Vous êtes sur la page 1sur 9

Preliminary design of precast, segmental box-girder bridges using optimization

CRAIGJ . MILLER
DP~(I~/IT
(<f'Ci~,il
I ~ I I /Et~gi~~ceritig,
C(I.>C>
We.~/crt7Rc,.\~r~,c
U t ~ i ~ ~ ( ~Ct ~l ~. \~i /~~~, ~ OH44106,
l(it1~1, U.S.A.
AND

JUAREZACCIOLY
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16
For personal use only.

Cn;.\.(t Posi(11593, So6 C(trlo.s,i t ~ ~ . 13560


\il
Received March 15, 1978
Accepted November 7, 1978
PI-ecast,prestressed segmental box-girder bridges are now accepted as an economical alternative fol-spans over 150 ft (46 m). Decisions about cross-sectional dimensions rnadecluringpreliminary design can have a substantial influence on the final cost of the bridge. To help the designer
obtain an economical startingpoint fora final design, a program was written to determine section
dimensions and midspan and pier prestressing steel areas to give minimum cost. Since a preliminary design is obtained, the analysis techniquesand design criteria have been simplified to reduce
computation. The design produced by the program will satisfy AASHTO specification requirements and the recommendations of the PC1 Bridge Committee. The optimization algorithm used
is the generalized reduced gradient technique. To demonstrate the program capabilities, three
example problems are discussed. The results indicate that optimum span-depth 12tios are
approximately 24 for the cost ratios used. The cost of the optimum design does not seem to be too
sensitive to the ~ x t i oof concrete cost to prestressing steel cost.
On admet maintenant que les ponts II poutres-caissons pl.ef:~briquees par voussoirs PI-econtraints constituent une variante economique pour des portees superieures II 150pi (46m). Les
cli~nensionscle la section transversale choisies II I'etape de I'avant-projet peuvent exercer une
influence significative sur le codt final de I'ouvrage. Pour aider le projeteur h s'ol-ienter d&sle
depart vers L I projet
~
Icononiiq~~e,
on a mis au point un programme permettant cle choisir des
dimensions de section et des bections d'acier de precontlxinte i mi-travee et sur appuis intermidiaires qui minimisent le cout. Puisqu'il s'agit cl'un avant-projet, les techniques d'analyse et
les critkres de chois ont ete simplifies de fafon h reduire les calculs. Le PI-ojetqui en resulte
satisfait aux prescriptions cle I'AASHTO et aux recommandations du Comite Ju PC1 (Ponts).
L'algorithme d'optimalisation utilise la technique generalisee de reduction du gradient. Trois
esemples illustrent les possibilites du programme. Les resultats indiquent que 1e rapporr optimal
porteelprofondeurest de I'ordre de 24 pour lescouts I-elatifsretenus. Le coGt du projet optimal ne
semble pas 6tre trop sensible au rapport des couts clu beton et de I'acier de precont~xinte.
[Tmduit par la revue]
Can. J . Civ. Eng., 6, 120-128 (1979)

Introduction
The trend in highway bridge structures has been
toward longer spans. Reasons for the trend include
traffic, safety, and aesthetic considerations. The prestressed concrete box-girder is currently thought t o be
an economical solutioll for spans over 150 ft (46 m),
because its high torsional stiffness provides excellent
transverse load distribution. In the United States,
cast-in-place box girders have been widely used in
California for spans in the 160 ft (49 m) range. The
precast, segmental box-girder bridge has been used
extensively in Europe for about 15 years. Spans range
from 150-350 ft (46-107 m) or more. Largely due to
the organization of the U.S. constructioll industry,
the precast, segmental technique has been slow to
take hold. In recent years, a number of these bridges
have been built, starting with a 200 ft (61 m) main
span structure over the Intracoastal Waterway near
Corpus Christi, Texas. Excellent sunlnlaries of the

state of the art in segmental bridge design and construction are available in papers by Muller (1975),
Libby (1976), and Breen (1976).
Decisions made during the preliminary design
phase of a segmental box-girder design regarding
such things as girder depth, segment length, and web
and flange thicknesses will have a significant illfluellce
on the final cost of the structure. F o r this reason, it is
important that careful attention be given to costs
during the prelimil~arydesign. To assist the designer
with that task, a computer program was written t o
optimize the cross section of a segmental box-girder
bridge using a ~ l o n l i ~ l e mathematical
ar
programming
technique. Based on simplified a~lalysesand a simplified set of design criteria, a preliminary cross-sectional
shape and steel areas are calculated t o minimize the
cost of the bridge superstructure.
There have been some previous attempts to produce such a program. The two that are specifically

