting design vector for the ith step.

5. N U M E R I C A L R E S U L T S

A2/. \2

(b.f~)2

z* = minimizing step-length, and

]'

C, = $78.80/ft 3

(12)

where V~is the coefficient of variation of the variable


x~ and the right hand side of equation (12) is to be
evaluated at the mean values of the design variables.

4. S O L U T I O N OF T H E O P T I M I Z A T I O N

PROBLEM
The constrained optimization problem stated in
section 2 is solved as a sequence of unconstrained
minimizations by using the interior penalty function
method due to Fiacco and McCormick[8]. In this
method, the objective function, equation (2), is
augmented by a penalty term which becomes large
......)

as the vector of design variables X approaches a

Cc = $( l "O06+ O'031f~ + O.OO5f))/ft 3.

Since the cost of concrete depends on its quality


or grade, the cost coefficient C~ has been expressed
as a quadratic in the strength of concretef~.
An example has been worked out by taking the
design moment M* = 2500 kip-in., the number of
standard deviations k = 3, the coefficient of variation of the design variables Vi = 0.05 (i = 1 to 5)
and the initial design variables as d = 29.4 in.,
fc = 3.0 ksi, fy = 50.0 ksi, As = 4.0 in 2 and
b = 12.0 in. For this example, the cost of the beam
has been reduced from $11.064 to $6.885/ft by
the optimization program. The optimum design
vector is given by d = 25.6 in., fc -- 4.98 ksi,
fy = 54.89 ksi, As = 2.89 in 2 and b = 4.13 in.
This design has been considered as the reference

design for the subsequent computations.

S. S. Rap

36

In a real s i t u a t i o n , the sensitivity o f the o p t i m u m


design w i t h respect to the v a r i o u s design p a r a m e t e r s
will be o f g r e a t i m p o r t a n c e . T h e results o b t a i n e d
by c h a n g i n g the different d e s i g n p a r a m e t e r s are
discussed in the f o l l o w i n g p a r a g r a p h s . In m o s t o f
the cases, the c o n c r e t e s t r e n g t h f ~ , the yield s t r e n g t h
o f steel f y a n d the a r e a o f steel A , h a v e b e e n f o u n d
to be n e a r their u p p e r b o u n d s at the o p t i m u m
design.

c h a n g e s o f 6, 3 and
16 per cent in the
optimum
cost
thereby
indicating
the
pred o m i n a n t effect o f the f o r m i n g cost. T h e o p t i m u m
cost is least effected by c h a n g e s in the c o n c r e t e
cost.
(ii) Effect o f the number o f standard deviations
T h e effect o f specification o f different values for
k in t h e statistical c o n s t r a i n t , e q u a t i o n (3), can be
seen f r o m T a b l e 2. It is to be n o t e d t h a t specification
o f a v a l u e for k d e t e r m i n e s , indirectly, the p r o b a b i lity o f failure o f the b e a m . A s the n u m b e r o f stand a r d d e v i a t i o n s is increased, the b e a m has to be
d e s i g n e d for m o r e s t r e n g t h and we e x p e c t a corresp o n d i n g i n c r e a s e in the o p t i m u m cost. T h i s has been
e v i d e n c e d by the present results also.

(i) Effect o f cost coefficients on the optimum design


T h e i n d i v i d u a l costs o f f o r m i n g , steel a n d c o n crete are v a r i e d by + 2 5 p e r cent in o r d e r to find
the effect o f cost coefficients on the o p t i m i z a t i o n
results. T h e results are s h o w n in T a b l e 1. It can be
seen t h a t a 25 per c e n t c h a n g e in the cost o f steel,
c o n c r e t e a n d f o r m i n g , respectively resulted in

Table 1. Effect of cost coefficients on optimization results


Reference values :
k = 3, My* = 2500 kip-in., Vi = 0 . 0 5 ( i = 1. . . . . 5)
xl ~ = 30.0", x2 ~u~ = 5.0 ksi, x3 ~"~ = 55.0 ksi
Starting design vector for all cases:
x = 29.4", x2 = 3.0ksi, x3 = 50.0ksi,

x4 = 4.0in 2, x5 = 12.0"

Objective function*
Increase in cost

Reference design
Forming cost - 25 %
+25 /
/o
Steel cost
-25~
+25~
Concrete cost - 25 %
+ 25 ~
All costs
-25~
+25%

Initial

Optimum

11.064
9'546
12.582
10.516
11.611
10.363
11-764
8.298
13'830

6.885
5.787
7.940
6.451
7.257
6.664
7-102
5' 154
8'580

Optimum X
x2 -~ 4.98 ksi, x3 -~ 54.97 ksi

Increase
compared
to Reference
Design (%t

xj
(in.)

x-~
(in 2)

.v5
(in.)

