You are on page 1of 9

ELSEVI ER

S0141-0296(96)00012-0
Engineering Structures, Vol. 18, No. 10, pp. 76%777, 1996
Copyri ght 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0141~)296/ 96 $15.00 + 0.00
Curved gi rders are s peci al
Dann H. Hall
Bridge Software Development International, Ltd (BSDI), PO Box 287, 562 Thomas Street,
Coopersburg, PA 18036, USA
Hor i zont al l y curved I -gi rders have been used in hi ghway bri dges
f or over 30 years. Thei r st ruct ural behavi our is known t o be qui t e
di f f er ent f r om st r ai ght gi rders because of t he al ways present non-
uni f or m t or si on. Earl y st udi es of these member s wer e based on a
st r engt h of mat er i al s approach. Modi f i cat i ons of these resul ts wer e
made t o account f or di st or t i on and ampl i f i cat i on effects. Mor e
recent i nvest i gat i ons in Japan have i nvol ved i nel ast i c f i ni t e
el ement studi es. From t hese st udi es var i ous modi f i c at i ons of
st r ai ght I -gi rder bendi ng st r engt h equat i ons have been presented.
Bendi ng tests of curved beams are compar ed t o these equat i ons.
It is suggest ed t hat cur r ent research may lead t owar d t he t hought
t hat curved gi rders are t he general case, and st r ai ght gi rders may
be consi dered a speci al case. Copyr i ght 1996 El sevi er Sci ence
Ltd.
Keywords: hor i zont al l y curved gi rders, bri dges, beams
1. Introduction
Horizontally curved girders may be thought of as special
straight girders. At times a girder may be designed as
straight but brought into the world accidentally curved. To
account for this case, design specifications may be based
on an assumed out-of-straightness. It is possible also to
think in the converse sense--straight girders are only spe-
cial curved girders. Analysis of curved girders is different
from straight girders in that torsion due to curvature is
always present, whereas torsion may, or may not, be pre-
sent in straight girders.
Curved girders are special in appearance, fabrication
techniques and structural behaviour. The earliest curved
girders were probably made from rolled shapes which were
cold bent about their weak axis. When girder welding
became accepted, curved girders grew in popularity. Hori-
zontally curved girders are used in buildings, such as for
balconies. However, the most widespread use of curved gir-
ders is in the highway bridge market where high speeds
require smooth changes of direction.
Dabrowski performed some of the earliest analytical
study of curved girders t. His work was limited to strength-
of-materials assumptions; however, he discussed other
effects such as cross-section deformation. The earliest lab-
oratory tests of steel curved girders were performed in the
United States during the late 1960s and early 1970s.
Pennsylvania Department of Transportation and the Federal
Highway Administration financed much of this work 2. The
FHWA financing was provided through the Consortium
University Research Team (CURT) Project. This work
resulted in the Guide spe.cifications f or horizontally curved
highway bridges 3 which AASHTO printed in 1976. These
provisions were based on working stress design. Later pro-
visions based on load factor design were developed 4,5.
These provisions have never been accepted as full
AASHTO specifications and remain essentially unchanged
at this time.
The Hanshin Expressway Corporation of Japan adopted
guidelines for the design of horizontally curved girder
bridges in 19886 . These provisions are based on working
stress design although the factors of safety are applied to
stresses determined from ultimate strength analyses. The
Japanese have performed more recent studies of curved gir-
ders including tests and finite element investigations.
There are no other specific provisions for curved girders.
Often horizontally curved bridges are designed using vari-
ations of straight girder design provisions.
The behaviour of curved I-girder structures is discussed
in this paper. Bending moments from several predictor
equations are compared to tests of single curved I-girders
tested in bending. Finally, curved girder research underway
in the United States is discussed. Although both curved I-
girders (open) and box girders (closed) are utilized, this
paper addresses only I-girders.
2. Structural system
Figures 1 and 2 each show a two-girder, 160-ft simple span
structure. Girders in Figure 1 are straight, while those in
Figure 2 are curved to a 400-fi radius. Figure 3 shows a
similar structure except that the same offset from tangent
accomplished in Fi gure2 with a 400-ft radius, is
accomplished with a kink. The beams all have the same
769
770
Figure 1
Cur ved gi r der s are speci al : D. H. Hal l
Cr os s F r a me Spa c i ng - 20 f eet
" ~ ~ ~ ~ p ~ p p o r t s - Ver t i cal at e n d s of gi r der s
s p a . . , e e ,
Ra d i u s - I nf i ni t e _ .
