Approximate Torsional Analysis of Curved Box Girders

by the M/R-Method
DAVID H . H . T U N G AND R I C H A R D S. F O U N T A I N

CURVED BRIDGES are becoming increasingly prevalent in 4. T h e thickness of each plate element is small as com-
highway construction because of improved geometric pared with its width, a n d in turn, the width is small as
designs a n d construction techniques. T h e current trend compared with the span length.
in this type of structure is to shape the girders so that
they follow the curvatures of the horizontal alignment, 5. Internal diaphragms are adequately provided, so
creating a continuous flow of the major structural that distortions of the cross sections will not occur.
elements. 1 6. Secondary stresses due to warping are considered to
While the appearance a n d structural efficiency are be negligible, which is justified when assumption 5,
often enhanced by the use of curved girders, the analysis above, is satisfied.
and design of these members are likely to be more com-
plicated, which, in some cases, m a y be the only major
DEVELOPMENT OF THE APPROXIMATE METHOD
factor that prevents the adoption of such a system. I t is (M/R-METHOD)
desirable, therefore, to develop approximate methods
which m a y help practicing engineers overcome this
Internal ForcesConsider an infinitesimal segment of
hindrance. 1 - 2 , 3
a curved girder (Fig. 1) for which three equilibrium
T h e objective of this paper is to present to design
equations m a y be formulated as follows:
engineers a simplified method for the torsional analysis
of single-span or continuous curved box girders, which, dV dV
by virtue of their excellent strength in resisting torsion, (1)
Rda dx
are generally recognized as ideal supporting elements
for horizontally curved structures. T h e accuracy a n d dM
(2)
limitations of the approximate method, as well as the Rda dx R
effects of the various parameters inherent in the problem, dT__ dT _ M _
are discussed herein. T h e results are then compared with (3)
Rda ~~ dx R
those obtained in exact solutions based on the transfer
matrix method. 4

ASSUMPTIONS
1. T h e dimensions a n d section properties of the girder
may vary in the spanwise direction. However, the cross
sections are symmetrical with respect to the vertical
^M
axis.

2. T h e curvatures may vary within each span, b u t are

not reversed in direction.
3. T h e line of bearings at each support is radial. T+dT
y z

(a) ( b )
David H. H. Tung is Professor of Civil Engineering, The Cooper
Union, New York, N. Y.
Fig. 7. Sign convention for external and internal forces

65

J U L Y / 1970
where
M = bending moment
R = radius of curvature
T = torsional moment -x

V = shear moment Developed length - L

p = distributed vertical load
t = applied torque
x = independent variable along the longitudinal axis Fig. 2. Torsional loading on developed girder
a = independent angular variable

