Agronomy Research Biosystem Engineering Special Issue 1, 5-12, 2011

On chaotic motion of a double pendulum

A. Aan, E. Aarend and M. Heinloo

Institute of Technology, Estonian University of Life Sciences, Kreutzwaldi 56,

EE 51014 Tartu, Estonia, e-mail: aare.aan@emu.ee

Abstract. This paper presents the results of studying the chaotic motion of a virtual double
pendulum. Large oscillations of this pendulum were modelled by the system of nonlinear
differential equations in the Hamilton form. This system was solved on the worksheet of
Computer Package Mathcad numerically, using the Runge-Kutta algorithm of fourth order. The
results are illustrated by some frames of video clips visualizing the chaotic motion of a double
physical pendulum. The results of this paper can be used in the teaching process of analytical
mechanics for students of engineering.

Key words: Analytical mechanics, equations of Hamilton, double pendulum, chaotic motion,
Computer Package Mathcad.

INTRODUCTION

Classical analytical mechanics contains a lot of differential equations in the

Lagrange and the Hamilton forms. These equations, mostly nonlinear, are practically
dead before their numerical solution and imagination of the results. It is convenient to
use different computer programs for this purpose, for example, the Computer Package
Mathcad. From the point of view of teaching it is useful to simulate, if possible, the
motion of a virtual object, modelled by the nonlinear equations of Lagrange or
Hamilton. Such an object is, for instance, a double pendulum. It appears that the
motion of virtual and physical models of a double pendulum is chaotic. It is possible to
find a lot of results on the study of chaotic virtual models of double pendulums on
different web sites, for example [1]–[12]. The physical models of a double pendulum
are presented, for example, on the web sites [13]–[18].
The purpose of this paper is to use the features of the Computer Package Mathcad
and solve numerically, using the Runge-Kutta algorithm of fourth order, the nonlinear
system of differential equations in the Hamilton form, modelling the large oscillation
of a virtual physical double pendulum, and to visualize its chaotic motion for different
initial conditions.

5
 MATERIALS AND METHODS

A mathematical model to study the motion of a double pendulum

Let us consider a physical double pendulum in Fig. 1

Figure 1. A physical double pendulum.

The double pendulum in Fig. 1 consists of two links of equal length and gravity
mg, where g is the acceleration of gravity, joined together by a pivot. The upper pivot
of the double pendulum is grounded. In the current position the upper link of the
double pendulum has inclination angle Tand the lower link has inclination angle
TThe co-ordinates of the centre of mass of the upper link are (x1, y1) and of the lower
link – (x2, y2).
The mathematical model in the Hamilton form, describing large oscillations of the
double pendulum is [18, 19]

6 2 pT1  3 pT 2 cos T1  T 2 6 8 pT 2  3 pT1 cos T1  T 2 (1)

T '1 , T '2 ,
ml 2 16  9 cos T1  T 2 2
ml 2 16  9 cos T1  T 2 2

1 § g ·
p'T1  ml2 ¨T '1 T '2 sin T1  T 2  3 sin T1 ¸ ,
2 © l ¹

1 § g ·
p'T 2  ml 2 ¨  T '1 T '2 sin T1  T 2  sin T 2 ¸ ,
2 © l ¹

where TandTare generalized co-ordinates [19], or inclination angles, of the double

1 2 1 2
pendulum (Fig. 1) and pT1 ml 8T '1 3T '2 cos T1  T 2 , pT 2 ml 2T '2 3T '1 cos T1  T 2 ,
6 6
are the generalized impulses [20] of the double pendulum. Note, that for small
oscillations of a double pendulum, considered in [21], the mathematical model is
following form equations (1), when cos T1  T 2 | 1 , sin T1  T 2 | 0 , sin T1 | 0 , sin T 2 | 0 .

6
 Let us assume that the special solution of the equations (1) satisfies the following
initial conditions:

T1 0 T10 , T2 0 T20 , pT1 0 pT10 , pT 2 0 pT 20 , (2)

where TT pT10  pT 20 are constants. Let us consider the first initial position of the
double pendulum, shown in Fig. 2 and assume that the motion of the double pendulum
begins without initial angular velocities.

10

 10 0 10

 10

Figure 2. First initial position of the double pendulum.

