Multiple Periodic Solutions To A Suspension Bridge Ordinary Differential Equation
Multiple Periodic Solutions To A Suspension Bridge Ordinary Differential Equation
=
3K
ml
cos [(y l sin )
+
(y + l sin )
+
]
1
+ f(t) (1)
y
=
K
m
[(y l sin )
+
+ (y + l sin )
+
]
2
y
+ g ,
where
1
,
2
are damping constants, g is the force due to gravity, and f(t) is the
external force at time t.
P. J. McKenna & K. S. Moore 185
The vertical
deflection of
the center of
gravity y(x,t)
the deflection from horizontal theta(x,t)
nonlinear cable-like springs
Figure 2: A horizontal cross section of the center span
186 Multiple periodic solutions to a suspension bridge O.D.E.
Assuming that the cables never lose tension, we have y l sin 0 and
hence (y l sin )
+
= (y l sin ). Thus, the equations (1) become uncoupled
and the torsional and vertical motion satisfy
=
6K
m
cos sin
1
+ f(t) (2)
y
=
2K
m
y
2
y
+ g (3)
respectively.
The equation for the vertical motion, (3), is simply the equation for a
damped, forced, linear harmonic oscillator and the behavior of its solutions is
well known, [2]. We will study the equation for the torsional motion (2), which
is the damped, forced, pendulum equation and is known to possess chaotic so-
lutions, [2].
To choose the physical constants K, m, :=
2
and the external forcing term
f(t), we rely on [1], [8], and [12]. From [1], we choose m = 2500 and = .01. To
determine K, from [1] we know that the main span would deect about half a
meter when loaded with 100 kgs per unit length, so we have 100(9.8)2K(.5) =
0 and we take K = 1000.
For a cross section similar to the Tacoma Narrows bridge, wind tunnel ex-
periments indicate that aerodynamic forces induce approximately sinusoidal os-
cillations of amplitude three degrees [12], so in (2) we choose f(t) = sin(t)
where [0, 0.06] is chosen to produce the appropriate behavior near equilib-
rium and the frequency is chosen to match the frequency of the oscillations
observed at Tacoma Narrows on the day of the collapse. The frequency of the
torsional motion was approximately one cycle every 4 or 5 seconds, so we take
[1.2, 1.6].
3 A Multiplicity Theorem
In this section, we prove that undamped equations of the form (2) have multiple
periodic weak solutions. To some extent, this result is inspired by the earlier
results of Castro and Lazer, [3], where they obtained multiple solutions of elliptic
boundary value problems once a critical value, measured by f(0), passes across
key eigenvalues.
Let = (, ), H = {u L
2
()|u(t) = u(t)} and dene Lu = u
.
For u H, let u = u
L
2
()
= (
_
|u|
2
dt)
1
2
. Using cos u sinu =
1
2
sin 2u,
removing the damping term, and imposing periodicity conditions, we rewrite
(2) as
Lu = b sin2u h(t) (4)
u() = u(), u
() = u
()
We observe that the eigenvalues and eigenfunctions of L are
n
= n
2
,
n
(t) = sin nt (5)
hence L
1
exists, is compact, and L
1
= 1.
P. J. McKenna & K. S. Moore 187
Denition We say that u H is a solution to (4) if
u L
1
(b sin2u h) = 0. (6)
We will prove the existence of multiple solutions u H to the problem (4).
Theorem 3.1 Let h H with h 1 and let b (
1
2
, 2). Then there exists
0
> 0 such that if || <
0
, (4) has at least two solutions.
Proof Let <
0
; we will determine the value of
0
later. Note that by (5)
and our choice of b, 2b is not an eigenvalue of L.
Dene T
1
: H H by
T
1
(u) = u L
1
(b sin2u h)
and note that, by (6), zeros of T
1
correspond to solutions of (4). Denote
the Leray-Schauder degree of the map T
1
in the domain U at the point p by
deg(T
1
, U, p). To prove the theorem, we will show
(D1): there exists R
0
> 0 such that for R > R
0
, deg(T
1
, B
R
(0), 0) = 1
(D2): there exists (0, R
0
) such that deg(T
1
, B
(0), 0) = 1
Then, since deg(T
1
, B
(0).
