Technion - Israel Institute of Technology

Faculty of Mechanical Engineering

Theory of Vibration (034011)

Numerical Analysis of Dynamical Systems
(i) Time Series
(ii) State Space
(iii) Poincare’ Maps

Using:
Matlab Simulation Procedure
(i) Simulation
(ii) Sampling
(iii) Graphics

Compiled By:
Sharon Pragai
Dr. Oded Gottlieb

October 2003

1
 Matlab Procedure: Dynamical System Simulation
Simulation of a dynamical system is obtained by numerical integration of a system of 1st
order differential equations with a corresponding set of initial conditions. Selection of the
integration routine (e.g. Runge-Kutta or Gear solvers) is based on the characteristics of the
dynamical system (e.g. smooth or stiff) and if automatic control of numerical error is
required (e.g. adaptive integration scheme).

The outcome of the simulation consists of the response time series. We emphasize that
dynamical systems subject to periodic excitation should be analyzed at steady state that is
to be determined. Consequently, the outcomes of a dynamical system simulation are:
i) time series (x(t));
ii) state space projections (y(x)) at steady state.

The Matlab function ODE45 is an adaptive solver that integrates a system of ordinary
differential equations using a 4th order Runge-Kutta formula (with a 5th order correction).
In order to obtain on-line help information, just type "help ode45" when inside the Matlab
window.

The Matlab command is:

[T, Y] = ODE45 (@F,TSPAN,Y0,OPTIONS)
where:
'F' - is the name of the file with the state space differential equations.
TSPAN - Tspan = [T0, Tfinal], initial and final integrating times.
If we look for a solution at specific times (T0, T1, T2, …, Tfinal), define
Tspan = [T0 T1 T2, … Tfinal], and Ti = t0:∆t:tf.
Y0 - is a column vector of initial conditions.
OPTIONS - contains the optional values for the desired accuracy like:
relative error tolerance 'RelTol' (le-3 by default) and
vector of absolute error tolerances 'AbsTol' (all components
le-6 by default). Omitting OPTIONS argument will choose default values .

2
Example
Solve the second order differential equation:
d 2x dx
2
+ δ + αx + βx 3 = γ cos(ωt) (1)
dt dt
dx
with the following initial conditions: x(0) = x 0 ; (0) = dx 0
dt
In time interval t = 0 ÷ 150 .
This is equivalent to a system of two first order equations, 5 parameters, and two initial
conditions:
dx 1
= x2
dt
(2)
dx 2
= γ cos ( Ωt ) − β x13 − αx1 − δx 2
dt
x1 (0) = x10

x 2 (0) = x 20

In order to solve the problem:

1. Create two separate external m-files:
(a) "system1.m": stores the information about the equation set structure.
(b) "solve.m": includes the remaining data and necessary commands.

2. Type "solve" from the matlab window and the file "system1" will be used automatically.
The results are stored in vector t and matrix x.

3. Plot solution graphs. The corresponding matlab command is "plot(x,y)" where x is an

argument column-vector (horizontal-axis), while y is a function column-vector (vertical
axis). File "solve.m" includes all the necessary commands, which shows how to plot the
graphs.

3
% File "system1.m" - Dynamical System
function xprime = system1 (t,x)
xprime=zeros(2,1);
% ========================== Parameter Definitions ===================

alpha=-1.0;
beta=1.0;
gamma=0.2;
delta=0.3;
w=1.2;
% ============================= State Space ========================
xprime(1)=x(2);
xprime(2)=gamma*cos(w*t)-alpha*x(1)-delta*x(2)-beta*x(1)^3;
================================================================

% File "solve.m" - Matlab Runge Kutta 4-5

close all; clear all; clc;
t0=0; % initial time
tfinal=150; % final time
x0=[0.5 0]'; % column vector of initial conditions
tspan=[t0 tfinal]'; % tspan can contain other specific points of integration.
option=odeset('RelTol', 1e-8, 'AbsTol', 1e-8);
[t,x]=ode45(@system1, tspan, x0, option);

% plotting the results

%=======================================================
% Free Vibration: gamma = 0:
figure(1)
subplot (2, 1, 1)
plot (t,x(:,1)); grid;
xlabel ('t [sec]'); ylabel ('x(1) [m]');
subplot (2, 1, 2)
plot (t,x(:,2)); grid;
xlabel ('t [sec]'); ylabel ('x(2) [m]');

4
% Forced Vibration: gamma = 0.2:
figure(2)
subplot (2, 1, 1)
plot (t,x(:,1)); grid;
xlabel ('t [sec]'); ylabel ('x(1) [m]');
% the steady state space is defined by the last 20 percent of the solution vector.
% n is the total number of time nodes during integration
n=size(t);
subplot (2, 1, 2)
plot (x(round(0.8*n):n,1), x(round(0.8*n):n,2)); grid;
xlabel('x(1)[m]'); ylabel('x(2) [m/s]');

5
 3
Numerical solution of x ''+ δx '+ αx + βx = γ cos ( ωt )

(1) Free vibration: γ = 0

Figure 1: Free vibration response: γ = 0, α = −1, β = 1, δ = 0.3, ω = 1.2 . a) x1(t). b) x2(t).

