Sie sind auf Seite 1von 6

2013

2013Sixth
6th International
InternationalConference
Conferenceon
onEmerging
EmergingTrends
TrendsininEngineering
Engineeringand
andTechnology
Technology

Analysis of Chaos in Double Pendulum


Himanshu Jain Ashish Ranjan Prof Kshitij Gupta
Mechanical Engineering Mechanical Engineering Mechanical Engineering
Department Department Department
Indian Institute of Indian Institute of Indian Institute of
Technology, Delhi Technology, Delhi Technology, Delhi
New Delhi, India New Delhi, India New Delhi, India
himanshujain.iitd@gmail.com ashish2993@gmail.com kgupta@mech.iitd.ac.in

Abstract-In this paper, we present the double pendulums remain together. For larger initial
analysis of chaotic vibrations in double pendulum displacements, the trajectories from each other
by experimental means. We shall illustrate the diverge due to chaotic behavior, which is clearly
sensitivity of its motion to initial conditions i.e., observed in a side view. A stroboscope flashing 25
the angles of each pendulum from the vertical, times per second was used to illuminate the
Tand I at initial time t=0. Here we present the pendulum, which was placed in a dark room. An
method of video processing as an experimental electromagnet was used to give initial displacements
technique to study the motion of double pendulum to the two pendulums. Lyapunov exponent is
in time domain. We have studied the dependence calculated from the experimental readings repeated
of the nature of motion of double pendulum for the same initial conditions. For large initial
broadly classified into chaotic and non-chaotic, on displacements, the Lyapunov exponent works out to
the initial conditions in which the angles θ and I be a positive quantity, thereby clearly establishing the
acquire values in the full range of -180 to 180 chaotic response.
degrees in various combinations. Levien and Tan [2] performed experiments
Also, we would be showing that the to study chaos in a simple passive double pendulum.
phenomenon of chaos does not only depend on the An optical encoder is used to measure the angles of
energy levels but there are other governing factors the two pendulums. First the two natural frequencies
also. As for example, in double pendulum, it is the are measured by initially putting the system in a
orientation of pendulums. natural mode. Then the large angle motion of the
Index Terms- Chaos, Double Pendulum, Energy pendulum is obtained experimentally. Authors have
levels, Orientation established the sensitive dependence of motion on the
initial angles of the pendulum, in the case of large
I. INTRODUCTION swings. This is characteristic of the chaotic motion of
the pendulum. Lyapunov exponents are evaluated to

T
HERE have been several studies on dynamics
of double pendulum addressing different verify the presence of chaos. Theoretical results for
issues. Some of the early experimental work the case under consideration show the onset of chaos
is by Shinbrot [1] and Levien and Tan [2]. as the initial angles approach the condition, T(0) =
Shinbrot [1] performed an experimental I(0)~65o.
study to analyze chaos in double pendulum. Zhou and Whiteman [3] analyzed chaotic
Experiments clearly show the sensitive dependence motion in a double pendulum. Following a
Newtonian approach, the equations of motion of the
of response to initial conditions. Two identical
double pendulum are derived in terms of two coupled
double pendulums are placed in tandem. Both the second order nonlinear ordinary differential
double pendulums are released from the same initial equations. This is the most generalized form of the
displacements at the same instant. For small initial governing equations. These are simplified for small
displacements, the motion trajectories of the two displacements T andI, the differential equations still