To establish (D1), dene
T
u = u L
1
(b sin 2u h)
for [0, 1] and note that this denition of T
1
is consistent with our previous
denition. Note also that T
0
is simply the identity map, hence for any R > 0
we have deg(T
0
, B
R
(0), 0) = 1. The homotopy property of degree ensures that
deg(T
, B
R
(0), 0) is constant provided that 0 / T
(B
R
(0)) for all [0, 1].
Fix [0, 1] and suppose u H solves T
u = 0, we have
u = L
1
(b sin 2u h)
[
0
+ b sin2u]
[
0
+ bm()
1/2
]
[
0
+ bm()
1/2
]
= [
0
+ b
2] < R
0
if we choose R
0
>
0
+ b
2.
188 Multiple periodic solutions to a suspension bridge O.D.E.
Thus, for R > R
0
we have
deg(T
1
, B
R
(0), 0) = deg(T
0
, B
R
(0), 0) = 1 (7)
and (D1) above holds.
To establish (D2), dene
T
u = u (1 )L
1
(2bu) L
1
(b sin 2u h)
for [0, 1] and note again that this denition of T
1
is consistent with our
previous denitions. We will again apply the homotopy property of degree and
a standard degree calculation to show that for some > 0
deg(T
1
, B
(0), 0) = deg(T
0
, B
(0), 0) = 1. (8)
We will show rst that for any > 0,
deg(T
0
, B
(0), 0) = 1. (9)
Consider the nite dimensional subspace M
N
:=span{
n
}
N
n=1
of H and recall
that, by compactness, 2bL
1
can be approximated in operator norm by the
operators B
N
: M
N
M
N
given by
B
N
(u) = 2b
N
n=1
c
n
n
.
By denition, for N suciently large,
deg(T
0
, B
(0), 0) = deg(I B
N
, B
(0) M
N
, 0)
=
u(IBN)
1
(0)
signJ
IBN
(u)
where J
(0) M
N
, 0) = sign
N
n=1
(1
2b
n
).
Since b (
1
2
, 2) and
n
= n
2
, the only negative value of 1
2b
n
occurs at
1
= 1.
Since the eigenvalues of L are simple, we have
deg(I B
N
, B
(0) M
N
, 0) = 1
and (9) holds.
To establish (8), and hence (D2), we must show that for some > 0,
deg(T
1
, B
(0), 0) = deg(T
0
, B
, B
(0), 0) is constant
provided 0 / T
(B
(0)) for all [0, 1]. Observe that u = 0 is the only zero
of T
0
since 2b is not an eigenvalue of L. Fix (0, 1] and suppose that u = 0
solves T
u = 0. Set u =
> 0.
Set =
u
u
=
u
. We claim rst that there exists a compact set K with
K. Since u is a zero of T
, u satises
Lu 2bu = (b sin2u 2bu h) (11)
and therefore
Lu = (b sin 2u h) + (1 )2bu.
Thus,
Lu b sin2u h + 2bu
b sin2u + 2b
+
0
bm()
1
2
+ 2b
+
0
2b + 2b
+
0
.
It follows that u L
1
(B
2b+2b +0
(0)), which is compact since L
1
is com-
pact, and thus there exists a compact set K with K.
Since u satises (11), we have
u L
1
2bu = L
1
(b sin 2u 2bu h). (12)
Denote the left and right hand sides of (12) by LHS and RHS respectively.
Since 2b is not an eigenvalue of L, we have L 2b = 0 and hence
inf
K
L
1
2b = > 0
and for our u we have
LHS = u L
1
2bu
. (13)
Now considering the right hand side of (12), we have
RHS b sin2u 2bu h
[
0
+b sin2u 2bu]
= [
0
+b sin(2
) 2b
].