03 15- 1468/79/0l0120-09S01.00I0
@ 1979 National Research Council of CanadalConseil national de recherches du Canada

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

MILLER A N D ACCIOLY

aimed at box-girder bridges are by Ramakrishna and


Powell (1974) and Lacey and Breen (1975). The
program written by Ramakrishna and Powell uses a
nonlinear programnling technique to produce a
minimum cost design for a ~nultiple-spanbox-girder
bridge of either constant o r variable depth. Because
the procedure does not allow cable anchorages at any
location but the girder ends, the program is not
suited t o the segmental bridge, which has anchorages
at every segment. The program is also not able t o deal
with the change in structural action that occurs when
the center closure joint is concreted t o transforn~a
pair of cantilevers into a continuous beam.
The work of Lacey and Breen (1975) deals with the
specific problem t o be dealt with in this paper: the
selection of key cross-sectional dimensions and prestressing steel areas t o produce a mininlum cost
design that can be used as the basis for a final design.
The design is based on ultimate loading, and no
checks on working stresses o r deflections are included
in the minimization scheme. The optimization is performed by treating the problem as unconstrained and
solving it using the Nelder-Mead algorithm. T h e
results obtained by Lacey and Breen indicate optimum span-depth ratios substantially higher than are
in use at present. This is because working stresses a n d
deflections are not considered.

L /2

- - L/2
I

FIG. 1. Typical three-span bridge.

whereas the rest are either fixed by other considerations, such as the total width of the bridge, or are
dependent on the design variables, such as the w ~ d t h
between inner webs in the double-box case. For the
bridge of two single-cell boxes, there a r e a total of 14
variables determined by the program, whereas there
are 12 in the case of the single-cell box. The web
thickness is included as a design variable at the
option of the designer. Most often t h e thickness of
the webs is determined by the amount of room
needed t o fit in the prestressing tendons and still
leave enough cover t o adequately protect them. If
the designer feels that would be t h e case in his
problem, he can fix the web thickness as an input
quantity. Otherwise, the program will determine it.
T h e objective function is the total cost per unit
length of the bridge superstructure. This is coinputed

Indicates Design Variable

Problem Description
The bridges considered in this paper are precast,
segmental box-girder bridges such as that shown in
Fig. 1. The bridge comprises three spans the lengths
of which are always in the proportions 1 : 2 : 1. T h e
reason for this is that the bridge is assumed t o be
erected using the balanced cantilever technique for
which these span ratios are ideal. The depth of the
bridge is assumed t o be constant throughout its
length. Current thinking (Breen 1976) is that the
constant depth structure will be the most econon~ical
solution for main spans in the 125-300 ft (38-81 m)
range. The only change in cross-sectional properties
is a change in the thickness of the bottom slab over
the supports t o provide additional conlpressive
resistance. The bottom slab tapers from a n~aximum
to the typical bottom flange thickness over some
length that is a variable in the optimization scheme.
Two different cross sections are considered, singlecell box and two single-cell boxes joined by a cast-inplace longitudinal joint.
The cross sections are shown in Fig. 2 ; Fig. 3 shows
a side view of the structure indicating the profile of
negative and positive moment cables and of the
bottom flange. In Figs. 2 and 3, the circled dimensions are those that are determined by the program,

L
I

I
mJ
8
I

1.

pier '
at midspan

FIG. 2. Bridge superstructure transverse configuration:


two single-cell boxes; (b) one single-cell box.

