Académique Documents
Professionnel Documents
Culture Documents
The deflection theory first published by Melan in 1888 j gravity stiffness and therefore leads to increased options
is a fairly accurate hand method for the structural analy- in the design process. The purpose of this paper is to
sis of suspension bridges. The general tendency over extend deflection theory so that it can be used to analyse
this century has been towards building bridges with rela- multicable configurations and derive some of their struc-
tively weak stiffening trusses for which other hand tural characteristics.
methods are less applicable and so deflection theory
has been given prominence amongst analytical tech-
Deflection theory formulae for a multicable span
niques. 2,3 In the past, deflection theory appears to have
been used in a computational role, the need for which The span of a suspension bridge will be considered to
has been eroded by the availability of digital computer be of multicable form if, when viewed in profile, there
techniques which use discrete rather than continuous is more than one cable visible and these cables are not
idealizations of the structural form. ~-7 Thus, in two parallel to each other. Thus, in Figure 1, spans (a) and
recent books on cable structures, s'9 which include sus- (b) are of classical form whilst (c)-(h) are multicable.
pension bridges, deflection theory has been mentioned For the purposes of this analysis, cables of the same
only briefly. It may, however, be used to investigate profile (as in Figure l ( b ) ) will be treated as if they were
the importance of different parameters on the structural a single cable. It will be assumed that each cable is con-
characteristics of suspension bridges. ~0 nected to the deck by means of closely spaced vertical
In modern long-span suspension bridges, most stiff- hangers and the normal deflection theory assumptions
ness derives from gravity effects in the cable. This gra- will be made.
vity stiffness is only sensitive to span lengths, cable Since the forces in hangers are assumed to be at all
support conditions and the horizontal component of ten- times vertical, the horizontal component of tension in
sion in the cables. The weight of the suspended structure each cable will be constant across the span. Also, since
only affects stiffness indirectly through its effect on the the hangers are assumed to be closely-spaced, equili-
tension in the cables. It is thus difficult to increase stiff- brium of an incremental section of cable i under the
ness for classical configurations without increasing the dead load condition yields:
size of the cables so that they can carry extra tension
d 2yi
load. Brown has suggested to the author the possibility wi = - H i - - ( 1)
of investigating the use of multiple cables to support dx 2
suspension bridges in such a way that the cables have
different profiles. This possibility does not appear to where y; is the vertical coordinate of cable i, w; is the
be discussed in the main texts on cable supported bridges dead load per unit span at spanwise position x, and Hi
although at least one such bridge at Chungking, China is the horizontal component of tension in cable i (Figure
has been built. This gives a method of increasing the 2). Here y; is measured downwards from a datum. If
0141-0296/87/02084-11/ $03.00
84 Eng. Struct. 1987, Vol. 9, April 1987Butterworth & Co. (Publishers) Ltd
Cable profiles for suspension bridges: A. Jennings
a b
a
c d
Wl W
W2
e f I
v~3
I1 12 13
L
w is the total dead load per unit span and H is the total
X
horizontal component of cable tension:
dy,
w = Ewi (2)
H = .~Hi (3)
Hi
If a superimposed load of p per unit is applied
to the deck, vertical equilibrium of cable i gives:
d2
qi = - (Hi + hi) - ~ (Yi + v) (4)
ofdeck ~
i, . . . .
,i/
!/ d4v d2v wihi
E l - 7 - 7~ux- ( H + h ) -;--~ox.= p - "Z H--T (7)
f*=T
I;L ,, sec -~tO,.dx
1 / - ~
kw-- ~
w + 2~
w--
w
w+
V, = 0 (18)
V 0
where:
Ci = sec2 &idx
0 v, = Jj vdx (19)
d2~7 - P (22)
dx 2 H z z 7
d2~ S A~'-~. 6 ~r C
- (23)
dx 2 H ! K /~J-
Similarly, working from the right-hand end for segments Figure 7 Gravity stiffness deflections of span having three
which are not loaded: piecewise parabolic cables
-v, l, (26)
(=<&-') - 4 0.08 f..~..~jj Classical /71><~Two-cable
-4 / /
/ / / "'x/~/ \ /Three-cable
li( l'-4~j] (27)
%' = 4 \lj - 3 ~ /
0.040 ~ , /_ _, __ z _--
Hence:
a
H8 z3(y + 3y) + y3y(fi:~ + 2-9)
(32)
P [:[r- + 4l(v + ~ + 12v~]
i /
~W which may be used to compute the gravity stiffness de-
I flection under a concentrated load as it moves across
W2
~___~ the span of a multicable bridge with piecewise parabolic
cable profiles.