Gi r der Spa c i ng - 10 f eet
Gi r de r De pt h - 6 f eet
t t
St r ai ght t wo- gi r der st ruct ure
Fi gure2 Curved t wo- gi r der st ruct ure
properties. They are singly symmetrical because the top
flange is smaller than the bot t om flange.
Table 1 gives selected results from finite element analy-
ses of each structure for a downward load of one ki p/ ft on
each girder. This loading can represent wet concrete of a
bridge deck. Reactions, mid-span deflections and bending
moments are presented in Table 1. Bending moment s in the
two straight girders are equal; however, the outside girder
in the curved and kinked structures carries the entire load.
The sum of the bending moment s in the two girders is
nearly equal in each case.
The interaction between the three pairs of girders differs
in significant ways. The bracing members in Figure 1 only
stabilize the girders, whereas they act as primary load carry-
ing members in the curved and kinked structures. However,
the offset, rather than the curvature, is the critical struc-
tural effect.
Cross frames introduce restoring torques to the girders,
Table 1 Reacti ons, def l ect i ons and moment s
Model gi r der Vert i cal Mi d- span Mi d-span
react i on def l ect i on moment
(ki ps) (i n) (K ft)
St r ai ght
Gi rder 1 80 3.18
Gi rder 2 80 3.18
Tot al 160
Curved
Gi rder 1 - 8 4.71
Gi rder 2 170 12.77
Tot al 162
Ki nked
Gi rder 1 14 6.10
Gi rder 2 148 6.63
Tot al 162
3292
3292
6584
- 547
7318
6771
- 1114
7862
6748
resulting in bimoments in the girders. The bimoments are
manifest in equal but opposite lateral bending moment s in
the top and bot t om flanges. Figure 4 gives lateral flange
bending moment s in the bot t om flange of the outside girder
for the two non-straight structures. When the girder has a
uniform curvature, lateral flange bending is distributed
along the girder in proportion to the vertical moment. When
the offset is created by a kink in the girder, the restoring
forces and lateral flange bending are concentrated near
the kink.
Figure 3 Ki nked t wo- gi r der st ruct ure
Cu r v e d g i r d e r s ar e s p e c i a h D. H. Hal l
3 O 0
~ _ . 2OO
I,
I ~o0
0
i---
z
LJ - I 00 . ~
0
~_ - 2 0 0
- 3 0 0
Figure 4
C U R V D
- - K I NK F CI ~'~
I
t S P A n
STRNG .IT \
' ~ \ "", ,
,. . "
S P A N
Bot t om f l ange l at eral f l ange bendi ng moment
2 13 23 30 32
0 -13 -24 / - 30 - 32
/
At eenterlin
\
I 0 -1 ~ 4 145
1 0 2 -5 -145
Figure 5 Cross f r ame f orces
Figure 5 shows the cross frame forces for the two offset
structures. Note the extremely large cross frame forces at
the kink. If the cross frame arrangement was modified such
that there was no cross frame at the kink, the lateral flange
bending moment at the kink would be extremely large
while the maximum cross frame forces would be miti-
gated substantially.
If deflections are small, an elastic small deflection analy-
sis that considers appropriate boundary conditions, bracing
members, girder bending and torsional stiffnesses can
adequately predict structural behaviour of curved structural
systems for design purposes. Figure 6 shows the exagger-
ated deflected shape of the curved model. Differential top
and bottom lateral deflections are often significant as well
as differential vertical deflections between girders. Designs
of curved bridges usually do not include consideration of
lateral deflections or tw:[st of the girders.
3. F r e e b o d y d i a g r a ms
Curved I-girders may be either singly or doubly symmetri-
cal. Testing and all of the theoretical research in the US has
involved only doubly symmetrical I-girders having equal
771
tension and compression flanges. This family of I-girders
has both the centre of gravity and shear centre at mid-depth
of the web. This discussion is limited to doubly symmetri-
cal sections, but the concept is equally applicable to singly
symmetrical girders with different size flanges. Figure 7
shows a free body diagram of a doubly symmetrical hori-
zontally curved I-girder between two torsional brace points
subjected to a near constant vertical bending moment.
Although the vertical bending moment effectively creates
non-collinear axial forces in the flanges, equilibrium
requires that torsion be created in the girder which in turn
causes restoring forces in the cross frames. Vertical as well
as lateral loads are created in the cross frames. The sum of
the vertical components represents the load transferred
from the inside girder to the outside girder required for
equilibrium.