W h e n a term carries two signs, the upper sign applies further and consider, as before, that in the torsional an-
to curves defined in Fig. l a and the lower to curves alysis the girder is straight and has a span length equal
defined in Fig. l b . to the developed length of the actual curved girder. T h e
Integration of Eq. (1) gives the well-known relation accuracy of the proposed method depends on the mag-
(just as in the case of a straight beam) that the change nitude of the central angle as well as the bending-
in shear forces between any two points on the girder is torsional stiffness ratio (EI/GJ) of the curved beam.
equal to the area of the load diagram between the same Since only the relative change in torsional moments
two points. Equations (2) and (3), however, are coupled can be determined from the corresponding (M/R t)
and the values of M and T cannot be determined as area, the torsional moment at any section can be com-
easily. For an exact analysis, one may first differentiate puted only after that at a reference section is known.
Eq. (2) or Eq. (3), and after making the substitution For a curved girder having a constant EI and GJ and
into the other, may proceed to solve the resulting second- torsionally-fixed at both ends of a span (but which
order differential equation. 5 otherwise may be continuous over several supports), the
W h e n the central angle a is small, and when the torsional moment at end A of span AB (Fig. 2), accord-
bending-torsional stiffness ratio lies below a certain limit ing to the approximate method, is
in the case of indeterminate structures, the bending
T
moments are not significantly influenced by the tor-
sional moments. One may therefore determine M ap-
proximately (but with sufficient accuracy for practical
* = i fX** - *)(L -x)dx (5)
which can be readily derived by the force method. T h e
purposes) by dropping the T/R term in Eq. (2), so that torsional moment at any section located at a distance s
from end A is therefore
^ = ^ = V (4)
(6)
Rda dx
T-T'-fX**-*)*'
In other words, the bending moments may be evaluated Equations (5) and (6) are analogous to the con-
closely by considering the curved girder as a straight jugate-beam method for the determination of deflec-
member with a span equal to its arc length, provided tions of beams. Thus, in the case discussed above, one
certain requirements are met. In fact, this approach has can consider a corresponding fictitious beam which is
long been adopted in practice, even though in some simply supported (because the angle of twist 6 = 0 at
cases its limitations are not fully recognized. both ends) and subjected to a distributed (M/R i)
T h e approximate method described herein for the loading (Fig. 3). T h e reaction at end A of the conjugate
torsional moment analysis of curved girders may be con- beam will then be equal to the torsional moment of the
sidered as a logical extension of the preceding solution real beam at A, and the shear at any other section will
for the bending moments. A similar procedure for the be equal to the torsional moment of the real beam cor-
approximate evaluation of torsional moments was also respondingly. T o satisfy Eq. (3), a positive (M/R i)
recently suggested independently in Ref. 6. quantity should be an upward load in the conjugate
Integration of Eq. (3) yields the basic relationship beam.
that the change in torsional moments between any two In most steel box girder bridges constructed in this
points on a curved girder is equal to the area of the country, it has been the practice to provide a bearing un-
(M/R ^ - d i a g r a m between the same two points. der each web of the girder, thereby rendering a torsionally
Whereas the torsional moments will be exact if the bend- fixed condition at every support. T h e approximate
ing moments have been computed in an exact manner, method will therefore be most useful in these cases. How-
the former may be determined approximately if approxi- ever, as demonstrated later, the method can be equally
mate M-values obtained from Eq. (4) are used in the applied to other situations in which point supports are
(M/R t)-diagram. Moreover, one may go one step present.

66

AISC ENGINEERING JOURNAL

 B
Load on conj. bm -(R Or

A ^<TTTT>*JS B
V ^
Ik
M - d i a g r a m of developed girder Conjugate beam

(a) (c)

e+de

Fig. 4. Sign convention for rotations

V - d i a g r a m of conjugate beam
t diagram of developed girder T - diagram of developed girder
In practical cases, 6/R2 is small as compared with
(b) (d)
the two remaining terms on the right h a n d side of Eq.
Fig. 3. Conjugate beam M/R-method (9), and can be neglected without significant effects.
Thus, Eq. (9) may be simplified as follows:

d2B 1 (M
D i s p l a c e m e n t s T h e longitudinal slopes and angles of
twist of a curved girder have a significant influence upon
dx2 ^ ]
EI\R) +
GJ[R <}
each other, and the vertical deflections are dependent
(10)
on both. Ell R GJ\ R )]
Since the equivalent straight-girder concept of bend-
Applying the conjugate beam analogy, one may con-
ing analysis may be extended readily to include the a p -
clude that for a curved girder with torsionally-fixed
proximate determination of vertical deflections, no
ends, the quantity Eld at any point of the real girder is
further attempts will be made herein to define the pro-
equal to the corresponding moment in the simply-
cedure, other than to examine the accuracy of the
supported conjugate beam under a distributed load of
approximate results.
M/R + (EI/GJ)(M/R - t). One may also note
In order to arrive at an approximate method of that the moment due to the second part of this load is
evaluating the angles of twist of a curved girder, examine equal to EI/GJ times the moment caused by the load
an infinitesimal circular segment as shown in Fig. 4. used previously in the approximate torsional moment
In the following discussions, downward deflections are analysis.
considered positive, whereas positive rotations are de-
W h e n t = 0, Eq. (10) reduces to
fined in Fig. 4, in which 6 = angle of twist, and 0 =
longitudinal slope of the girder.
T h e following equations may be written to relate 6 dx2
a n d <j>\ T h e quantity Eld is thus equal to (1 + EI/GJ) times the
moment in the conjugate beam under the distributed
J0_ _ dd _ 0 T_
(7) load of M/R. In other words, the numerical value of
Rda~ 7x ~ ^R GJ
the angle of twist in such cases is
dcj> __ d</> _ 6 M
l1-s(,+D<M>
(8)
Rda" dx ~
R EI <">
W h e n a term carries two signs, the upper sign refers to where w is the corresponding vertical deflection calcu-
the type of curves defined in Fig. 4a, and the lower to lated on the basis of the approximate bending analysis.
that in Fig. 4b.
S u m m a r y I n summary, under most conditions en-
Differentiating Eq. (7) with respect to x, and sub-
countered in practice (e.g., central angle ^ 3 0 and
stituting Eqs. (8) and (3) into the resulting expression,
EI/GJ ^ 2.5), the bending and torsional analysis of
d20 curved box girders may be uncoupled and investigated
dx2 ~
R~dx GJ~dx independently. By straightening the curved girder to its
full developed length, the bending moments and vertical
shear forces can be readily determined as customarily
(9)
R2 ?I\RJ
EI GJ\ R J done in the past. T h e proposed method suggests that