(a)

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

122

C A N . J . CIV. E N G . VOL,. 6, 1979

FIG.3. Bridge superstructure longitudinal configuration: (a) cantilever constri~ctionstage; (b) colnpleted s~lperstructure
configuration.

in a simplified way as: C = CcVc CpVp; C = total


cost/unit length; Cc = cost of concretelyd3; Vc =
volume of concretelft of bridge; C, = cost of pre~ ; V, = volume of prestressing
stressing ~ t e e l j i n . and
steel/ft. The cost of concrete, Cc, is assuined to
include the cost of mild steel reinforcing in webs and
slabs, formwork, material cost, and all other costs
except prestressing steel. The costs used in this work
are those used by Lacey and Breen (1975): concrete at
$144.00/yd3 ($188.00/m3) and steel at $0.70jlb
($1.54/kg). The costs were not updated because the
quantity that determines the design variables is the
ratio of the two costs, which was felt to have remained
relatively constant.

I n view of the preliminary character of the design


produced in the work described here, it was felt that
it would not make sense to use the analytical power
of the techniques described above. Based on the
recomniendations of the Prestressed Concrete Institute (PCI) Bridge Committee (1975), the transverse
analvsis considers tlie cross section t o be a fraiiie on
rigid supports at the lower corners. Loads are
positioned properly to obtain maximum values of
~ n o m e n t sand shears. Contrary t o the PC1 Bridge
Committee report, for si~nplicity tlie fillets at the
intersections of tlie slabs were ignored in the analysis.
Simple beam theory is used to analyze the bridge in
direction. This is in line with tlie PC1
tlie loneitudinal
Bridge Committee report and has been shown
Analysis
(Muller 1975) to be adequate except for cross secComplete analysis of a concrete box-girder struc- tions that are unusually wide compared to their
ture requires the use of sophisticated techniques depth. In tlie longitudinal analysis, the minor change
combined with the digital computer to account for in stiffness due t o thickening of the bottom slab over
all tlie complexities under load. Three such methods the supports is ignored, s o the structure is analyzed as
are the folded plate technique (Scordelis 1966), the a constant depth beam.
The AASHTO loads (American Associatioli of
finite segment technique (Lo and Scordelis 1969;
Brown and Burns 1975), and the finite element State Highway Officials 1973) for the H20-S44 truck
approach (Scordelis 1967). A number of programs and appropriate lane loading are used for analysis of
are available t o d o the analysis based on the above tlie completed superstructure. Influence lines are used
techniques with the program by Brown and Burns to positio~ithe loads for maximum effect. The loads
(1975) probably most useful since it is written considered prior to making the closure joint are the
specifically for analysis of segmental bridges during dead load of the section and a construction live load.
and after erection.
which is assumed t o be a concentrated load at the end

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

MILLER A N D ACCIOLY

of tlie cantilever. It is assumed that the segments are


lifted into place by a device that imposes 110 loads on
the structure, either a crane from below or a launcliing gantry supported directly by the bridge piers.
The secondary moments imposed on tlie structure
after the closure joint is completed and the positive
moment cables prestressed are accounted for in the
analysis by means of statically equivalent forces
applied to the structure. I n order to simplify the
analysis, it is assumed that the positive moment
cables can be lumped into a single cable the area of
which is equal to the total area required at midspan
and the profile of which represents an average profile.
This siniplification can be justified because the critical
sections to be checked during the optinlization are
not near the positive cable anchorages. The cable
profile is assumed to be a series of straight segments,
as can be seen in Fig. 3.

123

web thicknesses for construction reasons tend to


control, a simple check should be sufficient to
guarantee a reasonable design.

Optimization Scheme
The mathematical programming problem to be
solved can be stated as follows:
minimize.f(X),
subject to gi(X) = 0, i = 1, ..., 111,
ti 5 X i 5 ui,i = I, ..., n,
X = vector of design variables,
f(X) = objective function, in the case a t hand the
cost/ft of tlie superstriicture,
li, Lri = lower and upper bounds o n X i ,
11 = number of design variables,
IJI = number of constraints,
nz < 17.