If the effect of the web is ignored, the flange lateral bend-
ing moment, M~at, can be thought of as resulting from a
virtual uniform lateral load equal to the axial flange force
divided by the girder radius applied to the flanges. The
flange is assumed supported laterally by the cross frames.
Virtual loads are applied to top and bottom flanges in
opposite directions. The virtual loads are inversely pro-
portional to the girder depth.
The flange lateral moment is a function of the bracing
spacing squared. An approximate equation for the lateral
bending moment at brace points is given as equation (1).
This equation comes from the V-load method7:
MvL 2
Mlat- IODR (1)
where: My = vertical moment, L = unbraced length, D =
girder depth and R = girder radius.
The lateral flange moments at brace points tend to
+ M t I
Figure 7 Curved I -gi rder f ree body
+ M L
_ M L - M L
Figure 6 Def l ect ed shape of cur ved t wo- gi r der st ruct ure
772 Cur ved gi r der s ar e speci al : D H Hal l
. . o. . . . . . . . . . . .
Figure 8 Curved compression flange free body
restrain bowing of the flange due to the virtual load. The
net effect of the non-uniform torsion due to curvature is
always to increase curvature of the compression flange as
seen in Figure 8.
Figure 9 shows a tension flange free body, again ignor-
ing web effects. The virtual load tends to reduce curvature
while the lateral flange bending moment at the brace point
tends to restore it. The net effect on the flange can be to
either increase or decrease curvature of the tension flange,
depending on flange stiffness and curvature. Dashed lines
indicate deflected shapes in Figures 8 and 9.
3.1. Secondary effects
I f a strength-of-materials analysis were performed, it would
yield lateral bending stresses similar to those of equation
(1). However, there are important secondary effects in
curved girders.
A curved girder subjected to bending tends to deform
radially, causing both flanges to undergo increased bowing.
These deformations tend to amplify the curvature effect 8.
This in turn, causes a further increase in the lateral flange
bending in the compression flange and a decrease in the
lateral flange bending in the tension flange. As a result of
the radial effect, the magnitude of the lateral flange bending
stresses in the two flanges are no longer equal. The change
in curvature of the flanges due to the radial effect causes
the lateral bending moment s in the compression flange to
be amplified. The radial effect causes the cross frame forces
to be different on the compression and tension flanges of
a doubly symmetrical beam. The radial effect is evident as
soon as loading commences.
The lateral flange moment s are also amplified when the
curvature is increased due to lateral bending in the flanges.
The resulting increased lateral deflection leads to a larger
external moment. This effect has been observed experimen-
tally and theoretically to commence at approximately 30%
of the ultimate capacity of the section 8. These secondary
effects are thought to have an increasingly important influ-
ence on the strength of horizontally curved I-girders as the
curvature and flexibility of the girder increases 8.
Figure9 Curved tension flange free body
OUTSIDEOF
COMPRESSION
Figure 10

Split curved compression flange free body
I f the cross-section is assumed to retain its shape, a warp-
ing analysis could be performed using the strength-of-
materials method. In the case of bridge girders, such an
analysis would yield bi moment s and resulting lateral flange
stresses similar to those based on the virtual lateral loading.
Figure 10 shows a free body diagram of both halves of
a curved compression flange considering the web. The web
exerts a lateral force tending to restrain the flange from
bowing. It also restrains the web from buckling into the
plane of the web. The stress gradient in each hal f of the
flange is different. The lateral moment varies such that the
extreme fibre of the flange on the outside of the curve is
largest at brace points. On the inside of the curve, it is
largest mi d-way between brace points.
Forces can result in flange instability different from that
of a straight girder flange. Local buckling of curved flanges
is related to the stress gradient across the flange as well as
the average stress, curvature and width-to-thickness ratio of
the flange.
3.2. Web
A free body diagram of an unstiffened horizontally curved
I-girder web is shown in Figure 11. The compression edge
of the web bows outward while the tension edge of the web
is flattened if the potential effect of the tension flange is
ignored. In addition to vertical bending, the web undergoes
distortion about an axis parallel to the flanges as the flanges
attempt to move in opposite directions.
J
Figure 11
J
Curved web free body
Curved girders are special: D. H. Hall 773
The addition of transverse web stiffeners tends to reduce
web bending about a longitudinal axis. They are not likely
to have much effect on bowing about a vertical axis.
