Investigation Methods
Calculation of Lyapunov Exponents
Two methods are available:
- Calculation based on the method of Wolf et al.
(requires the implementation of the linearized system function by the user)
- Calculation based on the modified method of Wolf et al.
(no linearized system function is required)
For maps, hybrid maps, CMLs, ODEs, hybrid ODEs, DDEs and hybrid DDEs:
- Lyapunov-Exponents (Lyapunov spectrum)
- Lyapunov-Dimension
Calculation of Poincaré sections
Several methods are available:
- Calculation based on cross-sections with a hyperlayer
- fixed for all parameter settings
- dependent on current parameters
- Calculation using user-defined conditions
- For hybrid systems: calculation using discrete state change
Symbolic image analysis
Several methods are available to investigate the global structures of dynamical systems:
- Detection of trajectories and limit cycles including
- p-periodic trajectories
- stable and unstable limit cycles
- variable error tolerance
- Localisation of attractors and their basins
- Morse spectrum (experimental)
All methods are based on the theoretical works of Prof. George Osipenko of the St. Petersburg State
Polytechnical University (SPbSPU) and are developed in cooperation
with him.
Box counting
Calculation of:
- the invariant (natural) measure of an attractor
- fractal dimensions
- in progress: Lyapunov exponents (Lyapunov spectrum)
Period analysis
(For dynamical systems discrete in time)
Determination of:
- the period of an attractor
- one-dimensional parameter intervals where the corresponding
attractors show a certain period
- bifurcation diagrams based on periodic attractors
- bifurcation diagrams based on aperiodic attractors
(saving of last N points of the orbit)
Attractors usually will be saved as time series and -
for single-dimensional discrete maps - as a file containing
the graphical iteration data.
Region analysis
(based on Period analysis)
(For dynamical systems discrete in time)
Determination of:
- regions in two-dimensional parameter spaces where the
corresponding attractors show the same period
Cobweb diagrams
(based on Period analysis)
(For one-dimensional dynamical systems discrete in time)
The orbit data will be saved as time series containing the graphical iteration data.
Spectral analysis
(based on the Fast Fourier Transform (FFT) of the
FFTW library)
Calculation of the:
- real- and imaginary part of the Fourier transform
- Power spectrum
- Autocorrelation function
Principal component analysis
For multi-dimensional signals:
- eigenvalues
- eigenvectors
- covariance matrix
Symbolic sequence analysis
There are three possibilities to generate symbolic sequences based on:
- implicit partitioning of the state space according
to (L/R)-symbolic dynamics
- implicit partitioning of the state space according
to (+/-)-symbolic dynamics
- explicit partitioning of the state space according
to a user-defined partitioning function
Depending on the computed symbolic sequences the following values will be calculated:
- probabilities of single symbols within periodic parts of
the attractor
- probabilities of single symbols within the entire orbit
- entropy of the symbolic sequence within periodic parts of
the attractor
- entropy of the symbolic sequence within the entire orbit
General trajectory analysis
Saving of:
- the trajectory (several possibilities)
- mean value and standard deviation of the components of the state vector
- correlation coefficients between components of the state vector
minimum and maximum values of the component
- velocity of the components
- average velocities
- wave numbers
- system classes with a spatial component - i.e. CMLs and CODELs -
saved as PGM graphics
Conditions checker
This method performs checks for some conditions and saves the
time and the length of the trajectory from the beginning of the
iteration, until a condition is fulfilled for the first time.
The following conditions are implemented:
- The trajectory runs to a fixed point
- The trajectory runs to an arbitrary point given by the user
- The trajectory leaves an area, which can be a rectangle or a sphere
Last modified: 2005-04-25
Contact: AnT