Figure 7 shows gravity stiffness deflections computed
, L
using equations (30) and (32) for a three-cable span
b
in which lj : l2 : 13 = 4 : 3 : 4, subject to a concentrated load
b a t x / L = 5/11.
Figure 8 shows the graph of deflection under a concen-
trated load as it passes across the bridge taking account
of gravity stiffness only. In general, deflections are
! % reduced in magnitude by factors of between 1/2 and
/----7 1/3, as compared with those for the equivalent classical
span. The ratio ll/l 2 (assuming that II = 13) can be used
to modify the structural characteristics. The most appro-
lection - v priate value for this ratio will depend on what property
it is most important to optimize. Since the deflection
diagram for a concentrated load acting at a point is also
the influence line for deflection at that point, it can be
i 0 -x ~ L deduced by inspection of Figure 7 that the effect of mov-
5 ing from a classical to the multicable configuration is
C even more beneficial when considering deflections due
Figure 9 Two-cable s p a n with cubic profiles for cables: (a) geo- to distributed loads. However, deflection theory is
m e t r y a n d i m p o s e d load; (b) d e a d load distribution; (c) gravity unlikely to give simple analytical expressions for the
stiffness deflection d u e to c o n c e n t r a t e d load largest gravity stiffness deflections due to distributed
loads acting on multicable spans. Figure 8 also shows
the deflections under a concentrated load acting on a
and right-hand ends of the segment respectively, and two cable bridge having piecewise parabolic cable pro-
let y = z + y and y = g + ? . The subscript j will be files with Ii = 12 = El2. In this case, the deflections are
deleted throughout. In the deflection diagram, Figure similar to those of a classical span if the loading acts
6, let 4~ and & be the angles E A G and FCG, respec- at centre span. In general, similar deflections will arise
tively. The condition that the curvature of E M G equals in every case of symmetric loading. Reduced deflec-
that of G N F gives: tions, however, are obtained for asymmetric loading
cases.
(28)
z2 22 Cubic cable profiles
A simple non-segmented configuration is a two-cable
The area of H J L G K must equal the sum of the areas bridge in which dead load applied to each cable varies
of E M G K and G N F L in order to satisfy the V = 0 condi- linearly with spanwise distance according to:
1 2y ~ a-- 6 ( y + 3 v ) + (y+3-9)
w2= 2
(29)
By substituting equation (28) into equation (29): where ~ < w (see Figure 9). In this case, the cables have
a cubic profile when the bridge does not carry imposed
0.9 3(2/yy -- z237 -- z 2 y ) 8 load. Since H1 = H2 = H/2, the change in hanger load
(30)
Z2 z-2 Z 337(y + 3y) + z3y(35 + 3"9) per unit span obtained from equation (9) is:
P/3/3 L3
f,i(L-x)vdx=0 (36) V1 -t--V2 - 2--H 24H (S1 -4-$2) (45)
(The area and moment of area of the deflection diagram and also:
must be zero.)
If a concentrated load is applied at x =/3, the cubic P/3(fl-/3) L3
displacement ~ satisfying the end conditions and equa- V1 - V2 - 4H 96H (Sl - s2) (46)
tions (36) with the correct function is:
Hence, by using equations (43) and (44):
V= 3(L - x)xd + 1 0 ( f i - / 3 ) ( L / 2 - x)(L - x)xd
L2 L4 12H3fL I
L -I E A w 2 j (s 1 -t'- $2) = 12Pflfl (47)
(37)
where d is the peak value of 0 (Figure 9). The deflection
6 under the concentrated load is obtained by substituting ( 48HBfLI
x =/3 in equation (37) and subtracting this from d to L3q ~ / (Sl - s2) = 24P/3(fi-/3) (48)
give:
The deflection under the concentrated load, however,
- 1- 3q ., (38) is given by:
This equation has been used to obtain the comparison 6 - P/3fiH~ ~--~
(st/3/3 + s2) /3(/38__..._H~-/3)
(st - s2) (49)
shown in Figure 10.