Attaching transverse stiffeners to the flanges has been
found to increase the ultimate bending strength of curved
I-girders 9.
The behaviour of longitudinal web stiffeners on curved
webs has not been studied in the US. However, they are
permitted according to the AASHTO Guide Specification.
Their stiffness must be increased over that required for a
straight girder web.
4. Stability
4.1. Single girder
Resistance to lateral torsional buckling of a straight girder
can be thought of as tile ability of the girder to resist a
slight lateral deflection. As long as the internal lateral bend-
ing moment in the compressi on flange is larger than the
externally applied lateral moment, the girder has not
reached bifurcation (buckling load).
When a curved 1-girder bridge is erected, a single girder
is sometimes set. During this time it may be torsionally
restrained by external means such as lateral restraint at the
bearings and lateral ties of the top flange at the bearings.
This arrangement is adequate if the compression flange is
stable. Otherwise, additional supports, either torsional, ver-
tical, or both are required. A method of determination of the
stability beyond those limits given in the AASHTO Guide
Specification is not available at this time.
4.2. Multi-girder
Figure 2 shows a simple span system consisting of two
curved girders connected by cross bracing and supported
by bearings at their ends. This structural syst em is stable
when the bracings between girders and the bearings are
capable of equilibrium. Stability is achieved from the inter-
action between girders since neither girder alone would be
stable. Failure of one of the girders would cause a signifi-
cant reduction in the stability of the syst em and probably
cause collapse.
The results reported are based on an elastic, small
deflection analysis. Since the two girders are the same size,
the outside girder is likely to yield first. As the outside
beam commences to yiel!d, its stiffness decreases causing a
drop in the portion of load it receives from the inside girder.
Thus, the rate of change in moment of the inside girder
becomes greater than in the outside one. The system may
have significant reserve strength beyond the total load
determined by elastic analysis to cause plastification of the
outside beam.
5. Bendi ng strength
Most structural design specifications provide strength pre-
dictor equations for individual members comprising the
entire structure. Horizontally curved I-girders are different
from typical structural raembers in that their behavi our is
dependent on the boundary conditions provided by bracing
members and attachment of these bracing members to other
curved members. However, predictor equations for curved
beams are based on member behavi our as for other types
of structural members.
Tests of the curved beams have been limited to doubly
symmetrical members and the predictor models also have
this limitation. Many of the tests are of single curved beams
with rigidly connected torsional braces. Application of the
test predictor equations based on these investigations
requires that assumptions be made regarding singly sym-
metrical members and the effect of boundary conditions
other than those studied. There have been several types of
curved beams tested, but each test arrangement has led to
its own set of problems when applying t hem to practical
structures.
Practical curved bridge beams are usually singly sym-
metric with different size tension and compression flanges.
Such girders do not have coincident centres of shear and
gravity. Composite multi-girder cross-sections even have a
shear centre above the deck. The applicability of the
research limited to doubly symmetrical girders to more
practical girders needs further investigation.
Upon application of vertical load, the compression flange
commences to bend outward between brace points. Thus,
the spacing between brace points is always greater than the
effective bracing length of the curved compression flange.
Typically, the effective distance between braces of a
straight girder, K, equals 1.0 because the flange can buckle
in an S shape with nodes at the brace points. Equation (2)
gives an estimate of gamma used to determine the effective
unbraced length for a partially braced curved beam flange
from the Hanshin Expressway Guidelines 6. The effective
unbraced length equals gamma times the spacing between
cross frames, L.
~/= 1.0-1.97qbl/3+4.25qb--26.3~b3 (2)
l
where: qb = ~, where: l
radius.
= unbraced length and R = girder
5.1. Web behaviour
The compression side of a curved web tends to bow, caus-
ing an increase in the lateral flange bending stress. Further,
the tendency of the flanges to bend in opposite directions
tends to cause the web to distort such that the flanges no
longer remain parallel to each other. Nakai 9 found that
attaching transverse stiffeners to the compression flange
improves the pure bending strength. Both the Hanshin
Guidelines and the AASHTO Guide Specification limit
transverse stiffener spacing to the girder depth. However,
the AASHTO Guide Specification 3 permits transverse stiff-
eners to be eliminated when the web meets slenderness and
shear stress limits.
