> | restart: |
[1] Lynn Arthur Stewart: "Catastrophe Theory:the first Decade" , Science News, Vol 111, (http://www.stolaf.edu/people/steen/Papers/first_decade.pdf)
[2] Stewart, I.: "Elementary catastrophe theory" , IEEE Transactions on Circuits and Systems,ISSN: 0098-4094 ,pp.578 - 586
[3] John Guckenheimer: "The catastrophe controversy",The Mathematical Intelligencer,Volume 1, Number 1, 15-20, Springer,1977
Theory of Catastrophes
Catastrophic System : A system where a small and smooth change of a parameter may result in drastic (discontinuous) changes in system behavior
(e.g. equilibrium point looses its stability and new equilibrium point appears, this is refered as a bifurcation in theory of deterministic chaos )
Thom defined seven types of elementary catastrophes
E.C. Zeeman used the l Thom's theory for (mainly qualitative) modelling of nonlinear phenomena including processes in biological systems.This approach allows to describe nonlinear processes where small and smooth changes of parameters results in sudden change in behavior of a system
----------------------------------------------------------------------------------------------------------------------------
Catastrophic models (e.g. classical Zeeman's one ), can well describe behavior of modeled system from the qualitatvive point of view. Let's imagine a ball moving on a surface of system behavior (e.g. potential (defining) function f(x, p ), where x is the variable describing a state of a system and p is a vector of system parameters) , so we can see how is the system state going to change, what are the possible equilibria that can be reached for a given setup of p , we can observe hysteresis, achievable (stable) or unachievable (unstable) equilibra - bimodality, divergence.
Fig 1.: Five properties of Cusp catastrophy(adopted from Catastrophe theory by E.C. Zeeman, Trinity University, San Antonio, March 1995,http://zakuski.utsa.edu/~gokhman/ecz/c.html)
On a surface of possible system equailibria, that is defined by first derivative of defining potential function
> |
We can well see when there is a catastrophy to happen (biffurcation set is a border where a systems suddenly changes from on qualitative type of behavior to another one, i.e., there appears new equilibrium potins of a system).
Very basically, Zeeman's method says which equilibria will be achieved for a given setup.
Bellow in this text, you find implementations of catastrophic potential functions in nonlinear differential equations. This allows us to describe the time development of state variable (system behavior), i.e. how an equilibira will be achieved.
Thus from here, we are slowly approaching toward fundamentals of the field of nonlinear dynamical systems and their quantitative description in time <=> field of DETERMINISTIC CHAOS
(Note that according to Poincare-Bendixon theorem, time-continuous models may develop deterministic chaos starting from third-order of differenital equation systems )
For simplicity, we consider only a case where a nonlinearity is in the nonlinear potential function of a system fx (x, p ), where x is state variable, p is a vector of parameters (max 4 parameters) .
Let's consider simplest dynamical systems that may behave according to catastrophic models, so their potential function fx (x, p ) is nonlinear and the dynamics of the state variable is further defined by parameters a1, a2, a3, ...
...a3.x'''(t) + a2.x''(t) + a1.x'(t) +
d
/dx {
fx
(x(t),
p
) } = 0
p
=[a,b,c,d]=[p1,p2,p3,p4]
for example
> | ![]() |
Thus we can model system behavior quantitatively - describe system behavior from initial state to the resulting state in time.
If we have measured data, we may also optimize the parameters of models to those data.
=================================================================================================================
Catastrophe Fold
1-st type of catastrophes were stability is lost with the change of a parameter.. If a system is in equilibrium (but unstable one), then a small change of p1 results in loss of stability - a ball appears out of surface extremum where the potential function had been zero before p1 changed.
> | ![]() |
Defining function (potential function) fx (x, p )
> | ![]() |
------------------------------
Graph of potential function as a function of variable system parameter p1
> | xmin:=-5;xmax:=5; p1min:=-1.5;p1max:=1.5; |
> | ![]() |
A ball on the surface represents a system state. Unless p1 changes, the ball moves on a curve -a cubic parabola.