The problem above is general, since inequality


Design Criteria
constraints, which are common in structural optimiThe design of tlie bridge superstructure can be
zation, can be converted to equality constraints by
divided into three phases: transverse design, longiintroduction of slack variables. In the proble~ndistudinal design during segment erection, and design of
cussed here, the constraints ensure that stresses and
the continuous beam that results when the closure
deflections d o not exceed specified values.
joint is made. The transverse design uses the
In principal, any nonlinear constrained optiniizaAASHTO method for the action of wheel loads. The
tion routine could be used to solve t h e minimization
working stress method with balanced reinforcing
problem stated above. F o r this work, two package
ratio is used. Shear stress is not checked; however,
routines available at Case Western Reserve Univerfor normal proportions it will not be a problem.
sity were tried, with both giving acceptable results.
Tlie longitudinal design during the erection of the
The first attempt was made using a penalty fi~nction
segments involves both working stress and i~lti~iiate
routine (Fox 1973), which has been extensively used
load checks. The stresses are computed at the centerin structural optinlization work at Case. The second
line of the pier and at a distance clL/2 (see Fig. 2) fro111
routine is based on the generalized reduced gradient
the pier. Because the joints between segments are
(GRG) algorithm (Lasdon er at. 1974), wliich became
assumed to have 110 reinforcing across them, the
available during the course of this work. Both
allowable working stress in tension is zero. Deflecroutines gave tlie same answer when applied to a test
tions at the end of the cantilever are calculated for
problem. However, G R G appeared t o run about 30:d
the combination of dead load, prestressing, and l i ~ e
faster than the penalty function routine and olTered
load.
advantages in dealing with upper and lower bounds
After the closure joint is made and the top cables
on design variables. F o r these reasons, it was used
prestressed, the ultimate and working stresses must be
exclusively to obtain the results that follow.
checked. Tlie critical sections are midspan, the pier
The fundamental idea of the G R G approach is to
centerline, and the maximunl moment section in the
use the constraint equations to express m of the
side span. The deflections at midspan are calculated
variables in terms of the remaining tz - 172 nonbasic
with respect to live load and impact so they can be
variables. This is also the basic idea of the simplex
checked against the AASHTO deflection criterion.
method of linear programming. If the vector of design
In all phases of the design, allowable stresses used
variables X is partitioned into a vector of basic
are those of the AASHTO specification. The only
variables, y, and a vector of nonbasic variables, x,
exception to this is shear on the section at the pier on
then
the mainspan side. Because no attempt is made to
x = (Y, x)
design mild steel, the ultimate shear stress is checked
allowed by the ACI design
against the ~iiaxin~unl
g(X> = SdY, x) = 0
specification (American Concrete Institute 1971),
13 ,E.In this way, the possibility of a design with an
unrealistically small web is avoided. Since minimum
Since we can now express y in terms of s by means

124

CAN J. C I V . ENG. VOL. 6, 1979

TABLE
1. Example problem data

Segment iength

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

Example
problem

Centre
span
(ft (m))

Side
spans
(ft (m))

Number
of
boxes

Bridge
width
(ft (m))

At pier
(ft (ni))

At
cantilever
(ft (ni))

No. of
segments
for one
cantilever

Width of
longitudinal
joint
(ft (m))

Number
of
traffic
lanes

T A ~ L2.EGeometric data
Thickness of slabs
Deck
slabs
(ft (mni))

Webs
(ft (mni))

Bottom
flanges
(ft (nim))

Depth of
girder
(ft (nl))

0.5
(152)

1.0
(305)

0.5
(152)

5.0
(1.5)

Slope of
bottom
flange

Width
of curb
(ft (mm))

--

Minimum

of the constraints, the objective function can be


expressed in terms of x only:

I .O
(305)

used here will give the same optimum from any


reasonable starting point. Each of :he example
problems was run starting at a variety of points,
including one design that \itas infeasible. The results
By this device, the nonlinear problem has been trans- of some of these runs are shown in Table 5. I n all
formed to a reduced problem with only upper and cases, the same optimuni was obtained regardless of
starting point. indicating that tlie problem is well
lower bounds: ~ninimizeF(s)subject to I,, 5 s _< rr,,.
Here I,, and rr,, are tlie vectors of 1o\.verand upper behaved and the program is reliable. Further conbounds on the variables in s.G R G solves the original fidence in the results can be gained froni the fact that
problem by minimizing a series of reduced problen~s, both optin~izationalgoritli~iisgave tlie same optimum
which can be solved more easily than a general, non- design.
The next characteristic investigated for these three
linear problem. T o solve the system of nonlinear
equations that must be solved to express y in terms of problems was the spa~i-depth ratio and its influence
s, a Newton iteration scheme is used. The reduced on cost. Each example problem was run with the
minimization problem is solved using the Goldfarb depths constrained, so that a curve of cost versus
variable metric algorithm. A detailed explanation of span-depth ratio could be obtained. Each polnt on
G R G and its implementation is contained in Lasdon the curve represents the optimum design for a given
et al. (1974). Details of the constraints are given in maximum or mini~nuln depth. I t can be seen in
Fig. 4 that the optimum span-depth ratio for the
Accioly (1976).
three example bridges is around 24, which is subExample Problems
stantially lowel than the 33 obtained by Lacey a n d
Three exa~npleproblems will be discussed. The Breen but significantly larger than t h e value of 18-20
important data for the three are shown in Tables 1-4. which is often seen in the literature. The curves in
The first example is of particular interest, since it is Fig. 4 indicate that the rate of increase of cost is
the design used for the first precast segmental box- greater when the span-depth ratio is increased than
girder bridge built in the U.S. I n any optin~ization when it is decreased. This is particularly true for the
effort, the first thing that must be done is to deter- two single-cell cases.
I t is also interesting to study the change in shape of
mine how reliable the technique used is for the
probleln a t hand. A reliable program in the sense the cross section as t h e span-depth ratio changes.

