Analysis of Chaotic Dynamics  2. Stability
Formal Metadata
Title 
Analysis of Chaotic Dynamics  2. Stability

Alternative Title 
Analyse chaotischer Schwingungen  2. Stabilität

Author 

License 
CC Attribution  NonCommercial  NoDerivatives 3.0 Germany:
You are free to use, copy, distribute and transmit the work or content in unchanged form for any legal and noncommercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor. 
Identifiers 

IWF Signature 
C 1740

Publisher 

Release Date 
1990

Language 
English

Producer 

Production Year 
1989

Technical Metadata
IWF Technical Data 
Film, 16 mm, LT, 156 m ; F, 14 1/2 min

Content Metadata
Subject Area  
Abstract 
The stability; the spacial behaviour of nonlinear oscillations, is explored using Lyapunow exponentials and Ueda diagram. Bifurcations are shown. Very useful is the method of cell figures, which allows a quick calculation of longterm stability. Minimal changes of certain parameters decide wether the system behaviour is periodic or chaotic.

Keywords  dynamics, nonlinear dynamics, chaotic phase curve Lyapunow exponentials Ueda diagram bifurcation stability analysis cell figures Markov chains computer simulation system behaviour, periodic system behaviour, chaotic 
00:00
Trajectory
Group action
Dynamical system
State of matter
Length
Multiplication sign
Insertion loss
Parameter (computer programming)
Dimensional analysis
Mathematician
Diagram
Physical system
Potenz <Mathematik>
Logarithm
Sigmaalgebra
Generating set of a group
Mass flow rate
Perturbation theory
Time domain
Sequence
Orbit
Arithmetic mean
Frequency
LjapunovExponent
Vector space
Structural stability
Uniformer Raum
Steady state (chemistry)
Normal (geometry)
Bounded variation
Spacetime
Existence
Perturbation theory
Mathematical analysis
Distance
Force
Frequency
Average
Divergence
Structural stability
Initial value problem
Addition
Exponentiation
State of matter
Mathematical analysis
Mortality rate
Limit (category theory)
Numerical analysis
Computer animation
Quotient
Coefficient
04:13
Area
Potenz <Mathematik>
Dynamical system
Multiplication
State of matter
Variety (linguistics)
Multiplication sign
Graph (mathematics)
Range (statistics)
Algebraic structure
Parameter (computer programming)
Oscillation
Chaos theory
Computer animation
Different (Kate Ryan album)
Structural stability
Velocity
Calculation
Steady state (chemistry)
Diagram
Identical particles
Resultant
Physical system
Spacetime
08:13
Point (geometry)
Trajectory
Threedimensional space
INTEGRAL
State of matter
Multiplication sign
Sheaf (mathematics)
Parameter (computer programming)
Attractor
Theory
Force
Fraction (mathematics)
Frequency
Poincaré map
Plane (geometry)
Determinant
Initial value problem
Condition number
Physical system
Area
Surface
Model theory
Sampling (statistics)
Harmonic oscillator
Evolute
Time domain
Numerical analysis
Orbit
Chaos theory
Computer animation
Ring (mathematics)
Steady state (chemistry)
Chain
Spacetime
11:29
Area
Probability distribution
Computer animation
Green's function
11:58
Computer animation
12:18
Force
Area
Computer animation
Multiplication sign
3 (number)
Similarity (geometry)
Parameter (computer programming)
Time domain
Attractor
Phase angle
2 (number)
Physical system
00:25
Stability analysis normally deals with the effects of perturbations on the state of a system. For stable systems one expects that small perturbations of the state of the system have only limited effects. Regarding technical systems it is important to know whether there exists a practically useful domain of attraction of the steady state solution or not. In technical systems, one not only has perturbations acting on the state of the system but also perturbations of the system itself. To numerically study the stability of a nonlinear dynamic system we use Lyapunov exponents. By means of Lyapunov exponents we can obtain stability diagrams in parameter space. We have already observed that variation of parameters may often essentially change behaviour. Lyapunov exponents quantitatively characterize the average exponential divergence or convergence of neighbouring trajectories with respect to a reference trajectory. To measure the average speed with which a given displacement is amplified, we introduce a vector a which measures the initial distance between two neighbouring trajectories. As time increases, the distance may change and we would like to know how the vector a evolves with time. Therefore we take the norm of "a of t" which means the length of a. The mean exponential rate of divergence of the two initially close trajectories is then defined by the limit of 1 over t times the natural logarithm of the quotient a of t over a. Hence, we consider the stretching of an infinitesimally small distance a after an infinitely long time. Sigma is called Lyapunov exponent. The Russian mathematican Oselédec could prove the existence of sigma and that it is finite. This simple notion of the divergence of a normalized vector can be generalized to higher dimensions. The number of exponents coincides with the dimension of the state space. For a continuous dynamic system the behaviour along the flow is in the average constant, consequently, the exponent sigma along the flow is equal to zero. In addition to the case of the zero exponent along the flow, negative exponents will signal periodic orbits. At least one positive exponent indicates divergence of initially close trajectories and therefore a chaotic orbit. In the domain of periodic solutions zero exponents mean the loss of stability of one orbit and gain of stability of another orbit, i.e. zero exponents characterize bifurcations. By means of Lyapunov exponents we can generate stability diagrams, indicating the system's behaviour e.g. for damping coefficient d against forcing amplitude a. We generated a sequence of such diagrams by starting at different initial conditions. The initial