This assumption means that in the initial conditions (2) the constants must have
the following values:
S (3)
T10 , T20 0 , pT10 0 , pT 20 0 ,
2

Let us consider now the initial position of the double pendulum, shown in Fig 3
and assume also that the motion of the double pendulum begins without initial angular
velocities
10

 10 0 10

 10

Figure 3. Second initial position of the double pendulum.

This assumption means that in the initial conditions there must be the following
values of the constants:

T10 S , T 20 S , pT10 0 , pT 20 0, (4)

7
 Numerical solution of the system of differential equations (1) on the worksheet of
the Computer Package Mathcad

To use the algorithms programmed in Computer Package Mathcad for solution

system (1) of differential equations on the worksheet of Computer Package Mathcad
we have to use the following new variables with array subscripts y0 = Ty1 = T,
y2 = pT1  y3 = pT 2 and to present the system (1) in normal form

y'0 f1 m, l , y) , y'1 f 2 m, l , y) , y'2 f3 m, l, g , y) , y'3 f 4 m, l, g , y) , (5)

6 2 y2  3 y3 cos y0  y1 6 8 y3  3 y2 cos y0  y1
where: f1 m, l , y 2
, f 2 m, l , y ,
ml 16  9 cos y0  y1 2 m l 2 16  9 cos y0  y1 2

ml 2 §¨ 36 >2 y2  3 y3 cos y0  y1 @ >8 y3  3 y2 cos y0  y1 @ ·

sin(y0 ) ¸ ,
3g
f 3 m, l , g , y  sin y0  y1 
2 ¨ m 2l 4 16  9 cos y0  y1 4 l ¸
© ¹
ml 2 §¨ 36 >2 y2  3 y3 cos y0  y1 @ >8 y3  3 y2 cos y0  y1 @ ·
sin y0  y1  sin(y1 ) ¸ .
g
f 4 m, l , g , y
2 m l¨ 2 4
16  9 cos y0  y1 4 l ¸
© ¹
§ y1 ·
¨ ¸
¨ y2 ¸
Here y is the vector ¨y ¸.
¨ 3¸
¨y ¸
© 4¹
Computer Package Mathcad 14 [22] uses different functions, which presents
certain numerical algorithms for solution systems of differential equations. One of
them is the function rkfixed(a, t0, tN, N, D) , which uses the algorithm of Runge-Kutta.
Here
§ T10 ·
¨ ¸
¨ T 20 ¸
a= ¨p ¸ is the vector of constants in the initial conditions, t0 is the first value of the
¨ T10 ¸
¨ pT ¸
© 20 ¹
time t and tN – its last value; N is the number of intervals between t0 and tN; D – is the
§ f1 (m, l , y ·
¨ ¸
¨ f 2 ( m, l , y ¸
notation of the vector D(t,y) = ¨ of the right hand sides of the system (5).
f 3 ( m, l , g , y ¸
¨ ¸
¨ f 4 (m, l , g , y ¸¹
©
The function rkfixed(a, t0, tN, N, D) returns the matrix that is convenient to assign to
another variable Z: Z = rkfixed(a, t0, tN, N, D). The first column with number 0
contains the values of the variable y0 = t at the nodes. The second column with number
1 – the values of variable y1 = T1; the third column with number 2 – the values of
variable y2 = T2; the fourth column with number 3 – the values of variable y3 = pT1 ; the
fifth column with number 4 – the values of variable y4 = pT 2 

8
 RESULTS

The results of computations of y3 = pT1 and y4 = pT 2 for t0 = 0, tN = 100, N = 1,000,

l = 5, m = 1, g = 9,807 in the case of initial values (4) are given in the following tables,
copied from the worksheet of the Computer Package Mathcad. These tables show that
because of the approximate value of the S = 3.141592653589793…, the motion of the
double pendulum is possible even from the initial values (4).

0 0
0 0 0 0
1 -9.00728588395204· 10 -16 1 -9.00728588395204· 10 -16
2 -1.8014571767904077·10-15 2 -1.8014571767904077·10-15
3 -2.7021857651856114·10-15 3 -2.7021857651856114·10-15
4 -3.6029143535808155·10-15 4 -3.6029143535808155·10-15
5 -4.503642941976019·10 -15 5 -4.503642941976019·10 -15
6 -5.404371530371223·10 -15 6 -5.404371530371223·10 -15
pT 7 -6.305100118766426·10 -15 pT 7 -6.305100118766426·10 -15
1 2
8 -7.205828707161631·10 -15 8 -7.205828707161631·10 -15
9 -8.106557295556835·10 -15 9 -8.106557295556835·10 -15
10 -9.007285883952038·10 -15 10 -9.007285883952038·10 -15
11 -9.908014472347242·10 -15 11 -9.908014472347242·10 -15
12 -1.0808743060742447·10-14 12 -1.0808743060742447·10-14
13 -1.170947164913765·10 -14 13 -1.170947164913765·10 -14
14 -1.2610200237532854·10-14 14 -1.2610200237532854·10-14
15 ... 15 ...