67
J U L Y / 1970
the torsional analysis can be carried out in a similar
manner, except that (1) a straight conjugate b e a m sub-
jected to a distributed load of ( M / i ? t) in the tor-
KN
sional moment analysis, and (2) a straight conjugate M-diagram

b e a m subjected to a distributed load of [M/R + P l k /ft (a)

(EI/GJ)(M/R t)] for the determination of the
angles of twist are to be considered, where M is the bend-
ing m o m e n t obtained in the approximate flexural Exact
analysis, R is the radius of curvature, and t is the applied Approx.

distributed torque in the spanwise direction. Since M/R

is a parameter in the forcing function in the analysis,
the approximate method is called the M/R-method to
differentiate it from the classical conjugate b e a m method
for the determination of b e a m deflections. Plan
T h e following sections will examine the accuracy and
limitations of the approximate method as well as several Fig. 5. Circular girder Fig. 6. Comparison of M-
other aspects related to the problem in general. with fixed-ends and T-diagrams in exact and
approximate analyses
EFFECT OF CENTRAL ANGLE a0
T h e accuracy of the approximate torsional analysis de-
pends, among other factors, on the accuracy of the ap-
proximate bending analysis; both are influenced to a always smaller and the + M ' s always larger than the
great extent by the value of the central angle a0 and corresponding exact values. Moreover, the points of
the bending-torsional stiffness ratio EI/GJ. inflection in a curved girder are located slightly farther
As an indication of the effect of ao, consider a single- away from the supports than those in an equivalent
span fixed-ended (with respect to both bending and straight girder of its developed length. Further studies
torsion) circular beam subjected to a uniform load of show that this is also true in the case of continuous
1 kip/ft (Fig. 5). Assume that R = 300 ft and EI/GJ = girders. These observations, therefore, might well be
2.5. taken into account when change of flange areas is con-
Figure 6 shows typical qualitative comparisons be- templated in curved girder designs by the approximate
tween exact and approximate bending and torsional method.
moment diagrams. T h e approximate torsional m o m e n t Table 1 shows a comparison of the exact and ap-
diagram is defined by third degree curves because the proximate results for aQ = 30 and 45 respectively,
approximate bending moment curve is a second degree while the percentages of error in Mmax, -{-Mmaz, and
parabola. It may be noted that the approximate M- T at 0.2L are given in Table 2 for various values of o.
diagram is always above the exact, so that the M ' s are In the latter tabulation, the torsional moments at the

Table 1. Comparison of Exact and Approximate Solutions for aQ = 30 and a0 = 45

o = 30 0 =- 45
x/L Exact Approx. Exact Approx.
M (kip-ft) 0.0 -2096 -2056 -4818 -4626
0.1 -984 -946 -2313 -2128
0.2 -116 -82 -348 -185
0.3 506 535 1065 1203
0.4 880 905 1916 2036
0.5 1005 1028 2200 2313
T (kip-ft) 0.0 8.3 0.0 59.5 0.0
0.1 -71.3 -77.5 -217.0 -261.6
0.2 -99.1 -103.4 -317.9 -348.8
0.3 -87.8 -90.4 -286.1 -305.2
0.4 -50.4 -51.7 -165.3 -174.4
0.5 0.0 0.0 0.0 0.0
x = distance from support
L = developed length of curved girder

