Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Periodically forced damped pendulum

Carmen Hoek, Thomas Dijk, Jelmer Levenga


February 10, 2019

1
Contents
1 History 3

2 Small angle approximation 3


2.1 introduction to the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Solving for small angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 evaluating the small-angle equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Forced damped pendulum 8

4 Numerical solutions 11
4.1 Euler’s method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.1 The first derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.1.2 Code for the numerical approximation . . . . . . . . . . . . . . . . . . . . . 11
4.1.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 Bibliography 15

2
1 History
The pendulum has been known for quite a long time, the first recorded use of a pendulum was
in the first century AD, in China. The pendulum was used as a seismometer. In the case of
an earthquake a pendulum would hit a lever which would cause a small ball to fall down, which
indicated the direction of the earthquake. [4]
The first time the pendulum was described was by Galileo Galilei around 1582 [5]. He discovered
that for a pendulum the period of a swing is independent of the mass of at the end as well as the
amplitude the mass was raised to, but proportional to the square root of the length of the string
only. [6]. This is the property of pendulums that allow them to be used in clocks as timekeepers,
the period of a pendulum is set, as soon as long as the length of the rod it is attached to is known.
Newton used that fact to prove that the force of gravity acting on any object is proportional to its
mass [6]. In 1747 Daniel Bernoulli updated the previously known model and calculated that for
very small angles, the period of the pendulum is not constant. [7]

2 Small angle approximation


2.1 introduction to the problem
Before tackling the full problem, looking at a simplified version will greatly help in the understand-
ing of this problem, that simplest version is the undamped and not driven pendulum: a mass M
on a massless string of length l, which is attached to a frictionless pivot. The angle the pendulum
makes with the vertical axis will be called θ. A diagram of the problem is given in Figure 1

Figure 1: A diagram of the undamped undriven pendulum

The pendulum is not driven, so no force, besides gravity and the tension in the string, acts
on it once let go. To find the motion of the mass, these two forces should be equated. The force
of gravity is given by F = −mg where g the acceleration due to gravity. The force of gravity
can be decomposed into two parts, a part tangential to the path of the ball and a component
perpendicular to it. Only the tangential part is capable of accelerating the ball. The tangential
part is given by
−mg sin θ
The acceleration of a mass on a curved plane is given by the angular acceleration times the length
of the rod.
a = lθ̈

3
Using F = ma, the force associated with that acceleration is given by

F = mlθ̈.

Then equating the two gives


−mg sin θ = mlθ̈.
g
θ̈ + sin θ = 0
l
g
and per convention l is replaced by ω 2 :

θ̈ + ω 2 sin θ = 0. (1)

A possible damping force is given by friction, the magnitude of such a force scales with the (angu-
lar) velocity and opposes the motion: Fdamping = −bθ̇, where b is a constant.

The driving force is periodic so that is given by: Fdriving = Fd cos ωd t


. Here Fd is a constant and gives the amplitude of the force and ωd is the frequency of the driving
force, which in general is different from ω.

The full equation to be solved is now given as:

M θ̈ + bθ̇ + M ω 2 sin θ = Fd cos ωd t.

dividing out the mass to clean up the equation slightly gives

θ̈ + 2β θ̇ + ω 2 sin θ = F cos ωd t.
b Fd
where 2β = M and F = M.

2.2 Solving for small angles


Solving this problem analytically is only possible for small angles, because at small angles sin θ ≈ θ
and the equation becomes
θ̈ + 2β θ̇ + ω 2 θ = F cos ωd t. (2)
This can be rewritten in a slightly more usable form using the Euler identity:
F iωd t
θ̈ + 2β θ̇ + ω 2 θ = (e + e−iωd t ). (3)
2
This is a linear differential equation, thus the general solution will be the sum of the homogeneous
and particular solution.
ẍ + 2bẋ + ax = 0 (4)
The solution can be found by making an Ansatz, that

xh (t) = Aeαt .

substituting that into equation 4 gives

Aα2 + 2bα + a = 0.

This is a quadratic equation in α with two roots: α1 and α2


p
α = b ± b2 − a

These solutions can be added using the superposition principle:

xh (t) = Aeα1 t + Beα2 t .