MILLER A N D ACCIOLY

TABLE
3 Material properties and allowable stresses

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

(a)
(b)
(c)
(d)
(e)
(f)

(g)
(h)
(i)
(j)
(k)
(I)

Ultimate strength of cables


Allowable stress of mild steel
M O ~ U I LofI Selasticity of m ~ l dsteel
Initial prestressing stress
Average prestress~ngstress after losses
Bottom cable anchorage and bendup points

f,'

= 270 ksi (1863 kPa)


f, = 20 ksi (138 kPa)
E, = 29 000 ksi (200 100 kPa)
(0.701L'
(0.86) x (0.70)f,'

center span
side span
(0.30)L/2
(0.40)L/2 from pier
anchorage points
bendup points
(2.5)cl
(2.5)rl from anchorage
where L/2 and d a r e one half of the center span and the depth of the g~rder,respect~vely
Concrete compressive strength
f,' = 6 ksi (41.4 kPa)
Allowable stress at service load
Compression
(0 4)L' = 2.4 ksi (16.6 kPa)
N o tension
Allowable con~pressivestress
(0.4)fe' = 2.4 ksi (16.6 kPa)
M ~ X I ~ Iallowable
L I I ~ shear stress
13 f l =1 ksi (6.9 kPa)
Modular ratlos
Strength con~putations
6
Deflect~oncon~putations
8
1 5 ~ n (38
. mm)
Concrete cover for reinforcement and p~estressingcables

TABLE
4. Loads and deflections limits
(a)
(b)
(c)
(d)

Superimposed dead load (pavement)


Live load plus impact during cantilever construction
Concentrated load for transverse design
Live loads considered

(e) M a x i n ~ u ndeflection
~
due to dead load plus prestressing at
the end of cantilever
(f) Maximum live load plus impact deflection at midspan for
continuous girder

y, = 0.017 kp/ftz (0.814 kPa)


PC= 25.0kp(111.2 kN)
PZ0= 16.0 kp (71.1 kN)

HS20-44 truck and lane loadings from AASHTO (1973)


specifications
A, = 3 in. (76 mm)
A, = L/800

TABLE5. Program reliability

Example
problem

'Depth
(ft (m))

Boxes
b2(b,)*
(ft (nl))

Cost
(Vft ($/nl))

*bl = sinsle box cells.


tDesign for Intracoastal Waterway Bridge in Corpus Christi, T X .
$Infeasible design.

Figure 5 shows the optirnum cross sections obtained is greater than 24.3, the figure shows t h a t the slope of
for various span-depth ratios for example problem I. the webs becomes nlore nearly vertical and the
The optimum span-depth ratio in this case is 24.3. bottom flange becomes wider and thicker. For the
When the depth is forced t o be smaller, so that L / d largest L / d shown, the slope of the webs has actually

126

C A N . J. CIV. ENG. VOL. 6, 1979

-= -. -.
E

b4 b4

-TOTAL COST
CONCRETE COST
---- CABLE COST

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

-.-.-

300

I::.
::

.. .