68
AISC ENGINEERING JOURNAL
supports are also given. I t m a y be noted that in the EFFECT OF BENDING-TORSIONAL
conjugate beam analysis, the shear (and hence the STIFFNESS RATIO EI/GJ
torsional moment of the actual beam) vanishes at the T h e previous fixed-ended circular beam (R = 300 ft)
supports because the Af/i?-loading is self-equilibrated. is again considered in the following, except that ao is
It m a y be concluded from the above discussion that now maintained at either 25 or 30, while the value
for EI/GJ = 2.5, the approximate method is satisfactory of EI/GJ is varied in each case.
when ao ^ 30, and that the m a x i m u m torsional moments It m a y be concluded from Table 3 that the approxi-
obtained in the approximate analysis are always on the mate torsional analysis is acceptable for practical pur-
safe side for a fixed-ended curved beam. poses, provided that (1) EI/GJ ^ 4.0 when a0 ^ 25,

Table 2. Comparison of Exact and Approximate Results Showing Influence of a0

Bending Torsion
-Mmax +Mmax T a t 0 . 2 L = Tx T at Support = Ts
Error Error Error \TS/Ti\ .(100)
ao Exact Approx. (%) Exact Approx. (%) Exact Approx. (%) Exact (%)
10 229.0 228.5 -0.2 113.9 114.2 0.3 -3.81 -3.83 0.5 0.04 1.0
20 922.1 913.9 -0.9 452.1 456.9 1.1 -30.0 -30.6 2.0 1.12 3.7
30 2096 2056 -1.9 1005 1028 2.3 -99.1 -103.4 4.3 8.25 8.3
45 4818 4626 -4.0 2200 2313 5.1 -317.9 -348.8 9.7 59.5 18.7
60 8785 8225 -6.4 3779 4112 8.8 -704.9 -826.8 17.3 234.5 33.3
Notes: 1. AT and T in kip-ft
2. In all cases, Ts = 0 at the support in the approximate analysis
3. L = developed length of curved girder

Table 3. Comparison of Exact and Approximate Results Showing Effect of EI/GJ

(a) a 0 = 30
Bending Torsion

max +Mmas T at 0.2Z, 7 T at Support Ts

EI Error Error Error |r,/ri| -(ioo)
GJ Exact Approx. (%) Exact Approx. (%) Exact Approx. (%) Exact (%)
2.5 2096 2056 -1.9 1005 1028 2.3 -99.1 -103.4 4.3 8.3 8.3
3 2100 2056 -2.1 1000 1028 2.8 -98.4 -103.4 5.1 9.3 9.5
5 2116 2056 -2.8 984.5 1028 4.4 -95.9 -103.4 7.8 13.4 14.0
10 2149 2056 -4.3 950.3 1028 8.2 -90.6 -103.4 14.1 22.3 24.6
50 2286 2056 -10.1 808.2 1028 27.2 -68.4 -103.4 51.2 59.1 86.5
100 2350 2056 -12.5 742.1 1028 38.5 -58.0 -103.4 78.3 76.2 131.4
(b) ao = 25
Bending Torsion
~Mmai +Mmax r a t 0 . 2 L = T, T at Support = Ts
EI Error Error Error |7y:Ti|.(100)
GJ Exact Approx. (%) Exact Approx. (%) Exact Approx. (%) Exact (%)
1 1441 1428 -0.9 708.8 714.0 0.7 -58.9 -59.8 1.5 2.0 3.4
2 1446 1428 -1.2 704.5 714.0 1.3 -58.3 -59.8 2.6 2.9 5.0
3 1450 1428 -1.5 700.3 714.0 2.0 -57.8 -59.8 3.5 3.8 6.6
5 1457 1428 -2.0 692.3 714.0 3.1 -56.8 -59.8 5.3 5.5 9.7
10 1475 1428 -3.2 674.2 714.0 5.9 -54.4 -59.8 9.9 9.5 17.5
See notes in Table 2 for supplementary information

69
J U L Y / 1970
or (2) EI/GJ ^ 2 . 5 when a0 ^ 30. Such limits, although
arbitrary and subject to personal judgment, are neces- Exact

sary because the approximate method will give the same Approx.

answer for a given a 0 , irrespective of the value of EI/GJ.