This simplifies to √ √
b2 −at b2 −at
xh (t) = e−bt (Ae + Be− ). (5)

4
and if we define p
Ω= b2 − at,
xh (t) = e−bt (AeΩt + Be−Ωt ). (6)
Where A and B are constants, given by the initial conditions. Substituting this back into equation
4 solves the equation so the Ansatz made was a good one. For the pendulum the homogeneous
solution is given by √ 2 2 √ 2 2
θh (t) = e−βt (Ae β −ω t + Be− β −ω t ). (7)

The particular solution still has to be found. To solve an equation of the form

ẍ + 2bẋ + ax = C0 eiω0 t . (8)

where a,b and C0 are constants, an Ansatz is made:

x(t) = Aeiω0 t (9)

Substituting that into equation 8 this becomes:

−ω02 eiω0 t + 2biω0 Aeiω0 t + aAeiω0 t = C0 eiω0 t .

− ω02 + 2biω0 A + aA = C0 (10)


Solving for A gives:
C0
A= .
−ω02 + 2biω0 + a
C0
x(t) = eiω0 t . (11)
−ω02 + 2biω0 + a
Substituting this into equation 8 solves this equation so the initial Ansatz has been a good one.
Applying this to equation 3 is fairly straightforward, but care should be taken, because there are
two complex exponentials. Luckily the superposition principle tells us that the full solution will
be one term that corresponds to eiω0 t and one that corresponds to e−iω0 t :
F F
θp (t) = eiωd t − e−iωd t . (12)
2(−ωd2 + 2iβωd + ω 2 ) 2(−ωd2 − 2iβωd + ω 2 )

While this is the analytic solution to this particular solution, it is not very useful at this point,
because it is still in complex form. It can however be rewritten to give

F (ω 2 − ωd2 ) 2F βωd
θp (t) = cos ωd t + 2 sin ωd t. (13)
(ω 2 − ωd2 )2 + 4β 2 ωd2 (ω − ωd2 )2 + 4β 2 ωd2

We define a quantity R such that


q
R= (ω 2 − ωd2 )2 + (2βωd )2 .

Substituting that into equation 13 gives

F ω 2 − ωd2 2βωd
θp (t) = ( cos ωd t + sin ωd t) (14)
R R R
From the trigonometric identity
p
A cos x + B sin x = A2 + B 2 cos x − φ

Where
B
tan φ = ,
A

5
Which can be applied to the equation as
F
θp (t) = cos (ωd t − φ). (15)
R
Here φ can be found from the amplitudes of the sine and cosine terms from equation 14 :
ω 2 − ωd2
cos φ =
R
2βωd
sin φ = (16)
R
Though the previous identity already gave an expression for φ, the sine and cosine are still rather
useful in finding φ in certain extreme situations. combining these gives
2βωd
tan φ = . (17)
ω 2 − ωd2
Now that both the homogeneous solution and the particular solution are known, the full solution
is given by:
F
θ(t) = cos (ωd t − φ) + e−βt (AeΩt + Be−Ωt ). (18)
R
[2] [3]

2.3 evaluating the small-angle equation


From this equation it is immediately obvious that over a long enough time, the homogeneous
solution dies off and all that is left is the particular solution. At that time, the function becomes
F
θ(t) = cos (ωd t − φ).
R
1
The amplitude, from hereon called C, of this equations scales with R so it scales as
1 1
C= =p (19)
R (ω − ωd )2 + (2βωd )2
It is then possible to say something about φ, and the amplitude in three extreme cases, with small
damping:
1 In the situation where βωd << ω 2 − ωd2 , or for simplicity, ωd ≈ 0. we find from equation 16
that
sin φ = 0, φ = 0
In the amplitude the ωd term drops out so the amplitude scales as:
1
C∝ .
ω2
2 In the limit that ωd → ω from equation 17 we find
π
tan φ ≈ ±∞, φ=
2
Where φ is restricted to positive values only. In the amplitude we find that

lim C = ∞
ωd →ω

This is called resonance and dramatically increases the amplitude of the system.
3 When βωd << ωd2 − ω 2 ,or for simplicity, ωd ≈ ∞ it can be seen from equation 17 that

tan φ ≈ 0, φ=π

The amplitude then becomes


1
C∝
ωd2

6
A plot of equation 18 is given in Figure 10 and the corresponding phase portrait is given in Figure4
with the same initial conditions. For the plot the following initial conditions were used:
l=1
g = 9.81
ωd = 1.5
β = 0.1
F =2
A=B=1

Figure 2: A plot of the solution θ(t) in t for the given initial conditions

Figure 3: The phase portrait of θ(t) versus θ̇(t)

From The plot it can clearly be seen that at first the angle changes quite a bit, but very quickly it

7
settles into the particular solution. The phase portrait shows the same, it starts of quite chaotically,
but quickly spirals inwards and settles in a stable circular orbit.