978-1-4799-2560-5/13 $31.00 © 2013 IEEE


978-1-4799-2561-2/13 172
171
DOI 10.1109/ICETET.2013.50
remaining nonlinear. Further simplification is made the energy level is increased, although no threshold
by assuming a slow motion so that the angular level of energy is arrived at. In the present paper the
velocity terms are dropped. This results in linearized initial angles T and I, are varied in the full range of -
system equations. These equations are solved using
180 to 180 degree in various combinations, and the
the 4th order Runge-Katta method. Phase plane plots
are obtained from linearized equations, and the effect motion is identified to be chaotic or non-chaotic
of two parameters i.e., the mass ratio, and relative (periodic). This condition is plotted on a map with
lengths of the individual pendulum, is studied. Phase two parameters θ and I. This clearly demarcates the
plane plots are also obtained from the solution of regions of chaotic and non-chaotic motions
nonlinear equations, and an irregular motion is associated with the initial conditions on θ and I. Such
predicted, which is indicative of the chaotic behavior.
a map does not exist in literature, and provides a
The same is also confirmed from the sensitivity of the
system response to initial conditions. comprehensive picture as regards the chaotic motion
Stachowiak and Okada [4] studied chaotic of a double pendulum. The experimental results
motion by constructing Poincare´ sections and presented in the paper are for a pendulum with
Bifurcation diagrams for certain characteristic values specific data. Theoretical parametric studies with
of energy. Authors have attempted to correlate mass and length parameters however can be carried
qualitatively the energy level with the type of
out to assess the effect of these parameters on the
response. Analysis has shown that the system passes
into chaos as the energy level is increased. At low map.
energy levels a regular periodic motion is observed. Also, the effort has been made to show that
To illustrate this fact clearly, bifurcation diagrams are the energy is not the only governing parameter for
drawn with respect to energy level, as parameter. The chaos. Other parameters like orientations also play
largest Lyapunov exponent is also calculated to role in chaotic motion.
substantiate the conclusions derived from Poincare
maps and bifurcation diagrams.
II. EXPERIMENTAL SETUP:
In a recent study Roy et al.[5] analyze the
motion characteristics of double pendulum at low The two pendulums of the double pendulum
energy levels i.e., levels close to the critical value are bars with uniform cross-section, and hence have a
required to make the outer pendulum rotate. In the uniform mass per unit length. The two pendulums are
first part of the study, it is shown by both the mounted on rolling element bearings, which can be
harmonic balance and the Lindstedt–Poincare assumed to be frictionless. Figure 1 gives a schematic
analyses, that the natural frequency of the “out-of- of the double pendulum. The upper pendulum has
phase” mode drops sharply as the initial displacement
total mass m1= 0.42 kg and a length l1=0.335 m. The
is increases. The in-phase natural mode of the double
pendulum remains unchanged. In the second part of lower pendulum has a total mass m2= 0.12 kg and a
the analysis which involves numerical integration of length l2=0.1715 m.
nonlinear system equations, similar results as by To study the motion characteristics of the
Stachowiak and Okada [4] are obtained. It is shown pendulum, we need a reliable technique which does
that as the energy level associated with initial not infract with the system. The foremost objective
displacements is increased, the motion dynamics
was to record the orientation of upper and lower
changes drastically leading to chaotic behavior. One
result shows chaotic motion at an energy level of pendulum with the passage of time. Since any
2mgl even though it is not clear if it is the threshold hindrance in the motion of the pendulum can lead to
value. erroneous results, optical measurement techniques
It is evident that there are no comprehensive which are practically nonintrusive were needed. The
results as regards the effect of initial condition videos of the pendulum are the reliable source of the
(angles) on the chaotic behavior of the double information about the varying angles with time. The
pendulum. A general observation by several authors video processing techniques are hence, a reliable
is that small angle oscillations are periodic, while the experimental technique.
large angle oscillations turn into chaotic mode. There
are some attempts to correlate the chaotic behavior
with the energy level. Onset of chaos is observed as