As a consequence, spanwise variation of GJ can be
ignored entirely in the approximate analysis if, for
instance, the weighted average of EI/GJ is within the
above limits.
It may also be noted that in the case of a0 = 30, (a) Both ends fixed
the approximate bending moments will be of question-
able value when EI/GJ exceeds 10. For EI/GJ = 50,
the error is 1 0 . 1 % in Mmax and + 2 7 . 2 % in

EFFECT OF E N D RESTRAINTS
AND LQADING CONDITIONS
Five additional cases will be investigated to examine (1)
the effect of flexural end restraints, and (2) the effect of
(b) Both ends simply supported
two types of loading conditions. All of the beams con-
sidered are single-span circular beams, with R = 300 ft,
a0 = 30, and EI/GJ = 2.5. T h e supports are assumed
to be torsionally-flxed.
T h e cases investigated are:
1. A uniform load of 1 kip/ft over the entire span:
a. Both ends are simply supported with respect
to bending.
(c) Left end simply supported and
b. T h e left end is simply supported and the right right end fixed
end is fixed with respect to bending.
2. One concentrated load of 100 kips acting at the Fig. 7. Torsional moment diagrams showing effect of various end
mid-span: restraints in bending
a. Both ends are fixed with respect to bending.
b. Same as 1 (a).
c. Same as 1 (b).
approximate moment diagrams are linear), the general
While the torsional moment diagrams are defined shapes of both groups of diagrams nevertheless look quite
by third-degree curves in the case of uniformly-loaded similar when the support conditions are identical. For
beams (because the approximate moment diagrams are this reason, and for simplicity, the results are shown
second-degree parabolas), and by second-degree curves qualitatively in Fig. 7 without specifying the type of
in the case of concentrated-loaded beams (because the loading.

Table 4. Comparison of Exact and Approximate Results Showing Influence of Flexural End Restraints
Bending Torsion

-M (kip-ft) +Mnax (kip-ft) T, (see Fig. 7) (kip-ft) T at S-support (kip-ft)

Sup- AP- Error AP- Error AP- Error Ap- Error.

Load ports Exact prox. (%) Exact prox. (%) Exact prox. (%) Exact prox. (%)
Uniform F-F 2096 2056 -1.9 1005 1028 2.3 -99.1 -103.4 4.3 8.3 0 (-)
load = S-S 0 0 0 3175 3084 -2.9 -553.5 -538.3 -2.7
1 kip/ft S-F 3208 3084 -3.9 1720 1735 0.9 181.6 185.0 1.9 -264.3 -269.2 1.9
100 kips 1 F-F 2012 1964 -2.4 1937 1964 1.4 -1245 -1285 3.2 9.9 0 (-)
at mid- S-S 0 0 0 4019 3927 -2.3 -529.1 -514.0 -2.9
span S-F 3078 2945 -4.3 2426 2454 1.2 206.5 208.2 0.8 -251.7 -257.0 2.1
F = fixed with respect to bending
S = simply supported with respect to bending

70

AISC ENGINEERING JOURNAL

 Table 5. Comparison of Exact and Approximate Vertical Deflections and Angles of Twist
(EIw)maxX (10) ~3 (ElJe)max x (io)- 3
Error Error
Load Supports Exact Approx. (%) Exact Approx. (%)
Uniform F-F 0.203 R4 0.196 R4 4-3.4 - 0 . 6 6 5 R* - 0 . 6 8 7 R* 3.3
load = S-S 1.109 R4 0.982 R4 -11.5 - 3 . 6 2 5 R* -3.431 R* -5.2
1 kip/ft S-F 0.432 R4 0.408 R4 -5.6 - 1 . 4 2 1 R* - 1 . 4 2 6 i?3 0.4
100 kips F-F 0.770 Rs 0.749 R3 -2.7 - 2 . 5 4 4 R2 - 2 . 6 2 2 R2 3.1
at mid- S-S 3.379 R* 2.998 R* -11.3 - 1 1 . 0 6 #2 - 1 0 . 4 9 R2 -5.2
span S-F 1.415 R* 1.311 R* -7.4 - 4 . 6 6 R2 - 4 . 5 9 i?2 -1.5
F = fixed with respect to bending
S = simply supported with respect to bending