Figure 4: The Poincaré map of the linearized system

The Poincaré map also shows a stable orbit around the origin. The high density of points in the
center is due to the points that start out close to the origin that move outwards.

3 Forced damped pendulum


The damped, periodically forced pendulum system can be written as:

θ̈ + αθ̇ + sin θ = A sin ωt. (20)

This equation has three independent parameters: the strength of the damping α, and the amplitude
A and the frequency ω of the external field. For chaotic solutions, two conditions has to be satisfied.
The first one is that the equation must be nonlinear, which is the case in (20) because of sin(θ). The
second condition is that we must have at least three dynamical variables, i.e. a three-dimensional
state space, which is also the case, since we can rewrite (20) to

χ̇ + αχ + sin θ = f sin φ,
θ̇ = χ,
φ̇ = ω.

So we have non-linearity and the minimum dimension of three of a differential equation for chaotic
behaviour, but there are no chaotic solutions for the small driven amplitude A and large damping
coefficient α. This is because in the first case, the system will become lineal (small angle) where in
the last case we have that the differential equation system will reduce to dimension two. Chaotic
behaviour exists for the driven frequency to the resonant eigenfrequency of a pendulum. So chaotic
effects will occur near ω ≈ 1.

From the comparison of the driven amplitude A and the magnitude of the torque acting on the
pendulum, which is equal to sin θ, we can conclude the general properties of the solutions of (20).
For a weak external force, A < 1, the pendulum will stay at θ < π2 , where it will perform oscilla-
tions near the downward equilibrium point. But if A > 1, the strong driven force leads towards
full rotations.
For different initial values, we get big differences in the solutions, from which three different paths
to chaos have been found:
1. Period-doubling cascade preceded by the appearance of symmetric-breaking solutions. This
cascade holds for both locked and running solutions.
2. The loss of phase locking and random transitions between two locked states.

8
3. The intermittency form of the transition to chaos occurring for an external field of large
amplitude. The paths are then a combination of clockwise and counter-clockwise rotations
with damped oscillations in-between.
(20) has been solved by using a hybrid digital-analog computer (a computer that is uses changeable
aspects of physical phenomena to model a problem being solved) for α = 0.5 by varying ω and
A. The differences in the behaviour for different values of the driving amplitude and frequency is
given in the bifurcation diagram (Figure 5) [8].

Figure 5: Bifurcation diagram for a periodically driven damped pendulum for α = 0.5.
(Note that Ω = ω)

This graph gives us the different types of dynamic behaviour of the system (20). We get periodic
solutions if ω << 1 or if ω >> 1. In region A, for large amplitudes A, we get that the pendulum
makes rotations corresponding to the periodic motion from one potential well to another, where
is each well the pendulum performs simple oscillations around a certain point. In region B we
observe oscillations near the downward position. In this region, the set of period-doubling bifur-
cations appears what will lead to a chaotic state.

The chaotic behaviour is here characterized by strange attractors, which are attractors with fractal
structure [10]. The structure of these strange attractors can be understood in terms of two distinct
time scales of the motion: the shorter time scale corresponds to fast oscillations between some
close wells; the longer time scale corresponds with a slower diffusion throughout the lattice. This
chaotic motion is best characterized by the power spectral density in Figure 6 [9] for the parameters
A = 3.8 and ω = 0.64. Even though the sharp peak observed at the driving frequency is narrow,
it broadens at large driving amplitude.

9
Figure 6: Power spectrum of the attractors for a periodically driven damped pendulum with
A = 3.8, ω = 0.64 and α = 0.5

For a specially constructed experimental setup[11], the solutions of (20) have been obtained
for fixed values of the damping constant α = 0.24 and the driving frequency ω = 0.67. If the
driven amplitude A has small values, then the motion of the pendulum will be periodic, where the
pendulum oscillates back and forth where the spectrum of θ = θ(t) has a principal peak at the
driven frequency ω and smaller peaks at the harmonics of ω. We obtain a period-two orbit for
A = 0.69, where the power spectrum contains peaks at integer multiples of ω/2. Then another
period-doubling appears, and chaos will fully develop at A ≈ 0.70.