Substituting for sl + s 2 and s~ - s2 from equations (47)
Single span with cable extensibility included and (48) gives:
Whereas the ratio ~ / w does not affect the gravity stiff-
ness deflections provided that ~ > 0, it has an important Ht~ tiff 3afl2/32 31.'f12(l~--/3) 2
influence on the effects of cable extensibility. This is PL L2 L4 L4
(50)
particularly true when ~ tends to zero, for then the
changes in cable tension arising from an imposed load where:
tend to infinity. Because the equations including extensi-
bility are more complex, only the case of a symmetric 1 12H3f 1 48H3f
two cable bridge with piecewise parabolic cables will -= 1~ -= 1 -~ (51)
a EALZw 2 u EAL2i~ 2
be considered. Again, it will be assumed that there is
no movement of the towers. With no temperature
In this equation, the first two terms are correct for a
effects, equation (15) gives, for each cable:
classical configuration, whilst the third term gives the
1 fL
o
wivdx = fihi
--
EAi
L (39)
improvement obtained through using the two-cable con-
figuration. Where H = 182700 kN, E = 207 GN m -2, A
= 0.3224 m 2, L -- 990m, and w -- 123 kN m -1 (as per the
main span of the Severn bridge, 13 then f = 1.0555, a
From symmetry, however, H I = H 2 = H / 2 , AI=A 2 = 0.9276 and for ~ = w, v = 0.7621. Figure 11 shows
= A/2, and fl =f2 =f- Hence: the effect of using different ffffw ratios. Although ~ / w
ratios close to one are likely to be impractical, for 0.4w
=filL < ~ < 0.6w significant reductions in deflection can be
fL wivdx hi (40)
o EA obtained as compared with the classical configuration
represented by ~ / w = O.
The equivalent of equations (18) and (20) are: If d is the dip at centre span, 8Hd = wL 2. Hence:
w fL ( f L L3w2'\__
where o-H = H/A is the dead load stress in, the cables = - fdx+ +~)h+cOtL
u- u- Hr# , EA
where they are horizontal. Therefore, cable extensibility
effects become less important if Lid or ~n/E are (58)
reduced.
where:
0.2 -0
/ ''1 "~q/0,2 //1 "',\ /// "~. Classical
Z
If---'--"-Q4,/o
)~ ~1'o / / "
4 Z_~i
II
\
\ /
/
Festooned
\\/
\
0.1 I ,, , / ,,,
-- = 0 '1
0.08
f f/"~
\ / 02 / / kf
A/ " " \ / x
0.04
I
/,/ "~/~,/"'/ . / 10 1 "
!
\
' ,
/
Festooned
\
\
i~..---It \ I / \
,, ~1 X
0
0 0.5 1.0 1.5 2.0 2.5 3.0
x/L
C
Figure 12 Use of festooning to increase stiffness of bridge with three equal spans: (a) cable profiles for w / ~ = 0.4; (b) deflection
under concentrated load; (c) maximum deflections at different spanwise positions due to distributed loads of magnitude ~,
a rL\1
0.2 - Including cable ~. Gravity stiffness only
extensibility |~"~ \ //~'~"'~ \ f Classical
.,,4 ii \. / \
- [ I k / x-profile \
0.1
b 0 / , V il V , \
0 0.5 1.0 1.5 2.0 2.5 3.0
x/L
Gravity stiffness only
0.08
c,l
i'~'\ //~'~\\X~- Classical
Including cable \\ // \
extensibility \ / \
\ l \
0.04
i \ / x-profile
C 0
0 0.5 1.0 1.5 2.0 2.5 3.0
x/L
Figure 13 Use of x-profiles to increase stiffness of bridge with three equal spans: (a) cable profiles; (b) deflection under concentrated
load; (c) maximum deflections due to distributed loads of magnitude ~'
shown in Figure 1 (e). In this case, the lower cable sup- so a value h i may be assigned to both cables meeting
port points and the tops of the outer towers need to at a particular tower top.
be restrained as shown in Figure 13(a). If gravity stiffness For a three-cable bridge with identical span configur-
only is taken into account and the cables have a para- ations, the variables will be as shown in Figure 13(a).
bolic profile, it can be shown that the tower tops are Since, for span i:
all fully restrained thus leading to the same improve-
ment in stiffness compared with the classical configu-
ration as was obtained by festooning. For the analysis
f wL 3
i ~dx = I-~'-H-~2(hi + hi+l) (63)
including cable extensibility, it is necessary to allocate
horizontal displacement variables u i at each tower top.