Tabic 4 gives an indication of the errors of the ap- central angles in case of variable curvatures, should
proximate solutions. In general, the discrepancies of the not exceed 90.
torsional moments are largest at the supports and reduce 3. If one or more of the supports are not torsionally-
progressively toward the mid-span. Except for the case fixed, it is further recommended that:
of simply-supported beams (with respect to bending), a. There should be at least one torsionally-
the errors in torsional moments are on the safe side. fixed support in the entire span.
Moreover, all the errors appear to be acceptable in b. T h e central angle (or sum of central angles)
practice. should not exceed 40 (32) between two
adjacent torsionally-fixed supports, nor 25
VERTICAL DEFLECTIONS AND (20) between a torsionally-free end support
ANGLES OF TWIST
and the first torsionally-fixed support, and
Table 5 gives a comparison of the exact and approximate the weighted average of EI/GJ should not be
solutions for the circular girders investigated in the larger than 2.5 (4.0).
preceding section. c. T h e central angle of the entire girder, or the
sum of all the central angles in case of variable
CONTINUOUS CURVED GIRDERS
curvatures, should not exceed 90.
T h e approximate method is applicable to the analysis of When all the supports are torsionally-fixed, the pro-
continuous curved girders. However, aside from the cedure of analysis will be similar to what has been de-
influence of the EI/GJ ratios, the accuracy depends to a scribed previously, i.e.:
great extent on (1) the total central angle of the entire
girder from one end to the other (or the sum of the 1. Straighten the entire curved girder to its full de-
central angles of all the spans when the curvature varies), veloped length and provide corresponding sup-
(2) the central angle of each individual span, and (3) ports. Determine the bending moments M by any
the torsional restraint provided at the supports. T h e fol- method of indeterminate analysis.
lowing recommendations, based on the findings of a 2. Considering one span at a time, apply the dis-
number of numerical investigations (but again subject tributed (.M/R t) loads on the simply-sup-
to personal judgment), are suggested in order to main- ported straight conjugate beam and determine the
tain the same degree of accuracy in the approximate torsional moments in that span.
analysis as previously determined for single-span girders. 4. T h e algebraic difference of the two end torsional
T h e simplified method will be valid even if there are moments adjacent to a support is the torsional
reversed curvatures in the continuous girders-, provided moment reaction at that support.
that they do not occur between any two adjacent
Sometimes it may be necessary or desirable to pro-
torsionally-fixed supports.
vide point supports for a continuous curved girder. (A
T h e limitations are prescribed as follows:
point support, for instance, may be in the form of a
1. T h e central angle of each span should not exceed single bearing placed directly under an internal trans-
30 (25) and the weighted average of EI/GJ in verse diagram connecting the webs of a box girder.) In
each span should not exceed 2.5 (4.0). such cases, the span of the conjugate beams described
2. If all the supports are torsionally-fixed, the central in step 2, above, should be between two adjacent tor-
angle of the entire girder, or the sum of all the sionally-fixed supports.

71

J U L Y / 1970
 T h e procedure outlined above will yield the same
bending moments in the girder whether the supports are
torsionally-fixed or torsionally-free. T h e bending mo-
ments, however, are not significantly affected by these
conditions.
Further, consider the case in which torsion is due to
concentric loads alone (i.e., t does not enter into con-
sideration). W h e n R and EI are constant in the con-
tinuous girder, the approximate method will give only Fig. 8. Three-span continuous curved girder
one set of torsional moments, irrespective of the degrees
of torsional restraints provided at the intermediate
supports, in that the torsional moment reactions will beams must also be equal, so that there cannot be an)
always be equal to zero at these points. This is neces- torsional moment reaction at the intermediate supports
sarily so because the bending analysis must satisfy the T h e results of two investigations are examined below
condition that the longitudinal slope of the girder must
be continuous at an intermediate support, which requires 1. Three-span continuous girder (Fig. 8): D a t a : R = 300 ft
that the end shear of two neighboring conjugate beams EI/GJ = 2.5; central angle of each span = 3 0 ; tota
adjacent to a support be equal under the M/EI loading. central angle of girder = 9 0 ; all supports are tor-
Moreover, recall that the end torsional moments are the sionally-fixed, but permit free rotations with respect tc
end shears of the conjugate beams under the M/R bending; movable uniform load = 1 kip/ft.
loading. Since both EI and R are constant, it may be Table 6 gives a comparison of the results obtainec
concluded that the end shear in these latter conjugate by the exact and approximate solutions.