By changing the amplitude A for fixed α = 0.5 and ω = 0.47, a intermittent transition to chaos
was found from the numerical and analog solutions of (20). A critical value of A exists, denoting
AC , such that if A < AC , we get stable oscillations having the driven frequency. In the case
where Ais slightly larger than AC , the system becomes chaotic via the intermittent transition as
A is decreased. At this transition, θ̇ becomes unstable and periodic oscillations appear which are
randomly interrupted by bursts of θ̇. These bursts are shown in Figure 7[8].

Figure 7: Phase trajectory for a periodically driven damped pendulum with A = 0.9045, ω = 0.47
and α = 0.5.

10
The elaborated analysis of this intermittency case of the transition to chaos in (20) was done
in [12], where α = 0.5, ω = 2/3 and for different values of the amplitude A of an external field. If
A ≤ 1.5 there are two different periodic running solutions, clockwise and counter-clockwise. These
two modes remain separate and will become chaotic as A increases. When A = 1.4945, there will
occur intermittent switching between the two modes, which produces a large amount of noise if
the frequency is smaller than the driving frequency.

4 Numerical solutions
To determine how good the solutions of the small angle approximation are the problem is solved
numerically. To solve this problem the Euler approximation is used. Before the code is shown, the
numerical method is explained.

4.1 Euler’s method


Euler’s method is one of the most simple numerical approximations to make. It is based on the
definition of the derivative and is very useful for programming. The essential thing of the method
is using the definition of the derivative and instead of a limit use very small steps. To make more
sense of this, first a derivation for the first derivative is given.

4.1.1 The first derivative


The definition of the first derivative in the limit is given by:

f (x + h) − f (x)
f 0 (x) = lim (21)
h→0 h
For the numerical approximation according to Euler’s method the limit has to disapear and instead
use a very small value for h, most in the order of 0.0001. So for the numerical approximation the
following formula is used for the first derivative:
fn+1 − fn
f 0 (x) ≈ (22)
h

But there is a second order differential equation, so there must be found a numerical approximation
for the second order derivative. Again, using the definition of the derivative and put this in the
definition of the first derivative, the definition of the second derivative is given by:

f (x + 2h) − 2f (x + h) + f (x)
f 00 (x) = lim (23)
h→0 h2

For the numerical approximation, using the same technic as for the first derivative the following is
found:
fn+2 − 2fn+1 + fn
f 00 (x) ≈ (24)
h2

Numerical solutions are never perfect, there is always introduced a small error since it is an approx-
imation. This error can be estimated using the way of derivation of the method. The method is
based on a Taylor expansion, which is said to be n-th order accurate. This means that the highest
order of h is included in hn . For the case studied in this essay it means that the largest error made
in each step is given by hn+1 where n is equal to two since there are up to second order differential
equations.

4.1.2 Code for the numerical approximation


The Euler method is used in the following code.

11
%m a t p l o t l i b i n l i n e
import numpy a s np
import math
from m a t p l o t l i b . p y p l o t import f i g u r e , show
import numpy a s np

f i g = f i g u r e ( f i g s i z e =(16 ,16))
frame = f i g . add_subplot ( 2 , 1 , 1 )

the_0 = 0 . 3
the_1 = 0 . 3
tot = 0
w = (9.81)∗∗0.5
h = 0 . 0 0 1 ∗ 2 ∗ math . p i /w
wd = 1 . 5
b = 0.02
Fd = 2

s o l = [ the_0 , the_1 ]
tim = [ 0 , 0 . 0 0 1 ∗ 2 ∗ math . p i /w ]
an = [ ]
t d o t = [ 0 , ( ( the_1 − the_0 ) / h ) ]
andot = [ ]
ntdot = [ ]

w h i l e t o t <100:
the_2 = 2∗ the_1 − the_0 − h∗b ∗ ( the_1 − the_0 ) − h∗h∗w∗w∗np . s i n ( the_0 )
+ h∗h∗Fd∗np . c o s (wd∗ t o t )
the_dot = ( the_2 − the_1 ) / h
the_0 = the_1
the_1 = the_2
tot = tot + h
tim . append ( t o t )
s o l . append ( the_2 )
t d o t . append ( the_dot )