Figs. 4 and 5 show the dependence of the inclination angles T and T of links of the
double pendulum (Fig. 1) on time in the case of the initial conditions (3) (Fig. 2).

Figure 4. The dependence of inclination angle T1 on time in the case

of initial values (3).

9
 Figure 5. The dependence of inclination angle T2 on time in the case
of initial values (3).

Figs. 6 and 7 show the dependence of the inclination angles T and T of links of
the double pendulum (Fig. 1) on time in the case of the initial conditions (4) (Fig. 3).

Figure 6. The dependence of inclination angle T1 on time in the case of initial values (4).

Figure 7 The dependence of inclination angle T 2 on time in the case

of initial values (4).

The diagrams in Figs. 4-7 show that there is chaotic motion of the double pendulum.
The video clip http://www.youtube.com/watch?v=BFIJTAzYdvo shows a series of
frames from this video clip.

10
 10 10 10

 10 0 10  10 0 10  10 0 10

 10  10  10

10 10 10

 10 0 10  10 0 10  10 0 10

 10  10  10

Figure 8. Frames from composed video clip from initial position in Fig. 2.

The video clip http://www.youtube.com/watch?v=N9eWZR3I-VQ shows the

chaotic motion of the double pendulum from initial position in Fig. 3. Fig. 9 shows a
series of frames from this video clip
10 10 10

 10 0 10  10 0 10  10 0 10

 10  10  10

10 10 10

 10 0 10  10 0 10  10 0 10

 10  10  10

Figure 9. Frames from composed video clip from initial position in Fig. 3.

11
 CONCLUSIONS

This paper shows the features of the Computer Package Mathcad in solution of
nonlinear system of differential equations. The nonlinear equations of large oscillations
of a double pendulum are solved and the chaotic motion is found out. Note that it is
easy to make a physical model of a chaotic double pendulum. Some examples are
given in [13]–[18]. Fig. 10 shows a frame from video clip with the chaotic motion of a
simple physical pendulum [13].

Figure 10. A frame from video clip, showing chaotic oscillation of the
physical model of a double pendulum.

REFERENCES

[1] http://scienceworld.wolfram.com/physics/DoublePendulum.html
[2] http://demonstrations.wolfram.com/DoublePendulum/
[3] http://www.maths.tcd.ie/~plynch/SwingingSpring/doublependulum.html
[4] http://www.physics.northwestern.edu/vpl/mechanics/pendulum.html
[5] http://tabitha.phas.ubc.ca/wiki/index.php/Double_pendulum
[6] http://www.physics.usyd.edu.au/~wheat/dpend_html/
[7] http://www.physics.usyd.edu.au/~wheat/sdpend/
[8] http://www.youtube.com/watch?v=Uzlccwt5SKc&NR=1
[9] http://www.myphysicslab.com/dbl_pendulum.html
[10] http://www.chris-j.co.uk/rott.php
[11] http://freddie.witherden.org/tools/doublependulum/
[12] http://www.imaginary2008.de/cinderella/english/G2.html
[13] http://www.youtube.com/watch?v=8VmTiyTut6A
[14] http://www.youtube.com/watch?v=z3W5aw-VKKA
[15] http://www.youtube.com/watch?v=O2ySvbL3-yA
[16] http://www.youtube.com/watch?v=cXRXWy5KMgs
[17] http://www.youtube.com/watch?v=Lz8B2ek9o84
[18] http://www.youtube.com/watch?v=BrMQ7G1DtPw
[19] http://en.wikipedia.org/wiki/Double_pendulum
[20] Ü. Lepik, L. Roots. Teoreetiline mehaanika (In Estonian), 1971, Tallinn, 483 p.
[21] J. Kirs. Mehaanikaliste süsteemide võnkumised (In Estonian), 1998, Tallinn, 124 p.
[22] http://www.msmiami.com/custom/downloads/Mathcad14_Users_Guide_English.pdf

12