15

20

25

30

35
SPAN / DEPTH

FIG. 4. Optimum cost versus span to depth ratio for example


problems I , 11, and 111.

reversed, i.e., the webs slope toward the centerline of


the box from bottom to top rather than away from it
as they nornlally do.
For span-depth ratios less than 24.3, the section
becomes narrower at the bottonl with the bottom
flange getting thicker and thicker. At the smallest
L/d run, 16.7, which corresponds to a section depth
of 12 ft (4 m), the section has taken a nearly triangular shape with an extremely thick bottom flange.
Sections with L/d ratios close to the optimum appear
to have proportions that are visually pleasing and
similar to those used in practice.
Since the ratio of concrete to prestressing steel cost
could vary widely from one situation to another, it is
worthwhile investigating the effect on the optimum
design of various values of C,/C,. As would be
expected, when concrete is made cheaper relative to
steel, the span-depth ratio of the optinlum design
decreases. This may be one reason why the results
given above indicate higher span-depth ratios than
are currently thought to be optimum.
Another interesting aspect of cost variation to look
at is the change in cost that results if the design uses a

cost ratio that turns out to be substantially different


when the bridge is built. This can very easily happen
because the preliminary design will often be done
well in advance of bidding and construction. If we
consider example problem I with the design based
on Cc/C, = 137, and then evaluate that design with
cost ratios Cc/Cp = 103 and Cc/C, = 206, we find
that the opti~nunlcost based on Cc/Cp = 137 is
$550/ft ($1804/m) while the costs using that design
and either of the other cost ratios v a r y by less than
$I/ft ($3/m). This indicates that the cost of the
optimum design is not too sensitive to the cost ratio.
Accioly (1976) indicates for all of the example
problen~sit is only necessary to have the cost ratio
within f50% of the correct value to get a good
estimate of cost.
In all the example problems described, it was
assumed that the bottom cable bend points and
anchorage locations are as shown in Table 6. T o
show that the design that results is not affected
significantly by the choice of the cable profile,
example problem I was rerun with different profiles.
Table 6 gives the important nu~nericalresults. The
design does not seem to be sensitive to the location of
the anchorage and bend points. The change in cost
indicated in Table 6 is due to the change in volume of
the cables.
If a preliminary design prograill such as that
presented here is to be useful as a design tool, it must
be economical to use from the standpoint of time to
prepare input data and computer time. Since the
program can start with an infeasible design, it is not
necessary for the designer to spend a lot of time
finding a feasible starting point. However, running
time is adversely affected by starting with an infeasible design. All results presented here were obtained using a UNIVAC 1108 computer. The maximum running time for any of the examples was about
15 s of CPU time. At commercial rates for the 1108,
the cost including printing and card reading charges
would be about $6. Preparation of the input data
would require about half an hour of an engineer's
time. Based on these statistics, it is felt that the
program could offer some saving in the time and cost
required to get a good starting point for a final
design.

Canclusions
A program to find a minimum cost design of a
precast, prestressed segmental box-girder bridge has
been presented. Because of simplifications made in
the analysis and design, the resulting design is intended to be used as a starting point from which
detailed design can be carried out. The program
produces designs that appear realistic compared to
those used in practice, although the results appear t o

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

L/d
COST = $572.23/ft

= 33.33

COST = $544.55/ft

($1877/m)

(Sl786/m)

As, ~ 0 . 2 1 2ft2 (197 cm2)


As2 =0.207 ft2 ( 192 cm2)
~ s ~ = 0 . 0 4 7( f4t4~cm2)
0.815 f t
(24.8 cm)

COST = $495.21 / f t

($1625 / m )

Asl =0.194ft2(180 cm2)

(b)

As2 ~ 0 . 1 9 5ft2 (181 cm2)

As2=0.177 ft2(164 cm2)

As3=0.043 f t 2 ( 4 0 c m 2 )

~ s ~ = 0 . 0 4 7 f t ~ cm2)
(44

1.04ft (31.6 cm)

1.26 f t ( 38.4 cm)


L / d = 20.00

L / d = 22.20
COST = $ 499.08/ft

COST = $524.27 / f t

($1637 / m )

COST = $510.59/ft ($1675 /m)

COST = $539.10 / f t

($1720 /m)

0 1769 /m)

FIG.5. Cross-sectional configurations for example problem I: (a) configuration for Lid = 24.30 to 36.40; (6) configuration for L/d = 16.70 to 22.20.
TABLE
6. Effect of changing cable profile
Example
problem