Table 6. Comparison of Exact and Approximate Results of a Three-span Continuous Curved Girder
Error
Case Loading Condition M or T (kip-ft) Exact Approx. (%)
Bending
1 1 kip/ft over a b and cd +M at 0.4L from a 2519.9 2467.4 -2.1

2 1 kip/ft over a b and b e -iWatb 2992.9 2878.6 -3.8

3 1 kip/ft over b e -\-M at mid-span of b e 1862.4 1850.6 -0.6

4 1 kip/ft over a d -\-M at 0.4L from a 1992.8 1973.9 -0.9

M at b 2535.7 2467.4 -2.7

-\-M at mid-span of b e 549.8 548.6 -0.2

Torsion
1 1 kip/ft over a b and cd -Tata 439.2 430.7 -1.9

2 1 kip/ft over a b and b c + T at left of b 21.3 35.9 (~)

+ T at right of b 51.8 35.9 (-)
3 1 kip/ft over a b + T at left of b 246.7 251.2 1.8

+ T at right of b 265.5 251.2 -5.4

4 1 kip/ft over b e - T at left of b 225.4 215.3 -4.5

T at right of b 213.7 215.3 0.8

5 1 kip/ft over a d -Tata 324.9 323.0 -0.6

+ T at 0. 8L from a 228.7 228.2 -0.2

+ T at left of b 102.6 107.7 5.0

+ T at right of b 115.9 107.7 1.6

- T at 0. 3L right of b 40.3 47.4 (-)

72
AISC ENGINEERING JOURNAL
 Four loading cases are considered for bending, the
first three of which are critical loading conditions re-
quired to yield, respectively, a m a x i m u m moment at
each of the sections under consideration. T h e last case
deals with a uniform load extending over the entire
length, corresponding to the dead load condition.
It can be observed that the m a x i m u m error in bend-
ing moments occurs at the interior supports and is in the
Fig. 9. Two-span continuous curved girders
same order of magnitude as that of a single-span S-F
beam shown in Table 5. Similarly, the region of negative
moment extends farther from an interior support than
that in a straight girder of its developed length, and bending; uniform load = 1 kip/ft over the left span
should be taken into consideration in the design. only.
Five loading cases are considered for torsion. T h e T w o cases are investigated:
first case yields both the m a x i m u m torsional moment
a. Case 1: All supports are torsionally-fixed.
and the m a x i m u m vertical shear force at the end sup-
b. Case 2: T h e two end supports are torsionally-fixed,
port. T h e second case corresponds to the loading condi-
but the intermediate support is torsionally-free.
tion for m a x i m u m vertical shear at the two girder sec-
tions adjoining support b. It may be noted that the T h e results of the study are summarized in Table 7.
torsional moments are insignificant in this case. More- While the total central angle of the girder is 50 and
over, the errors, while considerably large in terms of EI/GJ is 3.0, so that this combination falls outside the
percentage, have no practical importance and therefore limits specified for good approximate results, it is
are not indicated in the table. As a comparison, the intentionally adopted herein to demonstrate the degree
m a x i m u m torsional moments at these two sections are of error to be expected in such cases, and to examine
given in cases 3 and 4. T h e fifth case again represents the effects of the two extreme torsional-restraint condi-
that of a uniform dead load. tions provided at the interior support.
T h e results summarized in Table 6 appear to be T h e results show that: (1) the bending moments are
acceptable for practical purposes. not significantly affected by the mode of torsional re-
straint encountered at the interior support, and (2) the
2. Two-span continuous girder (Fig. 9): D a t a : R = 300 ft; approximate torsional moments are closer to the exact
EI/GJ = 3.0; central angle of left span = 3 0 ; central values when the interior support is torsionally-fixed.
angle of right span = 20 ; total central angle of girder = In the latter case, however, the torsional moment at the
5 0 ; all supports permit free rotations with respect to two girder sections immediately adjacent to the interior