5.2. Predictor equations
5.2.1. Culver-McManus. Predictor equations for
compact and non-compact sections that were developed are
used in the AASHTO Guide Specifications 3. The working
stress design provisions permit only non-compact sections,
in which the strength of these sections is defined as when
the section reaches first yield. The load factor design pro-
visions permit either non-compact or compact sections. The
strength of a compact section is defined as full plasticiz-
ation of the cross-section. A flange can be treated as com-
pact without the web being compact. Vertical bending
moments and bi moment s are to be determined by analysis
774 Cur ved gi r der s ar e speci al : D. H. Hal l
of the entire superstructure. Bi moment s need be determined
only at brace points. Only the compression flange is con-
sidered because in the development of the equations, dou-
bly symmetrical sections were considered. The tension
flange is never critical in these sections.
The lateral flange stress is limited to 0.5 times the verti-
cal bending stress. The unbraced length divided by the
radius is limited to 0.1 for both non-compact and com-
pact sections.
The strength equations are based on modifications of the
equation for lateral torsional buckling given in equation (3),
used by the twelth edition of the AASHTO Bridge
Design Specifications m:
Fbs = Fy (1-3A 2) (3)
where: Fy = yield stress of flange and:
Equation (4) shows the modifications to equation (3)
used to determine the average critical flange stress, Fbc, for
a curved compact section.
Fbc = FbsPbPw (4)
where:
- 1
job =
0 0 1 ;
0 3
Pw = 0.95 + 18 0.1 - -t fb pb(FbJFy)
l = unbraced length, b = width of compression flange, R =
radius of curvature, fw = lateral flange bending stress, and
fb = vertical bending stress.
Equation (5) shows the modifications to equation (3)
used to determine the average critical flange stress, Fbc, for
a curved non-compact section.
Fbc = Fb,mpw (5)
where:
Pb =
P W 1 - -
Pw2 -
1
l l
I + - - -
R bf
' ]
1- - ~b 1 - 7 ~ f
1/b
0.95 +
30+8000(0.1 _~)2
1 + 0.6 (fw/fb)
When fwlfb >-0: Pw = Pw~ or Pw2, whichever is smaller;
when fw/fb < 0; pw = Owl.
5.2.2. Nakai. NakaV l shows equation (6) for pre-
dicting the bending strength of curved girders, Mu. This
equation modifies the lateral buckling equation and web
bend buckling of straight girders found in the Japanese
Bridge Specification ~2. Since the Japanese specification is
based on working stress design, Nakai ' s equation is modi-
fied by removal of the factor of safety.
[ L f ]
Mu = 1.92 + 0.357 Ma (6)
where: L = unbraced length, Ma = moment ratio based on
the Japanese Bridge Specification, R = radius and
b e = compression flange width.
5.2.3. Fukumoto. Fukumot o 13 developed equation (7)
for the prediction of the strength of curved I-girders. This
quartic equation must be solved using an iterative pro-
cedure and is thus difficult for designers to use.
A484 - 1 + ~ A4+ 1 82
- 8 + 1 = 0
(7)
where:
Mu
a n d Pe ---- elastic buckling load of section, Mp = plastic
capacity, Me = elastic moment, d = section depth, tcf =
comp. flange thickness, bcf = width of comp. flange, L =
unbraced length, and R = girder radius.
5.2.4. Hanshin. The Hanshin Expressway Corpor-
ation developed Guidelines 6. Equation (8) is presented in
these Guidelines to predict the strength of curved I-girders.
It is a linear interaction equation between vertical and lat-
eral bending stresses. Since the Hanshin Guidelines, as well
as the Japanese Bridge Specifications, are working stress
design based, equation (8) has been modified by removal
of the safety factors. Lateral flange moment s are specified
to be comput ed by the V-load method.
L fw
- - + < 1 . 0 ( 8 )
(Fba) c F~ao--
where: fb = normal flexural stress, ( F b a ) = critical normal
bending stress, fw = lateral flange bending stress, and Fbao
= yield stress.
The unbraced length divided by the radius is limited to
0.2. The sections must be compact. However, the definition
of compact flanges in the Japanese Bridge Specifications is
more liberal than that in the AASHTO Guide Specification.
5.2.5. Comparisons The predicted bending strength
of Nakai ' s Specimen M2 is shown in Figure 12 by
AASHTO, Hanshin and Fukumot o are compared in
Figure l 3. Since the flange is non-compact by the
AASHTO Guide Specification, only the first yield moment s
are considered. I f the AASHTO plastic moment capacity
were used, the strength would be much higher.
Curved gi rders are speciah D. H. Hal l
0. 472" I
t
~.1.5"
I
0 . 4 7 2 " I
7 . 1 5 "
w = 0. 177 i n.