Also, it will be assumed that the towers themselves do equation (15), when applied to each of the six cables
not restrain horizontal movements of the cables and in turn, yields:
u2 u3
0 (profiled)
span 2 f r
Odx = - (1 + r)(1 +4r)(2 + 4r) f 3 Odx (69)
12H2 ( 12H 3
\ h 2 + (1 + 2r)h3 = ~ J 2 17dx q- w----~L3u3 From which the deflection 8' of a point x =/3 on the
left-hand span is given by:
H~' r /3fi
12H 3 (70)
[ (1 + 2r)h 3 + h4 = w2L3 u3 @L2 (1 + r)(1 + 4r)(3 + 47") 2L 2
span 3 This method has been used to obtain the left-hand span
/ (64)
curves in Figure 13. From Figures 12 and 13 it is apparent
\ h 3 + (1 + 2r)h 4 = 0
where: that either festooned or x-profiled cables have consider-
able stiffness benefits. Furthermore, it would be possible
12H3f to use other configurations which have similar effects,
r = w2L2E~ (65) such as in Figures 1 (g) and (h).
It has been assumed here that the third span is not Configuration with hung sidespans
loaded, that there is no change of temperature, Ai In the case where a bridge has one mainspan and two
= A / 2 , Hi = 1-1/2, and wi = w/2. Solving these equations much shorter sidespans, greater stiffness against asym-
for a load acting on the centre span only gives: metric loads acting on the main span can be obtained
by using piecewise parabolic or cubic profiled cables
f 2 vdx = ~wL3 (h2 + h3) for the main span. There is no improvement in stiffness,
however, for loads applied to the sidespans or for sym-
metric loads applied to the mainspan. Whereas it is of
negligible benefit to use one of these cable configur-
_ 1 + 2r ( 0dx (66)
ations on the sidespans, it is of benefit to festoon the
(l+z)(l+4r) J2 sidespan cables. This has the effect of restraining move-
ment of the tower tops which increases the stiffness of
and, hence: the bridge for sidespan and symmetric mainspan loading
cases.
l+2r Consider a bridge which has piecewise parabolic
a = (1 + r)(1 + 4r) (67) cables for the main span and festooned cables having
the same ff/w ratio for the sidespans. Let the spans
is the reduction factor in ff caused by cable extensibility be L~, L 2 and L 3 and let the tower top displacements
effects. be u2 and u3 (see Figure 14(a)). The compatibility equa-
With span parameters similar to those quoted in a tions for the main span cables are:
previous section, except for f = 1.0972, then 7 = 0.08114
and a = 0.8116. (The value o f f has been calculated for f 2f2hlL
_ 2 + c20tL2
U3 -- U2 = -- fi2 2 WlVdX+ _ EA (71)
the case where the lowest points of the cables occur
at the 1/4 and 3/4 span positions.) This value of a has
been used to compute the centre span curves including
cable extensibility in Figure 13. If the left-hand span 2 f w2vdx+ 2f2h2L 2 + c20tL2
only is loaded, solution of equations (64) gives: u 3 - u2= - - H 2 EA
w
_ 3 + 14r + 12r 2 ( 0dx fl vdx +f2hL: + cxOtL2 (72)
(68) u3 - u2 = - ~1 j: EA
(1 + r)(1 + 47)(3 + 4r) )1
This is similar to equation (58) for a festooned span
and, hence, for the example, a = 0.8853. but with "O= 1. Hence, equation (61) can be used to
Although this value of a may be used to compute determine h.
the deflection under a concentrated load, in the compu- For the case where a concentrated load acts on a side-
tation of maximum deflections due to distributed load, span, the deflection under the load may be obtained