We claim that if
sin 2 2b
and note that (L1) above is satised.
To show that (L2) holds, we must show that f
() :=
b
sin 22b 0
uniformly on K as 0. To show that f
} is equicontinuous
on K.
Choose K, = 0 and note that
b
sin 2 2b = b
sin2
2
2 2b 0
as 0, hence
|
b
sin 2 2b|
2
0.
Moreover, since | sin w| |w| and since w
2
is convex, we have
|
b
sin 2 2b|
2
= 4|
1
2
b
sin2 +
1
2
(2b)|
2
4[
1
2
|
b
sin2|
2
+
1
2
|2b|
2
]
2[
b
2
2
|2|
2
+|2b|
2
] = 16b
2
2
L
1
since L
2
. By the Dominated Convergence Theorem, we conclude that
f
() f
)| = |
b
sin 2 2b
b
sin 2
2b
|
b
(sin 2 sin 2
) 2b(
)
sin2 2
+ 2b
2 2
+ 2b
= 4b
<
provided
<
4b
.
Since f
)].
Assume now that
0
<
1
2
<
, then (
) <
1
2
and hence
RHS <
.
But this contradicts (13). Therefore
is bounded below by
+ sin t. (14)
In [8], the author prescribed the initial position (0) and velocity
(0) of the
cross section and employed the Runge-Kutta method to solve the initial value
problem over large time. He demonstrated numerically that small or large am-
plitude periodic motion may result, depending only on the initial conditions.
Moreover, the amplitude and frequency of the large amplitude solutions matched
the behavior observed at Tacoma Narrows on the day of its collapse, [1].
As in [8], we compute periodic solutions to (14). Moreover, we describe a
numerical continuation algorithm and we use it to plot the amplitude of periodic
solutions versus the amplitude of the external forcing term. We demonstrate
that for small , three periodic solutions to (14) exist and that whether large
or small amplitude motion results depends on the initial position and velocity
of the cross section. Moreover, we demonstrate that bifurcation from single to
multiple periodic solutions occurs for small .
4.1 The Algorithm
We wish to generate the bifurcation curve for periodic solutions to (14) as the
amplitude of the external forcing term varies; i.e., we wish to plot the am-
plitude A
_
. (15)
Then solutions of (14) correspond to solutions U of the rst order system
u
= v (16)
v
= v
v
0
which yield a T-periodic solution to (17). We then solve the initial value
problem (17) for t [0, T] and test whether F(
0
) = 0. If not, we update
0
via Newtons method:
1
=
0
DF(
0
)
1
F(
0
)
where
DF =
_
F1
1
F1
2
F2
1
F2
2
_
(19)
is the Jacobian matrix of F.
To compute this matrix, observe that by our denition of F in (18), we have
DF =
_
F1
1
F1
2
F2
1
F2
2
_
=
_
1
u
1
(T)
u
2
(T)
v
1
(T) 1
v
2
(T)
_
. (20)
To compute
u
i
,
v
i
, i = 1, 2, we recall that the solution to the initial value
problem is continuously dierentiable with respect to initial conditions and pa-
rameters, [5], and hence we can dierentiate (17) with respect to
1
. This yields
P. J. McKenna & K. S. Moore 193
the rst order system in
u
1
,
v
1
_
u
1
v
1
_
=
_
v
1
2.4 cos(2u)
u
1
.01
v
1
_
;
_
u
1
v
1
_
(0) =
_
1
0
_
(21)
Hence [
u
1
,
v
1
]
T
solves the initial value problem
_
w
z
_
=
_
z
2.4 cos(2u)w .01z
_
;
_
w
z
_
(0) =
_
1
0
_
(22)
where u is the solution to the initial value problem (17).
Similarly, dierentiating (17) with respect to
2
, we see that [
u
2
,
v
2
]
T
solves the initial value problem
_
w
z
_
=
_
z
2.4 cos(2u)w .01z
_
;
_
w
z
_
(0) =
_
0
1
_
. (23)
Therefore, to compute the entries of the Jacobian matrix (20), we must solve
the initial value problems (17), (22), and (23).