R = 1157. 5 in.
Fyf = 56.1 ksi
F y w = 46. 2 ksi
L = 78. 7 in.
Figure 12 Cross-secti on f or speci men ' M2'
Figure 13 shows a plot compari ng the bending strength
versus slenderness. The abscissa is the unbraced length div-
ided by radius (L/R). The solid line plots are for varying
L while R is constant at the value of the specimen. The
dashed line plots are for varying R while holding L constant
at the value of the specimen. Of course, the plots cross at
the L/R value of the specimen. The ultimate moment
capacity is non-dimensionalized by the plastic moment
capacity of a straight girder of the same scantlings and yield
strength. The lateral flange bending stress is comput ed by
the V-load met hod to be 0.56 times the vertical bending
stress in the test specimen.
The AASHTO solid line is terminated where L/R equals
0.1. The AASHTO dashed line is terminated where the V-
load lateral flange stress divided by the vertical bending
stress is 1.0. The Hanshin plots are terminated where
L/R=0.2. The Fukumot o plots extend over the entire
range. In no case is the strength a pure function of the
variable L/R. However, all these curves behave similarly
and they each underestimate significantly the test speci-
men strength.
r
0. 9
0.111
0. 7
0. 6
:
0 5
0. 4
0. 3
H AASHT O - vanat)Je t.
O-- O A A S HT O - Vanaol e R
H Har l snm - Varla01e L
O Q Hans hm - Vanaol e R
Fuk umot o - Var i abl e L
. " l F u k u m O W . Var i abl e R
4, Test Val ue
" ~ t .
~ " 7 . 7, "
Q " ~ " L ''O ''I 3 . "/3
0 2 ] I s , ~ , , ~ , , , o , ~ , t
o .1 [ t / R - o o e o j
0. 0 J"
O.0S 0. 10 0. 15 0. 20 0 2 S
L I R
Figure 13 Bendi ng st r engt h pr edi ct ons ver sus L/R f or speci -
men ' M2'
7 7 5
5.3. Component bending tests
Figure 14 shows a typical test arrangement used by Mozer
e t al) 4A5, to test curved girders in pure bending. Figure 15
shows a typical arrangement used by Nakai e t a l . 9, t o per-
form similar tests. Mozer ' s tests are in positive bending
while Nakai ' s are in negative bending. The bending span
is longer than the torsional unbraced length in Mozer ' s
tests. The test fixture restrains the girder torsionally and
vertically in Nakai ' s tests.
Table 2 gives a comparison of the test results for pure
bending compared to the Cul ver - McManus predictors for
both first yield and plastic moments. Some of the specimens
had rather slender flanges that would be defined by Cul ver -
McManus as non-compact. These specimens would be
designed by AASHTO for first yield. It is seen that the
plastic moment better predicts the strength of the test speci-
mens.
Table 3 gives a comparison of the same test results com-
pared to predictions using Hanshin, Nakai and Fukuoto.
The Cul ver - McManus provisions are limited to L/R less
than 0.1 and Hanshin provisions are limited to 0.2. It is
possible that the Japanese tests have excessive torsional
restraint that may cause the apparent high values compared
to the predictor equations.
All but the Cul ver - McManus predictors apply to only
compact sections. However, the definition of compactness
is more liberal than that used by AASHTO.
6. Shear
There have been no unique mathematical predictor models
for shear in curved girder webs. The total shear stress is
composed of bending and torsional shear. The web tends
to bow as well as buckle as in straight girders. Thus,
additional rigidity of the web stiffeners is required for
curved webs.
Culver and others report that postbuckling strength can
develop in curved webs. However, there is some fear that
deflections will be large compared to straight girders. As a
result, the elastic buckling value is used in the AASHTO
and Hanshin design specifications.
6.1. Bending-shear interaction
Interaction between shear and bending is ignored in the
AASHTO curved girder provisions because it is ignored in
the straight girder provisions when elastic buckling limits
are used for the web. Research has shown that the typical
moment - s hear interaction equations seem to give rather
good results for component tests of moment and shear ~.
7. Current research
The National Cooperative of Hi ghway Research Program
( NCHRP) of the Transportation Research Board initiated a
project ( NCHRP 12- 38) to update the present AASHTO
curved girder provisions in the Guide Specification for
Horizontally Curved Bridges. This work is limited to appli-
cation of the present state of the art. Special emphasis is
placed on the construction of curved girder bridges. A sur-
vey of curved girder bridges fabricated across the United
States is underway. Results from this survey will provide
insight into the range of curvature experienced as well as
how the bridges have been detailed. This work, including
a new specification ready for ballot by AASHTO Bridge
Committee, is scheduled to be compl et ed in 1996.