172
173
boundary as in figure 4 and records their orientations.
Thus, using the image processing techniques [6] we
can get the data points pertaining to the angular
position of the two pendulums from the vertical.
The coordinate system is defined such that
angle of a pendulum from the vertical measured in
the anti-clockwise direction is considered to be
positive, or else negative. So the range of
measurement lies in the interval of (-180q, 180q). In
fact the boundary angle -180q is same as 180q. We
set the initial angle of upper pendulum at a gap of
every 15q in the range of [-90q, 90q] (i.e. at -90q, -
75q, -60q, -45q, -30q, -15q, 0q, 15q, 30q, 45q, 60q, 75q
and 90q). For each initial angular position of the
upper pendulum the lower pendulum was set at an
interval of 45q in the range (-180q, 180q) (i.e. at -
135q, -90q, -45q, 0q, 45q, 90q, 135q, 180q).
Initial velocity of both the pendulums was
Fig 1: Double Pendulum: It is hinged at O1and connected with zero. So the initial condition was defined by angular
each other at O2. The angle of upper pendulum with vertical is
displacements imparted to the two pendulums. Thus,
T and that of lower isI.
for these 117 initial conditions, variation of angles Vs
The videos at 23 frames per second were time was obtained for further analysis.
recorded and to maintain a proper contrast, the The varying parameters of concern are the
pendulums were made white and the background was angle and angular velocities of the two pendulums.
kept dark. For the ease of identification of the Hence we have 4 parameters. To obtain the Poincare
pendulums by the program, some part of the upper maps we need to fix two parameters so that we can
pendulum was painted black. The videos were define a Poincare plane and take the projection on it.
processed using a program in which videos were We select those data for which the angle of upper
disintegrated into still photos. Figure 2 shows a frame pendulum is close to zero and the angular velocity of
of video in which the two pendulums are at some the upper pendulum is positive. The plot of angular
angular positions in the motion. Thus analyzing these velocity Vs angle for the lower pendulum for the
still photos we can get the data points at 0.0435 selected data gives the Poincare map. It was observed
seconds interval i.e., 23 data points per second. Each that choosing another Poincare plane does not alter
photo was programmed to be converted into a binary the characteristics of the motion but the number of
image as in figure 3 and then the program identifies data points in the map changes.
the upper and lower pendulum and encloses them in a

Fig 2: Binary Image Fig 4: Bounded image showing the identified


Fig 3: Image of pendulum in motion upper and lower pendulum

173
174
III. EXPERIMENTAL RESULTS AND ANALYSIS
We use here a (T,I) plot as a tool to study
the nature of the motion and also discuss its
importance in the study of motion of chaotic systems.
The motion in the present study is broadly classified
as chaotic or non-chaotic. The Poincare maps
generated for all the data points (initial conditions)
were analyzed. It helps in identifying whether the
motion is chaotic or non-chaotic.
The poincare maps for the initial conditions
as stated in each figure was plotted for both upper
pendulum and the lower pendulum. Figures 5 and 6
show some typical Poincare maps for the lower and
upper pendulum respectively for different initial
conditions. Fig 6. Poincare maps for the upper pendulum for different
initial conditions showing the effect of increasin θ (ɸ =0)

In figure 7, the red stars show the non-


chaotic motion, while the blue stars show the chaotic
motion. The dotted curve in figure 4 approximately
demarcates the regions of chaotic and non-chaotic
motions of the double pendulum. It may be noted
that, T and I being the angles of lower and upper
pendulum, the characteristics of motion for the cases
of initial conditions (T,I) and (-T,-I) would be
similar. This is obvious from the fact that magnitude
and relative orientation of the two pendulums are
same and only the direction is reversed. This also
implies that the upper and the lower regions of the
Fig 5. Poincare maps for the lower pendulum for different map in Fig. 4 will be a mirror image of each other.
initial conditions showing the effect of increasing θ (ɸ =0)
The plot shows that motion is non-chaotic
The nature of the maps changes as the for smaller values of angle T for the upper pendulum.
motion changes from non-chaotic to chaotic. The For T> 45q, the motion is always chaotic and for T<
phase plane becomes more and more sparse as we go 45q, there are different values of I for which the
towards chaotic motion. The plot appears to be chaotic motion occurs. Also as we increase T, the
wandering in space. Initially the plot either shows a range of I for which the motion is non-chaotic
damping characterstic or is clustered around few decreases.
points which is a characterstic of non-chaotic motion. We define the orientation of pendulum in
As we move towards chaotic motion the plot which the product (T*I) of the angle is negative as an
becomes random distribution of points having no anti-orientation and if it is positive then it is in a
clear pattern. sym-orientation.
Then, there are cases when the system with
the anti-orientation (e.g.: (T,ф)= (30,-45), ( 45,-45) or
(45,-90)) show the chaotic motion, while the
corresponding sym-orientations (i.e. (T,ф)= (30,45), (
45,45) or (45,90)) show non-chaotic motion. It may
be noted that the potential energy in an anti-