The above algorithm (Shooting Algorithm with Newton Update) suggests a
method by which we can plot the amplitude A
_
F1
1
F1
2
F1
F2
1
F2
2
F2
N
1
N
2
N
_ (26)
194 Multiple periodic solutions to a suspension bridge O.D.E.
is nonsingular even if the matrix DF is singular, [7]. Finally, we observe that
the entries in the third row of DF can be computed directly from the denition
(24) of N and that the computation of
Fi
+ .01
+ 2.4 = sin t
of (14) and note that for the undamped equation, resonant solutions occur at
,
we cannot compute large amplitude periodic solutions; the algorithm does not
converge to a periodic solution (see Experiment 4). If >
, the amplitude
of periodic solutions increases with , but bifurcation from single to multiple
periodic solutions does not occur.
Experiment 1 = 1.0; See Figure 3
In this experiment we see that the path of periodic solutions is S-shaped and
that bifurcation from single to multiple periodic solutions occurs at 0.0126.
If < , (17) has a unique periodic solution of small amplitude, but if
(, .623), multiple periodic solutions exist. Whether the cross-section oscillates
with small or large amplitude depends only on the initial conditions. This is
consistent with the results in [8]. Moreover, we observe that the amplitude of
the large oscillations is close to one radian, as was observed at Tacoma Narrows
on the day of its collapse, [1].
Experiment 2 = 1.2; See Figure 4
Again we see that the path of periodic solutions is S-shaped and that bifurcation
from single to multiple periodic solutions occurs at 0.0117.
Experiment 3 = 1.4; See Figure 5
Again we see that the path of periodic solutions is S-shaped and that bifurcation
from single to multiple periodic solutions occurs at 0.0088.
Experiment 4 = 1.5; See Figure 6
From the curve in Figure 6, we see that multiple solutions exist for [.0197, .0213].
P. J. McKenna & K. S. Moore 195
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.2
0.4
0.6
0.8
1
1.2
1.4
lambda
a
m
p
l
i
t
u
d
e
o
f
p
e
r
i
o
d
i
c
s
o
l
u
t
i
o
n
mu= 1.0
Figure 3: Experiment 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.2
0.4
0.6
0.8
1
1.2
1.4
lambda
a
m
p
l
i
t
u
d
e
o
f
p
e
r
i
o
d
i
c
s
o
l
u
t
i
o
n
mu= 1.2
Figure 4: Experiment 2
196 Multiple periodic solutions to a suspension bridge O.D.E.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.2
0.4
0.6
0.8
1
1.2
1.4
lambda
a
m
p
l
i
t
u
d
e
o
f
p
e
r
i
o
d
i
c
s
o
l
u
t
i
o
n
mu= 1.4
Figure 5: Experiment 3
However, because = 1.5 is close to the resonant frequency
, the amplitude
of periodic solutions grows with , but that the growth is slower for higher
frequencies. Moreover, for larger , the amplitude of the periodic solution grows
nearly linearly with .
5 Conclusion
In this paper, we have presented substantial progress on the understanding
of the structure of the periodic solutions for the forced torsional equation in
the parameter ranges where the large-amplitude oscillation was observed at
Tacoma Narrows. The multiple solutions exist over roughly the right interval
of amplitude and period, consistent with the historical evidence.
Nonetheless, some intriguing questions remain.
First, we have not presented any proof that the large amplitude solutions
persist if there is small damping. This is certainly supported by the numerical
evidence but remains open as a mathematical question. Obviously, there must
P. J. McKenna & K. S. Moore 197
0 0.005 0.01 0.015 0.02 0.025
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
lambda
a
m
p
l
i
t
u
d
e
o
f
p
e
r
i
o
d
i
c
s
o
l
u
t
i
o
n
mu= 1.5
Figure 6: Experiment 4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
lambda
a
m
p
l
i
t
u
d
e
o
f
p
e
r
i
o
d
i
c
s
o
l
u
t
i
o
n
mu=1.6, mu=1.7, mu=2.0, mu=2.6
Figure 7: Experiment 2
198 Multiple periodic solutions to a suspension bridge O.D.E.
be some relationship between the small forcing term and the small damping,
since in the absence of any forcing, all motion must eventually decay.