0-00
- 15'93
15-31
- 6-29
5.39
- 3.22
3.15
- 25' 12
24.63

25.60
27.96
22.97
23.89
26'48
24.80
25.15
24.57
25.81

4-13
3-55
5-05
4-65
3.81
4-50
4.21
4.42
3.99

2 89
2.62
3.23
3.11
2.80
2-96
2.95
3'02
2.88

* S/ft.

Table 2. Effect of k lin equation (3)] on optimum design


All reference values except k and the starting design vector are same as those given in Table I.
Objective function corresponding to the starting design vector = 1 !.064.
Objective*
Optimum
Value of k

1
2
3
4
5
* S/ft.

6"263
6"557
6' 885
7'252
7"712

Optimum A7

Decrease
compared to
starting
design (%)

x~
(in.)

x2
(ksi)

x3
(ksi)

x,
(in 2)

x~
(in.)

43 "45
40-80
37"79
34"42
30'31

22-73
25'62
25"60
26'04
26"79

4'98
4"97
4'93
4"98
4"98

54'97
54"97
54-89
54'97
54"97

2"67
2"60
2'89
3'21
3'58

4'20
3-66
4" 13
4"43
4"73

Minimum Cost Design of Concrete Beams with a Reliability-Based Constraint


(iii) Effect of the design moment capacity
The results obtained by varying the design
ultimate moment M~ are shown in figures 1 and 2.
It can be seen that the optimum cost increases as
M~: is increased. Figure 1 indicates that the standard
deviation at the optimum increases nearly linearly
with M*. The graph between the variables d, A, and
b versus M* is shown in figure 2. From this figure,
it can be seen that the variable b increases rapidly
as M~ is increased.
900.0

o.....o,,

35<~

37
,'

x,~-30oin

d VS"I ~ , all V~"0~)

25~

/\
/
/

2o<

,;

b~r&M~,all'V, / ' /
-0-10
/

11-

8000

700.0

.=_
-~ 6 0 0 0

C.o

IOooo

150Oo

25000

3oooo

3~3Oo

,*~mme mame~ t:ap~ M,~, k~-~.


Fig. 2. Effect of M,* on d, b and A~ ( M . >-_ 114,*+ 3au.).

/-

400~

(iv) Effect of the coefficients of variation


The effect of changing all Vi from 5 to 10 per cent
can also be seen from figures 1 and 2. It can be
seen that, as the coefficients of variation increase,
the standard deviation and hence the optimum cost
(for any fixed value of k) also increases rapidly.

O'Mu~. I ~ , aNVi- 0-05

x.3-30 Oin.
x~ ) -5-0 ksi

x(~} 55"0 k$i

K30.O

lO000

15000

20000

25000

ultimate momlmt capodty M~,

Design

&5000

. 3000.0

kip-in.

Fig. 1. Effect of design moment capacity on CrMv and


optimum cost (M. > M.* + 3aM.)

(v) Effect of placing the upper bound on A s


The optimization results obtained by removing
the upper bound on the area of steel are shown
in Table 3. These results indicate that the optimum

Table 3. Effect o f varying the upper bounds on x,


i

All reference values are same as those given in Table 1.


Starting design vector for all cases:
xx = 2 0 ' 0 " , x 2 = 3.0ksi, x3 = 50.0ksi, x 4 = 4 " 0 i n

2,

x s = 12.0".

Objective function corresponding to the starting design vector = 8'789.


Optimum results
Upper bounds on
d
(in.)

25'0
.
30"0
.
25'0
.
25.0
.

f~
(ksi)

5-0
.
5-0
. .
6-0
. .
5.0
. .
.

* Active constraint.
t S/ft.

fy
(ksi)

A,
(in a)

Balanced
steel
area at
optimum

55'0

no
yes
no
yes
no
yes
no
yes

2"49
3"24
2"37
3.13
2.44
3.18
1.47
2.57

.
55"0
.
55"0
.
80.0
.

Steel
area

3-59
3"23*
3'38
3.12"
3.53
3.17"
2.71
2.55*

Objectivet

Decrease
compared
to starting
objective

(%)

6-752
7"193
6"745
7.163
6"531
6-722
6.150
7.138

23"177
18-163
23"254
18"502
25.630
23.511
30.023
18"779

38

S. S. Rao

cost of the beam increases if there is an upper


bound on A~, It can also be seen that when there is
no upper bound on A~, the area of steel is more than
the balanced steel area which means that the failure
of the beam is no longer governed by the failure of
steel. This shows the necessity of placing an upper
bound on A~ in all practical design problems. The
effect of changing the upper bounds on d,f~ and./~,
are also shown in Table 3.
6. C O N C L U S I O N S
The feasibility of designing concrete beams for
minimum cost with a reliability-based constraint
has been demonstrated. The approach is quite
general and is applicable to the design of any
concrete member or structure. The present method
has the potentiality of automating the design of
complex structures with relatively low cost with
the help of high-speed digital computers[l l]. A
rnore realistic method of design includes the control
of mix, field conditions of curing, local product