04:14
velocity was kept constant at xdot = 0, and only the initial displacement x was varied, starting at x = 2. Blue areas indicate negative, yellow and red positive exponents, and green areas zero exponents, which indicate bifurcations. In these graphs we stepwise increase the displacement of x by delta x = 0.5. The varying structure within the stability diagram indicates different longterm behaviour depending on initial states for identical values of the system's parameters. Superimposing all diagrams resulting from our calculations by using the Lyapunov exponents approach yields this picture. Hatched regions indicate multiple steady states. That is, the same oscillator with fixed values of parameters may have different longterm behaviour. This picture represents a small portion of Ueda's diagram already used before. For a large range of the parameters a and d the complete diagram shows a variety of different behaviours. Ueda also observed different kinds of behaviour for the same parameter values. Again these regions are indicated by the vertically hatched subdomains. Let's now restrict ourselves to the most interesting area for a from 5 to 15 and d from 0 to 0.2. The results obtained via Lyapunov exponents confirm Ueda's observations. The phenomenon of multiple or coexisting final behaviours in the same system is an important common feature of nonlinear dynamic systems.
07:21
Besides different periodic oscillations even chaotic motions may coexist; that is, different initial states may lead to completely different steady state solutions. As an example we pick out of the parameter space a = 12, d = 0.1, all other parameters being fixed. Here we have coexistence of a periodic oscillation and a chaotic motion. The time history of the chaotic motion is already well known to us.
08:14
For the same parameter values but starting at a different initial state we obtain the time history of the harmonic oscillation. In the threedimensional ring model the coexistence of the periodic motion and the chaotic motion is obvious. The Poincaré map will finally be used to visualize the different final behaviours in the surface of section. Starting with this initial state, the motion settles down to a strange attractor of a familiar shape. For another initial condition the final behaviour is represented by a single point. Different initial states obviously become entrained to different final behaviours. In cases where different final motions are possible, the determination of the domain of attraction of the final motions must be done for the given state space. Simply using simulations and evaluating many starting conditions can be very time consuming. A more pragmatic approach is the cell mapping method, first described by C.S. Hsu from Berkeley, U.S.A. It determines the evolution of a system based upon a large collection of very small cells. For our example we subdivide the domain of interest of the x, xdot plane or surface of section into small subdomains. For convenience we use rectangular cells, but in general the shape of the cells can be arbitrary. Within each cell we use a certain number of initial sample points. Starting from these points, in the course of time the system evolves along trajectories. The end point of each trajectory after one forcing period is determined by time integration. The relative probability of a cell mapped onto other cells is calculated by determining what fraction of the points ends within a target cell. This information can be used in the context of Markov chain theory to determine steady states in the surface of section. For the parameter value a = 12 the chaotic solution is represented by this strange attractor. The coexisting oneperiodic orbit is represented by a single point. The steady state solutions from the cell mapping cover the same area of the surface as the Poincaré map. Let us now supplement the attractors with their domains of attraction: From the green area,
11:20
the system ends up in the strange attractor with 100 % probability. Starting in the yellow area with 50  100 %,
11:36
and with 0 to 50 % probability from the violet area. The probability distribution for the periodic solution is obtained in the same way.
11:49
Starting in "green" we end up with 100 %, starting in "yellow" we end
11:59
up with 50 to 100 % and starting in "violet"
12:09
we eventually end up with 0 to 50 % in the periodic solution.
12:19
Finally we will study the time behaviour for the two coexisting domains of attraction. To settle down to the strange attractor from the most distant area of the domain of attraction it takes about 14 seconds.
12:39
From the outermost cells of the domain of attraction of the periodic solution it takes on the average 16 seconds to end up at the steady state. By using the cell mapping method, similar portraits for other system's parameters and other phase angles can easily be obtained.