Second, the original paper [8], contains two other areas which have not been
studied here. One is the observation of subharmonic solutions at integer multi-
ples of the frequency of forcing studied here. More sophisticated continuation
methods may be needed to see exactly where these families of solutions lie on
the big bifurcation picture.
Third, the relationship between the vertical and torsional motions has not
been addressed. Presumably, there are multiplicity theorems for the coupled
ordinary dierential equation system. But more importantly, the role of the
loosening of the cables in large vertical motion and the instantaneous transition
to torsional motion, so dramatically portrayed in [8], has not been addressed
here. More light will be shed on this question when the global bifurcation picture
of the coupled system is available. The relationship between the instability of
the vertical large-amplitude solutions and the torsional solutions should prove
to be particularly interesting. Progress in this direction has been made in [6],
[9], and [10].
Fourth, the question of the spatial dependence and nodal structure of the
solutions of a nonlinear beam equation will give more insight into the observed
large-amplitude motions captured on the famous lm of the Tacoma Narrows.
This question is studied in [11].
In short, we expect this subject to be a fruitful area of study (and insight
into history) for some time to come.
References
[1] O.H. Amann, T. von K arman, and G.B. Woodru. The Failure of the
Tacoma Narrows Bridge. Federal Works Agency, 1941.
[2] Blanchard, Devaney, and Hall. Dierential Equations. Brooks/Cole Pub-
lishing Company, Pacic Grove, 1998.
[3] A. Castro and A. C. Lazer. Critical point theory and the number of solutions
of a nonlinear Dirichlet problem. Ann. Mat. Pura Appl. (4), 120:113137,
1979.
[4] Y.S. Choi, K.C. Jen and P.J. McKenna. The structure of the solution set
for periodic oscillations in a suspension bridge model. IMA J. Appl. Math.,
47:283-306, 1991.
[5] E.A. Coddington and N. Levinson. Theory of Ordinary Dierential Equa-
tions. McGraw-Hill, New York, 1955.
[6] S.H. Doole and S.J. Hogan. Non-linear dynamics of the extended Lazer-
McKenna bridge oscillation model. Dyn. Stab. Syst., 15 (2000), no. 1
4358.
P. J. McKenna & K. S. Moore 199
[7] H. B. Keller. Lectures on Numerical Methods in Bifurcation Problems.
Springer-Verlag, Berlin, 1987.
[8] P.J. McKenna. Large torsional oscillations in suspension bridges revisited:
xing an old approximation. The American Mathematical Monthly, 106:1-
18, 1999.
[9] P.J. McKenna and K.S. Moore. The global structure of periodic solutions
of a suspension bridge mechanical model. In preparation.
In preparation.
[10] P.J. McKenna and Cilliam OTuama. Large torsional oscillations in suspen-
sion bridges revisited yet again: vertical forcing creates torsional response.
In preparation.
[11] K.S. Moore. Large amplitude torsional oscillations in a nonlinearly sus-
pended beam: a theoretical and numerical investigation. Dissertation, Uni-
versity of Connecticut, 1999.
[12] R.H. Scanlan and J.J. Tomko. Airfoil and bridge deck utter derivatives.
Proc. Am. Soc. Civ. Eng. Eng. Mech. Division, EM6, 1717-1737, 1971.
Joe McKenna
Department of Mathematics
University College
Cork, Ireland
e-mail: mckenna@math.uconn.edu
Kristen S. Moore
Department of Mathematics
University of Michigan
Ann Arbor, MI 48109-1109, USA
e-mail: ksmoore@math.lsa.umich.edu