776 Cur ved gi r der s are speci al : D. H. Hal l
Figure 14
400, 000 lb. T I N I U 5 O L S E N -..,..]1 II
T E S T I N G MACHI NG
B - -
A m
n':n
I
' i l '
F
T E S T S P E C I M E N
W E S T END _ ~ _ , ,,
A - - -
\ ' ~ \ \ \ \ \ \ ~,
UO AO ~ ~ . . . . . Y . _ . . I .
i I S P E C , . E N . , - - , , . , U S S
[ 0 , . . . . . , , o . , , . ,
SECTION A - A \ 1"1 --S"EcI"EN E
~ ' ~ - ~ " \ II / / ~ - - = = = - - - ) - - . O C K E . s u ~ , o . r
E A S T ENO
~ . ~ \ l I o I wa l strut
~ C O L U M N
(lOB 11.5)
~/ / / / / / "
Cul ver-Mozer bending test arrangement
I . oacl
C a n t l l e v e r
I I I I k l l i r ' " ' " " " " - " * * ' , , , , m
5 u D D o r t
S u p p o r t
Loadi ng and s uppor t i ng devi ce for
bending test.
Figure 15 Nakai bending test arrangement
Table 3 Comparison of predictors to test data
Hanshin Nakai Fukumoto
Mu test
(K ft) Pre- Ratio Pre- Ratio Pre- Ratio
dictor dictor dictor
L1A 1830 1107 0.60 1999 1.09 1509 0.82
L2A 1830 1136 0.62 1993 1.09 1540 0.84
GI5 1377 651 0.47 1279 0.93 991 0.72
G08 2120 1167 0.55 2213 1.04 1923 0.91
M1 8098 7428 0.92 5657 0.70 7989 0.99
M2 7752 4653 0.60 6508 0.84 6169 0.80
M3 6131 3498 0.57 5356 0.87 5145 0.84
M4 7203 2618 0.36 7972 1.11 4261 0.59
M5 5902 2291 0.39 6790 1.15 3589 0.61
M6 6287 2311 0.37 6649 1.06 3558 0.57
M7 6547 2366 0.36 7176 1.10 3827 0.58
M8 2935 * 1451 0.49 1069 0.36
M9 5548 * 6934 0.25 3126 0.56
Table 2
data
Comparison of Cul ver-McManus predictors to test
*Equation not applicable (central angle exceeds allowable).
Mx- - Reference 15; Lx- Reference 3; Gx- Reference 14.
Mo t e s t
(Kft)
First-yield Plastic
Predictor Ratio Predictor Ratio
L1A 1830 843 0.46 1716 0.94
L2A 1830 872 0.48 1749 0.96
GI5 1377 453 0.33 1058 0.77
G08 2120 931 0.44 1995 0.94
M1 8098 6930 0.85 7844 0.97
M2 7752 3692 0.48 7062 0.91
M3 6131 2620 0.43 5718 0.93
M4 7203 * 4774 0.66
M5 5902 * 3995 0.68
M6 6287 * 4024 0.64
M7 6547 * 4338 0.66
M8 2935 * *
M9 5548 * 2575* 0.46
*Equation not applicable (central angle exceeds allowable).
Mx- Reference 15; Lx- Reference 3; Gx- reference 14.
The Federal Highway Administration has financed a five-
year basic research effort. This work is to assist in the
development of an AASHTO Load and Resistance Factor
Design (LRFD) code for horizontally curved bridges. The
project includes both analytical and experimental studies.
Open and closed sections are included.
The experimental work involves laboratory testing of
components and testing of I-girder and box girder bridges.
The I-girder bridge is planned to be a full-scale simple span
structure. Full-scale tests eliminate the need for compensa-
tory dead loads and are able to reflect accurately fabrication
influences such as initial distortions and residual stresses.