174
175
orientation and the corresponding sym-orientation extent of divergence of the motion which we have
will be same. This shows that the potential energy at used in theoretical analysis.
initial condition is not the only variable on which Experimental results clearly show that initial
nature of motion depends. It also depends on the energy level of the system in the form of potential
relative orientation of the upper and the lower energy is not the only parameter on which chaos
pendulums at t=0. Thus choosing the initial energy depends. The relative orientation of pendulums is
level as the only parameter to determine the nature of one of the fundamental parameters that affect the
the motion would not be appropriate. motion. This is illustrated in the figure 5 below.
Hence (T,ф) diagram as shown in Fig. 7 is Since the angle of the upper pendulum has a greater
an effective tool to study the motion of the double contribution in overall energy we have made a plot of
pendulum as it incorporates both the parameters i.e. (abs (ϴ1)-abs (ϴ2)) v/s ϴ1.
energy level and the relative orientation of the two
pendulum.

Fig 8: (abs (ϴ1)-abs (ϴ2)) v/s ϴ1. The colour shows the energy
level. The squares and circles are non-chaotic and chaotic
points respectively. We can see that even for the same or low
energy levels, configurations with high angle difference
Fig 7: The experimental plot of (ϴ, ф) diagram showing the between the pendulums at t=0 exhibit chaotic character.
chaotic and non-chaotic regions
The present work has been an attempt to
IV. CONCLUSION find some pattern in the system as regards the
For the study of chaos, we need some dependence of nature of the motion, chaotic versus
reliable techniques. The video processing technique non-chaotic, on the initial conditions of the
is useful because of its accuracy. In fact this pendulum. As expected, we found that the
technique can be used for other system analysis also. experimental chaotic region is symmetric (Fig. 4)
Also, in the observations the time difference between about the vertical axis. Our work also shows that
the consecutive data point is small (here, 1/23 of a techniques such as video processing for determining
second). And this is reasonable in this case. As the the changing orientation with time is really useful
frequency of oscillation is small, a camera with and can be used in various other systems.
capable of recording 23 frames per second is suitable.
For other system where frequency of vibration is very REFERENCES
high a high speed camera can be used. [1] Troy Shinbrot, Celso Grebogi, Jack Wisdom, James A.
Yorke “Chaos in Double Pendulum” on 14 December
Secondly, as discussed earlier, the present
1991.
work has emphasized on the usefulness of the [2] R.B. Levien (Department of physics, Princeton
Poincare map as a one of the means to study the university, New Jersey 08544) and S.M. Tan
changing characteristics of the motion. The effect of (Department of Physics, University of Auckland, New
Zealand) “Double Pendulum: An experiment in chaos”
slight variation in the initial condition has been
1 April 1993.
another way of justifying the occurrence of chaos and [3] Ziliang Zhou and Chad Whiteman, “Motion of double
the Lyapunov Exponent is a way to quantify the pendulum” in Nonlinear Analysis, Theory, Methods &
Applications Vol. 26 No 7. Pp 1177-1191, 1996.

175
176
[4] Tomasz Stachowiak, Toshio Okada, “A numerical dynamics of a double pendulum at low energies” 19
analysis of chaos in the double pendulum”, Feburary 2013.
NonlinearAnalysis, Theory, Methods & Applications, [6] Image processing toolbox from Matlab
Vol. 26, No. 7, 1996, pp. 1177-1191. http://www.mathworks.in/products/image/
[5] Jyotirmoy Roy · Asok K. Mallik · Jayanta K.
Bhattacharjee in “Role of initial conditions in the

176
177

Das könnte Ihnen auch gefallen