Beruflich Dokumente
Kultur Dokumente
Horizontally Curved
DAVID B. BEAL
JOHN G. RUBY
RESEARCH REPORT 30
ENGINEERING RESEARCH AND DEVELOPMENT BUREAU
NEW YORK STATE DEPARTMENT OF TRANSPORTATION
Hugh L. Carey, Governor/Raymond T. Schuler, Commissioner
The Engineering Research and Development Bureau conducts and administers the engineering research program of
the New York State Department of Transportation in cooperation with the U. S. Department of Transportation,
Federal Highway Administration, which provides financial assistance for research activities. This publication was
prepared in cooperation with the Federal Highway Administration; its contents do not necessarily reflect the official
views or policiesof the State of New York or the Federal Highway Administration, nor does it constitute a standard,
specification, or regulation.
Research Report 30
July 1975
iii
INTRODUCTION 1
A. Dead Load 4
6. Superimposed Dead Load 6
C. Live Load 6
D. Lateral Flange Bending 7
II. DISCUSSION 9
CONCLUSIONS 15
ACKNOWLEDGMENTS 17
REFERENCES 19
The basic objective of this study was to devise empirical expressions for the
transverse distribution of in-plane bending effects in composite horizontally
curved bridges. It was clear that the structure's behavior should be described,
if possible, in terms of the basic geometry available to the designer at the
earliest stages.
The approach used parallels that of Heins and Siminou (1) in their paper on
live-load distribution factors for horizontally curved structures. Factors
have been derived that relate the behavior of a curved system to that of a sin-
gle straight member. With this procedure, design of the horizontally curved
girder system is reduced to the familiar, comparatively simple design of indi-
vidual straight girders.
A curved girder analysis program (CUGAR I), developed at the University of Rhode
Island, was used in developing a preliminary design method. This program (5),
idealizes the structure as a planar grid and uses the stiffness method of analy-
sis to evaluate member deformations and internal forces. Full-scale field tests
performed by New York State (.2,_3,6) have shown that this analysis can be used
reliably to predict the response of simple- and continuous-span curved girder
systems under dead loads.
Several cycles of bridge design, analysis, and refinement were required to pro-
duce the final relationships presented here. Although initially the relation-
ships derived were dimensionally correct and had some theoretical basis, the
final expressions are entirely empirical and cannot individually be given any
physical meaning. Although the equations used are awkward, they reliably pre-
dict the behavior of curved girder bridges over a broad range of practical de-
sign parameters.
The method has been tested for a wide range of bridge geometries and has pro-
duced designs differing from the analytical solution by less than 3 percent.
Variables included and the practical range of each are as follows:
number of girders = 3 to 8,
3
From the Digital Collections of the New York State Library.
4 Design Method
length of the curved girder.) The basic factors are used to predict behavior
of the fascia girder on the outside of the curve. Factors for the other gir-
ders are obtained by apportioning the basic factors relative to the position
of the girder in question in the curved system.
A. Dead Load
F (1)
^ DL ^ ^LS
Where tLTc is the dead-load design moment of the equivalent straight girder
having a length equal to the developed length of the curved girder, and
(2)
(3)
where
c = 4(1 - a)
(N - I)2
and
(4)
where
(5a)
K3 = 1 + 9300(Lc)0-9(Rc) 2 9
- (5c)
(N - I)2 £
1 +
2(N + 1) R
Ki+ = (5d)
1 + *!
3 R
In these expressions,
N = number of girders,
S = girder spacing, ft
Note'that when n = 1,
f (6)
dl(1) f
dl
Thus, is evaluated in the design of the first girder and remains constant
for all subsequent girders of a particular structure. Because is dependent
on the weight of Girder 1, some iteration is required. Equation 2 defines a
parabola with a maximum value of at the longest girder and a minimum value
of 2 - at the shortest girder. The centerline value [n = (N + l)/2] is
equal to a.