The proposed steel framing for the I-girder tests is shown
in Figure 16. This framing may be used to test some girder
components in pure bending and shear with bending. By
using a bridge as a test frame for components, ideal bound-
ary conditions can be obtained. Other component tests may
Curved girders are special: D. H. Hall 777
go'-O ALONG GIRDER 2
t ~ " 2 ~ ~ R I~LI" I11 I I -- r~auIUSI / " " , ~ l ~ , . . ~ I l
. < , , 5 - /
Uuffs ~ DECK--' ~/" ~ AT16 E j ~ /
PLAN
Figure 16 Three-girder test frame
be performed on single girder specimens which will have
flexible lateral braces aT: various locations.
Testing of a single curved member that derives part of its
strength from the entire system is complex. Further, curved
members in a stringer bridge receive load directly and may
receive load through the: bracing members from other strin-
gers. Analytical studies indicate that as one girder com-
mences to yield, its stiffness is reduced sufficiently to affect
the load distribution in the girders. The cross frame forces
do not necessarily increase concomitantly with load under
this condition. Instead of adding load to the exterior girder,
the sense of force in the cross frames may change and load
would be transferred to the next interior girder.
When the component tests are completed, a concrete
deck will be placed on the steel frame and the bridge will
be tested to failure.
8. Conclusions
Curved I-girder researc!~ has been directed toward making
adjustments to the straight girder predictor equations to pre-
dict curved girder strength. This approach has not been very
successful as evidenced by the wide range of predicted
values for the test specimens. To date, there has been no
comprehensive theory presented which explains the inelas-
tic behaviour of curved I-girders. Most work being done at
present revolves around inelastic finite element analyses.
The AASHTO and Nakai predictors involve the straight
girder as a basis. The Hanshin provisions are indirectly
based on straight girder provisions but deviate significantly
from them via the interaction equation for vertical and lat-
eral bending stresses.
Fukumoto' s equation is most unique and seems to predict
test results rather well. However, it is unwieldy from a
designer' s view. It may be more fruitful to develop a pre-
dictor equation for curved girder strength that can be
reduced to that for a straight girder when the radius
becomes infinite. Such .an equation may include parameters
such as Fukumot o' s equation.
Such an approach would then permit straight girders to
be considered as special curved girders.
References
1 Dabrowski, R. ' The analysis of curved thin-walled girders of open
sect i on' , Stahlbau 1964, 33 (12), 364-372
2 Culver, C. G. ' Design recommendations for curved highway bri dges' ,
Final Report for Researcher Project 68- 32, PENNDOT, Civil Engin-
eering Department, Carnegie-Mellon University, June 1972
3 AASHTO 'Guide specifications f or horizontally curved highway
bridges' American Association of State Highway and Transportation
Officials, Washington, DC, 1993
4 Galambos, T. V. ' Tentative load factor design criteria for curved steel
bri dges' , Research Report No. 50, Washington University, Depart-
ment of Civil Engineering, May 1978
5 Stegmann, T. H. and Galambos, T. V. ' Load factor design criteria
for curved steel girders of open sect i ons' , Research Report No 43,
Washington University, April 1976
6 The Hanshin Expressway Public Corporation, 'Guidelines f or the
design of horizontally curved girder bridges (draft )' , October 1988
7 V-Load analysis, USS Highway Structures Design Handbook, Vol-
ume 1, July 1981
8 McManus, P. F. ' Lateral buckling of curved plate gi rders' , PhD thesis
Carnegie-Mellon University, Pittsburgh, PA, 1971
9 Nakai, H., Kitada, T. and Ohminama, R., ' Experimental study of
bending strength of web plate of horizontally curved girder bri dges' ,
Proc. JSCE 1983, 15, 19-28
1O Standard specifications f or highway bridges, 12th Edn, American
Association of State Highway and Transportation Officials, Wash-
ington, DC, 1977
11 Nakai, H., Kitada, T. and Ohminama, R., ' Experimental study on
ultimate strength of web panels in horizontally curved girder bridges
subjected to bending, shear and their combi nat i on' , 1984 Annual
Technical Session - - Stability Under Seismic Loading, SSRC, San
Francisco, CA, 10-11 April 1984, pp. 91- 102
12 Specification f or highway bridges, Japan Road Association, March
1984
13 Fukumoto, Y. and Nishida, S. ' Ultimate load behavior of curved I-
beams' , J. Engng Mech., ASCE 1981, 109 (1), 192-214
14 Mozer, J. D., Cook, J. and Culver, C. G. Horizontally curved highway
bridges stability of curved plate girders - - P3, Federal Highway
Administration, August 1975
15 Mozer, J., Ohlson, R. and Culver, C. Stability of curved plate gir-
ders - - P2, Federal Highway Administration, August 1975

You might also like