Graph of system equilibria
If x is a state variable and
fx
(x,
p
) is potential function of a system where
p
is vector of parameters,
then the set of possible resulting equilibria to where system can arrive is implicitly defined as
> |
> | implicitplot(dfx,p1=-10..10,x=-5..5,grid=[100,100]); |
Here we can see, that there are not appearing not dissapearing new equilibria states for p1>0 . One branch represents stable equilibria while the other branch represents unstable equilibria. The system state (a ball) can move toward instability (fall down) or retunr to a stable equlibria depending on its intial state.
------------------------------
Differential Equation
(A simple dynamics of a ball on a surface of behavior)
> |
> | a2:=2: a1:=1: p1:=4: u(t):=0: DR; |
Equilibira:
> | xp:=solve(dfx,x); |
> |
Gradient field in state space
> | dfieldplot([deqns[1],deqns[2]],[x1(t),x2(t)], t=-1..1,x1=-3..3, x2=-5..5,title=`Fold`,arrows=medium,color=blue); |
---------------- Solving De ---------------------
Initial conditions
> | x0:=-1.99999999999; dx0:=0; |
> | tmax:=60: xmin:=-5: xmax:=5: dt:=0.05: |
> | DEplot(DR,x(t),t=0..tmax,[[x(0)=x0,D(x)(0)=dx0]],x=xmin..xmax,stepsize=dt,color=black); |
--------------------------------------------------------------
--------------------------------------------------------------
Catastrophe Cusp
Second from elementary catastrophes. With this catastrophe, stability dissapears and new equilibrium point may appear.
> | ![]() |
Warning, the name changecoords has been redefined
Definiing function (potential function) fx (x, p )
> | fx:=x^4/4+p1*x+p2*x^2/2; |
------------------------------
Graph of potential function as a function of variable system parameter p2
> | xmin:=-3;xmax:=3; p1:=1; p2min:=-3;p2max:=1; |
> | 'fx'=fx; plot3d(fx,x=xmin..xmax,p2=p2min..p2max,grid=[100,100]); |
On this surface, we may imagine a ball whose position represents a state of a system. Unless p2 changes, a ball moves on a curve given by 4-order polynomial.
Graph of system equilibria
If x is a state variable and
fx
(x,
p
) is potential function of a system where
p
is vector of parameters,
then the set of possible resulting equilibria to where system can arrive is implicitly defined as
> |
> | implicitplot3d(dfx,p1=-1..1,p2=-3..1,x=-3..3,grid=[30,30,30]); |
Plocha určuje možné rovnovážné stavy systému (pozor, stabilní i nestabilní).
Fig .2: a 3: Adopted from Catastrophe theory by E.C. Zeeman, Trinity University, San Antonio, March 1995,http://zakuski.utsa.edu/~gokhman/ecz/c.html
------------------------------
Differential Equation
A simple dynamics of a ball on a surface of behavior
> |
> | a2:=2: a1:=1: p1:=0.5: p2:=-3: u(t):=0: DR; |
Equilibira::
> | xp:=solve(dfx,x); |
> |
Gradient field in state space
> | dfieldplot([deqns[1],deqns[2]],[x1(t),x2(t)], t=-2..2,x1=-2.5..2.5, x2=-2..2,title=`Cusp`,arrows=medium,color=blue); |
Do you see the equilibrium points in gradient field?
---------------- Solution of DE ---------------------
Ininital Conditions
> | x0:=-3; dx0:=0; |
> | tmax:=60: xmin:=-5: xmax:=5: dt:=0.05: |
> | DEplot(DR,x(t),t=0..tmax,[[x(0)=x0,D(x)(0)=dx0]],x=xmin..xmax,stepsize=dt,color=black); |
--------------------
-------------------------------------------------------------