availability, etc. The author hopes to include these


considerations in his future work.
The problem of specifying small numbers of the
order of 10- 6 for the probability of failure has been
avoided by taking the statistical constraint as shown
in equation (I). This constraint can be directly
converted into a probability of failure constraint
if the distribution functions of the strength parameters are known. For example, a value of 1, = 2
in equation (I) corresponds to a probability of
failure of 0.0227, a v a l u e o f / , : 3 c o r r e s p o n d s t o a
probability of failure of 0.0013, etc., if the moment
capacit) M,, follows a standardized normal distribution.

Acknowledgement--The numerical results reported in this


paper have been obtained on Univac 1108 computer of the
Case Western Reserve University. The author wishes to
thank Prof. Fred Moses of the Division of Solid Mechanics,
Structures and Mechanical Design, Case Western Reserve
University, Cleveland, U.S.A. for his valuable suggestions
during the progress of the present work.

REFERENCES
I. H. C. SHAw, The rational probabilistic code format. J. Am. Concr. Inst. 6@ 690 (1969).
2. R. (3. SEXSMITH,Reliability analysis of concrete members. J. Am. Contr. Inst. 66, 413
(1969).
3. L. A. HILL, Automated optimum cost design of building girders, ACI Publication SP-I 6,
Computer Applications in Concrete Design and Technology, Paper SP 16-8 (1967).
4. G. G. GOnLE and W. S. LAPAY, Optimum design of prestressed beams. J. Am. Contr.
Inst. 68, 712 (1971).
5. F. MOSES and D. E. KtNSER, Optimum structural design with failure probability constraints. AIAA Journal 5, 1152 (1967).
6. F. MOSESand J. D. STEVENSON,Reliability-based structural design. J. Struct. Div., Proc.
Am. Soc. Cir. Engrs, 96, 221 (1970).
7. C. J. TURKSTRA,A statistical investigation of under-reinforced concrete beam moment
capacity, Structural Concrete Series No. 6. Dept. of Civil Engineering and Applied
Mechanics, McGill University, Montreal, Canada (1969}.
8. A. V. FIACCOand G. P. McCORMiCK, Nonlinear Programming, Sequential Unconstrained
Minimization Techniques. Wiley, New York (1968).
9. G, WINTER, L. C. URQUHART,C. E. O'ROURKE and A. H. NILSON, Design of Concrete
Structures, Seventh Ed. McGraw-Hill, New York (1964).
10. R. FLETCHERand M. J. D. POWELL, A rapidly convergent descent method for minimization. ComputerJ. 6, 163 (1963).
11. J. F. BRO'rCHIEand M. P. T. LtNZEY, A model for integrated building design. Build. Sci.
6, 89 (1971).
La conception de coot optimum de poutres de b6ton sous-renforc6es est consid6r6e
en traitant t o u s l e s param~tres de design comme variables au hazard. Le design est
consid6r6 sOr et satisfaisant si la capacit6 de moment pr6vue de la poutre d6passe la
capacit6 de moment de design d'un certain nombre de d~viations standard. Le probl~me de conception de coot optimum est 6tabli comme un probl6me de programmation
math6matique non-lin~aire et il est r6solu comme s6quence de probl~mes de minimisation sans limites. Des exemples num6riques sont pr6sent6s pour illustrer l'efficacit~ de
la m6thode. Le programme d'ordinateur qui en r6sulte est utilis6 pour examiner les
caract6ristiques de conceptions optimum relies qu'elles sont affect6es par des changements dans les coefficients de coots et les param~tres de design.
Es wird der Entwurf fiir Optimumkosten yon unterbewehrten Betontr/igern durch
Behandeln yon allen Entwurfsparametern als beliebig Ver/inderliche erwogen. Dex
Entwurf wird als sicher und ausreichend angesehen, wenn die erwartete Momentenkapazit~it des Tr/igers die Entwurfsmomenten-kapazit/i.t um eine bestimmte Anzahl
von Normalabweichungen tiberschreitet. Das Problem des optimum Kostenentwurfs
wird als ein nicht-lineares, mathematisches Programmier-problem geplant und wird
als eine Folge uneingeschr/inkter Reduzierungsprobleme auf ein Minimum geltist.
ES werden zahlenm/issige Beispiele zur Illustrierung der Wirksamkeit der Methode
gegeben. Das sich ergebende Computerprogram wird benutzt, die Eigenschaften fiir
Optimumentwurf zu untersuchen, wie sie sich dutch .~nderungen in den Kostenkoeffizienten und den Entwurfsparametern ergeben.

Potrebbero piacerti anche