(2) Forced vibration response: γ = 0.2

Figure 2: Forced vibration response: γ = 0.2, α = −1, β = 1, δ = 0.3, ω = 1.2 .

a) Time series
b) Phase plane at steady state (120sec < t < 150sec)

6
Poincare' Map for Periodically Excited Forced Vibration
A Poincare’ Map consists of a sequence of points in state-space generated by the
penetration of a continuous evolution trajectory through a generalized plane in space. For
periodically excited dynamical systems, a Poincare’ Map can be obtained by
stroboscopically sampling the system state every forcing period (H. Poincare’ 1854-1912).

Poincare' Maps are used to characterize stability of periodic system response (e.g. wether
the response is periodic with the same period as that of the excitation).

The Poincare’ Map of a periodically excited dynamical system is constructed by sampling

system response at a fixed interval that equals the (minimal) period of the excitation.

A system with periodic response has a finite number of Poincare’ points in state space:
i) a single point: periodic response with the same period of excitation (T).
ii) m points: subharmonic response (of order m) corresponding to periodicity with an
integer multiple of the exciting period (e.g. mT, m>1).
iii) m points: ultrasubharmonic response (of order m/n) corresponding to periodicity with
an integer submultiple of the exciting period (e.g. mT/n, m,n>1).

A system which exhibits aperiodic response has an infinite number of Poincare’ points in
its state space:
i) an infinite number of points organized along an invariant curve (or topology):
quasiperiodic response describing a beating like phenomena.
ii) an infinite number of points that are not! organized along an invariant curve describe
nonstationary response (and are a candidate for chaotic vibrations in deterministic
systems).

Example:
Solve the system of two first order equations (equation 2), with γ = 0.2, 0.28, 0.51, 0.46 ,
α = −1, β = 1, δ = 0.3, ω = 1.2 :

7
 Poincare Map Procedure In Matlab
% File " poincare.m" , Poincare sampling procedure
clear all; close all; clc; format long;
%%%%%%%% initial conditions %%%%%%%%
t0=0;
tf=1000;
x0=[0.5 0]' ;
%%%%%%%%%%%%%%%%%%%%%%%%%
w=1.2; % input frequency
Te=2*pi/w; % cycle time
RES=1000; % resolution - number of sample points per cycle
Tspan=[t0:(Te/RES):tf];
option=odeset ('RelTol', 1e-8, 'AbsTol',1e-8); % option argument can be omitted
[t,X]=ode45 (@system1, Tspan, x0, option);
m=length(t);

%%%%%% POINCARE SAMPLING %%%%%%%%%%%%

per=0.5; % beginning of steady state
poincare_x=X(round(m*per):RES:m , 1) ;
poincare_y=X(round(m*per):RES:m , 2) ;
figure (1);
subplot (3,1,1) , plot (t,X(:,1));
xlabel ('t');ylabel ('x1');grid;

subplot (3,1,2) , plot (X(round(m*per):m,1),X(round(m*per):m,2));

xlabel ('x1');ylabel ('x2');grid;
subplot (3,1,3) , plot (poincare_x, poincare_y,'+');
xlabel ('x1');ylabel ('x2');grid;

8
(3) Forced vibration response: γ = 0.2

Figure 3: Forced vibration response: γ = 0.2, α = −1, β = 1, δ = 0.3, ω = 1.2 .

a) Time series
b) Phase plane at steady state (120sec < t < 150sec)
c) Poincare’ Map at steady state (period 1)

(4) Forced vibration response: γ = 0.28

Figure 4: Forced vibration response: γ = 0.28, α = −1, β = 1, δ = 0.3, ω = 1.2 .

a) Time series
b) Phase plane at steady state (120sec < t < 150sec)
c) Poincare’ Map at steady state (period doubling)

9
(5) Forced vibration response: γ = 0.51

Figure 5: Forced vibration response: γ = 0.51, α = −1, β = 1, δ = 0.3, ω = 1.2 .

a) Time series
b) Phase plane at steady state (120sec < t < 150sec)
c) Poincare’ Map at steady state (period tripling)

(6). Forced vibration response: γ = 0.46

Figure 6: Forced vibration response: γ = 0.46, α = −1, β = 1, δ = 0.3, ω = 1.2 .

a) Time series
b) Phase plane at steady state (5000sec < t < 10000sec)
c) Poincare’ Map at steady state (aperiodic with 760 points)

10