s o l u t i o n = np . a r r a y ( s o l )
time = np . a r r a y ( tim )
t h e t a d o t = np . a r r a y ( t d o t )
newthetadot = np . a r r a y ( n t d o t )

frame . g r i d ( )

frame . p l o t ( time , s o l u t i o n , l a b e l = ’ n u m e r i c a l s o l u t i o n ’ )

frame . s e t _ x l a b e l ( ’ time ’ )
frame . s e t _ y l a b e l ( ’ t h e t a ’ )

l = 1
g = 9.81
omega = ( g / l ) ∗ ∗ 0 . 5
omegad = 1 . 5
A = 1
B = 1

12
beta = 0.1
F = 2
R = ( ( omega ∗∗2 − omegad ∗ ∗ 2 ) ∗ ∗ 2 + ( 2 ∗ b e t a ∗omegad ) ∗ ∗ 2 ) ∗ ∗ 0 . 5
bigomega = ( b e t a ∗∗2 − omega ∗ ∗ 2 ) ∗ ∗ 0 . 5
p h i = np . a r c t a n ( 2 ∗ b e t a ∗omegad / ( omega ∗∗2 − omegad ∗ ∗ 2 ) )
theta = [ ]
time = np . l i n s p a c e ( 0 , 1 0 0 , 8 0 0 )
f o r t i n time :
t h e = np . r e a l (F/R∗np . c o s ( omegad∗ t − p h i ) )
t h e t a . append ( t h e )

frame . p l o t ( time , t h e t a , l a b e l = ’ s m a l l a n g l e approximation ’ )


frame . l e g e n d ( )
show ( )

4.1.3 Results
To check how good the numerical solutions are there is done an approximation of one of the most
simple linear second order differential equations, which can also be solved by hand. Below the
derivation of the equation:

ÿ + 2ẏ + y = 0 (25)
−x
The most simple solution for this equation is e . The plot with this together with the numerical
solution is shown in figure 8.

Figure 8: Numerical solution and analytic solution of the simple linear second order differential
equation.

From this code, the following figure comparing the numerical solution with the small angle ap-
proximation is found. This plot is showing the results on the long term, the start of the pendulum
is not shown.

13
Figure 9: A numerical solution of the full equation and the small angle approximation of the full
equation

As shown in figure 9 the small angle approximation solution approximates the numerical solu-
tion very well. This is very logical because the angular distance from the origin is around 0.3 rad
which is equal to approximately 17 degrees which is just outside the maximum angle of the small
angle approximation. If the same code is used, but a bigger angle is used, the following is obtained:

Figure 10: Another numerical solution and the small angle approximation

As seen in the figure the small angle approximation is not a good approximation of the numerical
solution anymore. This is the result of an angle of 0.8 rad which is equal to approximately 45
degrees.

14
5 Bibliography
References
[1] Figure 1: Chetvorno, 2008, Simple gravity pendulum, wikimedia, accessed 24-12-2018
{https://commons.wikimedia.org/wiki/File:Simple_gravity_pendulum.svg}
[2] Thornton and Marion, 2008 Classical dynamics of particles and systems pp. 117-123

[3] David Morin, 2007 Classical mechanics with problems and solutions pp. 109-111
[4] Needham, Science and Civilisation in China: Volume 3, Mathematics and the Sciences of the
Heavens and the Earth, pp627-629
[5] Murdin 2008 Full Meridian of Glory: Perilous Adventures in the Competition to Measure the
Earth, pp 41.
[6] Wikipedia, pendulum accessed 24-12-2018
{https://en.wikipedia.org/wiki/Pendulum#History}
[7] Thomson and Poynting, 1907, A textbook of physics, pp 10

[8] M. Gitterman, 2010, The chaotic pendulum, pp27-30.


[9] B. A. Huberman, J. P. Crutchfield, and N. H. Packard, 1980, Noise phenomena in Josephson
junctions pp 750-752.
[10] Wikipedia, attractors accessed 5-2-2019
{https://en.wikipedia.org/wiki/Attractor#Strange_attractor}

[11] J. A. Blackburn, Z. J. Yang, S. Vik, H. J. T. Smith, and M. A. H. Nerenberg,Physica D26,


385 (1987)
[12] E. G. Gwin and R. M. Westervelt, 1984 Intermittent Chaos and Low-Frequency Noise in the
Driven Damped Pendulum Phys. Rev. Lett.54, 1613 (1985); Phys.Rev. A33,4143 (1986).

15

You might also like