Table 7. Comparison of Exact and Approximate Results of Two-span Continuous Curved Girders
Bending (M in kip-ft) Torsion (T in kip-ft)
Case 1 Case 2 Case 1 Case 2
Error Error Error Error
x/L Approx. Exact (%) Exact (%) Approx. Exact (%) Exact (%)
Left 0.0 0 0 0 0 0 -376.8 -375.8 0.3 -367.4 2.6
Span 0.2 1604 1616 -0.7 1619 -0.9 -284.2 -282.5 0.6 -274.0 3.7
0.4 2221 2227 -0.3 2233 -0.5 -75.4 -72.5 3.9 -63.5 18.7
0.6 1851 1829 1.2 1838 0.7 146.4 148.7 -1.6 158.5 -7.6
0.8 458 424 8.0 436 5.0 277.8 275.4 0.9 286.2 -2.9
1.0 -1851 -1972 -6.1 -1957 -5.4 215.3 202.9 6.1 215.1 0.1
Right 0.0 -1851 -1972 -6.1 -1957 -5.4 215.3 231.3 -6.9 215.1 0.1
Span 0.2 -1480 -1589 -6.9 -1577 -6.2 99.0 107.0 -7.5 91.7 8.0
0.4 -1110 -1198 -7.4 -1190 -6.7 8.6 9.6 -10.4 -4.9 (-)
0.6 -740 -802 -7.7 -796 -7.0 -56.0 -60.3 -7.1 -74.3 -24.6
0.8 -370 -402 -8.0 -399 -7.3 -94.7 -102.3 -7.4 -116.0 -18.4
1.0 0 0 0 0 0 -107.7 -116.3 -7.4 -130.0 -17.2
Case 1: Interior Support is torsionally-fixed.
Case 2: Interior Support is torsionally-free.

73
J U L Y / 1970
support is more or less the average of the two exact the determination of beam deflections. T h e correspond-
values. It may also be noted that the errors in m a x i m u m ing fictitious beam in this case, however, is straight and
torsional moments in the loaded span are small and subjected to distributed loads expressed in terms oi
therefore acceptable. T h e torsional moments in the un- M/R. For this reason, the proposed method is denoted
loaded span, while exhibiting larger errors, are of no as the M/R-method to differentiate it from the classical,
practical interest because different loading conditions It is found that within the limits specified herein, the
will be required for maximum vertical shear forces in method will give results with sufficient accuracy for
that span. practical purposes.

SUMMARY
REFERENCES
Curved box girders have been recognized recently as
1. Highway Structures Design Handbook, Vol. 7, Chapter 72
excellent supporting elements in bridge structures which U. S. Steel Corporation, Pittsburgh, Pa.
must follow a horizontal curved alignment. T h e exact 2. Highway Structures Design Handbook, Vol. 7, Appendix C
analysis of these members, however, is often complicated U. S. Steel Corporation, Pittsburgh, Pa.
and tedious, unless a computer solution is readily 3. Gillespie, J. W. Analysis of Horizontally Curved Bridge;
accessible. AISC Engineering Journal, Vol. 5, No. 4, Oct., 7968.
4. Tung, David H. H. Transfer Matrix Analysis of Horizonta
This paper presents a simplified method for the ap- Continuous Curved Girders. Submitted to ASCE for consider a
proximate solution of torsional moments and angles of tion for publication.
twist in curved box girders of single or multiple spans. 5. Tung, David H. H. Analysis of Curved Twin Box Girde:
T h e box girders are assumed to be adequately stiffened Bridges ASCE Structural Engineering Conference, Seattle, Wash
by internal transverse diaphragms so that both warping ington, May 7967. Preprint No. 484.
6. Witecki, A. A. Simplified Method for the Analysis of Tor
and distortion stresses may be considered as negligible. sional Moment as an Effect of a Horizontally Curvec
T h e proposed method adopts an approach which is Multispan Continuous Bridge Concrete Bridge Design, AC
quite similar to the classical conjugate beam method for Publication SP-23, pp. 793-204, 7969.

Computer Program for Column Design

T o help meet the more sophisticated needs of today's engineering profession,
a computer program based on the 1969 A I S C Specification is now available for the
design of steel columns. This new design aid offers a valuable means to utilize the
computer to speed selection of the most efficient, economical column sections for a
structure a n d to reduce design costs.
T h e program was developed by A I S C in cooperation with the Committee of
Structural Steel Producers and the Committee of Steel Plate Producers of American
Iron and Steel Institute.
For further information, contact your local A I S C regional engineer, or write
to AISC, 101 Park Avenue, New York, N. Y. 10017.

74
AISC ENGINEERING JOURNAL