*From Fig. 3.

a2 *

CI*

El2 and El3*


(deg)

62
(ft (m))

d
(ft (m))

(ft2 (cm2))

cost
(Sift ($im))

0 . 4 (L/2)

0 . 5 (L/2)

22

10.59
(3.23)

8.21
(2.50)

0.191
(1 77)

384.80
(1262.50)

A,

128

C A N . J . CIV. ENG. VOL. 6, 1979

Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Curtin University on 07/05/16


For personal use only.

indicate that somewhat shallower bridoes than are


currently used would be optimum. For the only
problem for which an actual design was compal-ed,
example problem I, the program produced a design
about 5 Y , less costly than the design that was built.
It is felt that this is the order of inagnitude of saving
that could be expected in the typical case. The
prograin is easily used, econoinical t o run, and can
probably produce a good starting point for final
design at a lower cost than is possible with traditional
design methods.

LACEY,
G. B., and BREEN,J . E. 1975. The design and optimization of segmentally precast prestressetl box-girder. bridges.
Center for Highway Research. The University of Texas at
Austin, Austin, TX, Research Report 121-3.
LASDON.L. S., Fox, R. L., and RATNEK,M. W. 1974. Nonlinear optimization using the generalized [reduced gradient
method. Revue Fr~uisaisecl'Autoniatiq~~e.Inforni;~liqueet
Recherche 0per;ltionnelle. 3, pp. 73- 104.
L I B B YJ., R. 1976. Segmental box-girder bridge supe~.struct~~re
clesign. Journal of the A~neric;tnConcrete Institute. 73, pp.
279-290.
LO, N. S.. and SCOKDEL.IS,
A. C. 1969. Finite segment analysis
of folded plates. ASCE J o ~ ~ r nof
a l the S t r ~ ~ c t ~ lDivision,
rill
95.
pp. 83 1-852.
MULLER,
J. 1975. Ten yehrsofesperience in precast, segmental
A c c r o ~ u ,J. 1976. Optimization of segmental precast, preconstruction. Journal of the Prestressed Concrete Institute,
stressed concrete box-girdel- bridges. Department of Civil
20, pp. 28-61.
Engineering, Case Weste1.n Reserve University, Cleveland.
PRESTRESSED
CONCRE-!.EINSTII.UT'EB R I D G ECOMMITTEE.
OH, Report No. 76-2.
A M E R I C AASSOCI;\I-ION
N
OF STATEH I G H W A,\KD
Y TRANSPOR- 1975. Tentative design and construction specifications for
precast. segmental box-girder bridges. Journal of the PreTA'rION OFFICIAL.^. 1973. Stantl>irclspecifickitions for highway
stressed Concrete Institute, 20, pp. 34-42.
bridges. I Ith ed. AASHTO, Washington, DC.
L.AV.,
. and POWELL.C. H. 1974. MSBOX-a
A ~ I E R I C ACN
O N C R E I IEN S . I . I . ~ U I1971.
~ E . Building code 1.e- R A M A K R I S H N
coniputel. program for automated design of prestressed conquirements for reinfol-ced concrete. Detroit, MI. ACI 318-71.
crete box-girder bridges. Universityofcaliforniaat Berkeley,
BRE~N
J . .E. 1976. Design of segmental bridges. Federal HighBerkeley, CA, Report No. UCSESM 74-18.94 p.
way Administration Bridge Meeting, Pennsylvania State UniSCORDELIS,
A. C. 1966. Analysisof simply supported box-girder
versity, University kirk, PA.
bridges. University of California at Berkeley, Berkeley, CA.
BROWN,R. C . , and BURNS.N. H. 1975. Computer analysis of
a l the S ~ ~ L I C ~ L Report
I I . ~ ~No. UC SESM 66-17.
segmentally erected bridges. ASCE J o ~ ~ r nof
1967. Analysis of continuous box-girder bridges. UniDivision, 101. pp. 761-778.
versity of California at Berkeley, Berkeley, CA. Report No.
Fox, R. L. 1973. Optimization methods for engineering design.
UC SESM 67-25.
Addison-Wesley Publishing Company Inc., Reading, MA.

Vous aimerez peut-être aussi