M
SL (t1) ■ F
SL(n> M
SLS (7)
M
where SLg is the superimposed dead load moment in the equivalent straight gir-
der, and
F
SL(n) F
SL R
LL(n) (8)
in which
F = e (9)
SL
and
2,1
0.005 L
c
P = (10)
0.012
= 1 - (n - 1) (11)
(N - 1)
C. Live Load
•V"' - F
LL<n) "LLS (12)
Here is the design live load moment in the equivalent straight girder,
based on the AASHTO transverse distribution factor of S/5.5, and
F
LL^n) F
LL R
LL(n) (13)
2
0.012 L '
c
f 0.85 + 0.0005 L +
ll c
1.00 I"
0.95
0.0 0.7
1.00
0.95
0.0 0.2 0.3 0.4 0.5 0.6 0.7
l
/r
C. LIVE LOAD
0. 10
CEHTCRUNC LE«6T«
♦ 50*
• too'
* 125*
qc ^ i>^>ii.• y.^■■■■».!(MM■ ■ ■ ■ ■ i■ AA "y
— ISO'
f 200'
I QQ | I rrJfcimiliWii L.n.nnnr
- ^ >i|M« i/mi
■
0,95 mm t,>iiiiin"i>Hiii
♦
mm & -
0. 90
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
l
/r
Figure 1. Ratio of analytical to design moment for the longest fascia girder.
The ratio of analytical bending moment from the CUGAR program to the design
moment is shown in Figure 1A. For dead load, this ratio is within 3 percent
of unity in 14 of 21 designs and within 5 percent for 20 of 21. The extreme
value of 92.5 percent (conservative) is for a structure with a length of 100
ft and a radius of 200 ft.
A comparison of analytical and design dead-load values for two designs at the
limits of applicability of the design relationships is shown in Figure 2. Un-
derestimation of bending moment in girders at the inside of the curve is not
critical, because dead load accounts for only about 25 percent of total stress
in this case. The low percentage of dead-load stress results from the design
requirement that all girders in a structure have the same web depth. Because
of the large difference in bending moment between the longest and shortest gir-
ders, the flange size of the latter is reduced in thickness and width. The
result of this reduction in flange size is an increase in magnitude of lateral
flange bending stress. For the results in Figure 2 (L = 125 ft, R = 300 ft),
the lateral flange bending stress in Girder 5 is approximately 8 ksi.
For superimposed dead load, the ratio of analytical to design bending is with-
in 3 percent of unity for 18 to 20 design (Fig. IB). Its transverse distribu-
tion is shown in Figure 4A. The comparison of analytical and design values is
less satisfactory for live load (Figs. 1C and 4B). Design values are generally
conservative (except L = 50), the deviation being greatest for intermediate
values of central angle. These design values are inherently inexact because
the approximate straight-girder live-load distribution factor (S/5.5) is used
in the design to maintain the consistency of the design procedure.
The results shown in Figure 1C imply that the present design factor is too
large, and therefore in comparing live-load amplification factors their rela-
DESIGN
ANALYTICAL
2 3 4 5
GIRDER CENTRAL ANGLE
GIRDER
DESIGN
ANALYTICAL
12 3 4 5
GIRDER
25 —
Ppffipj AASHTO EQUATION GIVES SMALLER VALUES
1:1 PRESENT EQUATION GIVES SMALLER VALUES
0l 1 1 I I I
200 400 600 800 1000 1200
RADIUS, FT
tive magnitudes are significant. Figure 6 shows the values of span length and
radius for which one or the other equation gives the smaller value, and thus
with respect to the results shown in Figure 1C, the most suitable design equa-
tion. The present expression is preferable at all radii for spans less than
125 ft. For longer spans, the AASHTO equation seems appropriate at the shorter
radii.
Dead-load and superimposed dead-load reactions are reliably predicted for all
structures with central angles less than 0.2 rad. In contrast, live-load re-
actions are consistently overestimated but for central angles less than 0.2
rad, these values are also considered acceptable. Therefore, for shear, as
for deflection, a separate analysis is required for structures with central
angles greater than 0.2 rad.
A series of approximate equations have been presented for evaluating the inter
nal forces, deflections, and support reactions due to dead load, superimposed
dead load, and live load in horizontally curved simple-span bridges. Compari-
sons of design values with analytical results indicate that for structures with
central angles less than 0.2 rad, the preliminary design can be taken as the
final design without further analysis. For those with central angles from 0.2
through 0.5 rad, the preliminary design is adequate for strength, but should
be analyzed for deflection and reaction. When the central angle exceeds 0.5
rad, the preliminary design is an estimate only, and several cycles of analysis
and design may be required to achieve a satisfactory structure.
15
The study reported here was undertaken cooperatively by the Structures Design
and Construction Subdivision, Robert N. Kamp, Deputy Chief Engineer (Structures),
and the Engineering Research and Development Bureau, William C. Burnett, Direc-
tor. Much of the preliminary work on this design procedure was done by H. Dan-
iel Rogers, Jr., Senior Civil Engineer (Structures). Robert P. Murray, Senior
Engineering Technician, prepared input data for the analysis of trial designs
and tabulated the results of these computations.
17
19