Exsheet 1

Engineering Tripos Part IIB
Module 4F3: Nonlinear and Predictive Control

Examples Paper 4F3/1
1. Suppose two nonlinear dynamic systems are coupled together, as shown in fig. 1,
and that their defining equations are:
d3 y(t)
System A: = g[y(t), ẏ(t), ÿ(t), u(t), t]
dt3
System B: ü = h[u(t), u̇(t), y(t), t]
where g and h are nonlinear functions. Show that they can be represented by a
single vector differential equation
ẋ(t) = f [x(t), t].
What is the dimension of x?
u y
Figure 1: Feedback interconnection
2. Show that the difference equation
yk = h[yk−1 , yk−2 , . . . , yk−n , uk , k],
where h is some nonlinear function, can always be written in the standard form
xk+1 = f (xk , uk , k).
Do the reduction for the system
yk = yk−1 yk−2 + k sin(yk−1 uk ).
1
3. A phase-locked loop (used in communications networks) is described by the equation
ÿ(t) + [a + b cos y(t)]ẏ + c sin y(t) = 0
where a > b ≥ 0 and c > 0. Express this in state-space form. Find the equilibrium
points and examine their stability.
4. By considering the two curves y = ax(1 − x) and y = x explain diagrammatically
how the logistic map
xk+1 = axk (1 − xk )
has one stable equilibrium for a = 2 and tends to an oscillation between the state
x = 0.48 and x = 0.83 for a = 3.3.
5. The Verhulst system
xk+1 = xk + r(1 − xk )xk
becomes chaotic for most values of r above 2.6, approximately. But there are values
of r in this range for which only a small number of stable periods exist. Using a
computer or calculator, show that r = 2.83 gives stable period-3 points near 0.2113,
0.6830 and 1.2957, but that for r = 2.845 each of these points has undergone ‘period
doubling’. Verify that for r = 2.825 and r = 2.86 the system exhibits chaotic
behaviour.
(Matlab is recommended. The following commands
xx=[]; x = 0.5; r = 2.83;

for k = 1:50,
xx = [xx,x];
x = x +r*(1-x)*x;
end
xx(40:50)
plot(xx)
would do the iterations for r = 2.83. The period-3 oscillation takes about 20 itera-
tions to become established with x0 = 0.5; so if you use a calculator you should take
x0 = 0.2113, or one of the other stable points, to avoid a lot of button-punching.)
6. In each of the cases below show that the general solution of the equation ẋ = Ax is
as given: (Hint: reduce the matrix equation to appropriate scalar equation(s).)
! ! !
0 0 1 0
(a) A = , x(t) = c1 + c2 ,
0 0 0 1
! ! !
0 1 1 t
(b) A = , x(t) = c1 + c2 ,
0 0 0 1
(where c1 and c2 are arbitrary constants.) Characterise the stability of the equilib-
rium in each case. Find the eigenvalues of A. Comment.
2
7. The Rössler system has been used to model a variety of systems to demonstrate
evolution from well-behaved limit cycles to chaotic solutions, e.g. dripping water as
flow rate is increased, onset of cardiac fibrillation. The system is described by
ẋ1 = x1 − x1 x2 − x3
ẋ2 = x21 − ax2
ẋ3 = bx1 − cx3 .
Find the equilibrium solutions and investigate their stability as a function of the
parameters. (Note: The Routh-Hurwitz Stability Criteria given in the Control
section of the Electrical and Information Data Book will be useful for part of this
question.)
8. Consider the following piecewise-linear system representing a non-linear PD-controller

for an inertial mass. The torque input to the inertial mass is provided through a
relay with dead-zone: u = ±U if |z| > 1, u = 0 otherwise. Let r = 0, kp = 2,
kv = 1 and U = 4J. Obtain a state equation description of the system and find the
equilibrium states. (Hint: You need a different equation for each condition of the
relay.)
Sketch the state-plane trajectories. Notice that, strictly speaking, the vector field for
the overall system is discontinuous. Determine the points at which discontinuities
occur (“switching surfaces”). Do solutions exist starting at these points? If not,
speculate about the possible behaviours of the physical system if it finds itself on
one of these switching surfaces.
Relay with Plant

dead zone
r + e + z u ẏ y
1 1
- m- kp - m- - Js
- s
-
−6 −6
kv
Figure 2: Nonlinear PD controller
3
9. By considering dtd (x2 + y 2 ), or otherwise, show that the unit circle is a stable limit
cycle of the system
x
ẋ = y + √ [1 − (x2 + y 2 )]
x2
+ y2
y
ẏ = −x + √ 2 [1 − (x2 + y 2 )].
x + y2
10. Consider the phase-locked loop of question 3.
(i) By writing the system in the standard state equation form with x1 = y and
x2 = ẏ and using the function
1
V (x1 , x2 ) = c(1 − cos x1 ) + x22
2
where c > 0, show that the origin is a stable equilibrium point.
(ii) Use LaSalle’s theorem to show that the origin is asymptotically stable.
(iii) Sketch the function V (x1 , x2 ) and deduce the stability of each equilibrium.
Answers
3. Equilibrium points: y = nπ for n an integer. Stable equilibria: n is even or zero.

√ √
7. Equilibrium points: origin and, if d = a(1 − b/c) > 0, (± d, 1 − b/c, ±b d/c). At
the origin stability requires a > 0, b > c > 1. Under the assumption that d > 0,
stability at other equilibria requires:
c > 0
a + c > b/c
!
b 3b
a (a + c − )(c + 2 − ) − 2(c − b) > 0
c c
Past papers
The following past Tripos questions are suitable for further practice:
Part IIB Module 4F3: 2009, Q. 1. 2008, Q. 2. 2007, Q. 2. 2006, Q. 4. 2005, Q. 1. 2004,
Q. 1. 2003, Q. 2.
Part IIB Module I3: 2002, Q. 2, 3(b). 2001, Q. 1. 2000, Q. 1(a) and (b). 1999, Q. 1.
1998, Q. 1,4(a). 1997, Q. 1. 1996 Q. 1,2.
I. Lestas, January 2009

(J.M. Maciejowski, January 2008)

Exsheet 1

Uploaded by

Copyright:

Available Formats

Exsheet 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Exsheet 1

Uploaded by

Copyright:

Available Formats

Engineering Tripos Part IIB

Module 4F3: Nonlinear and Predictive Control

ẋ(t) = f [x(t), t].

What is the dimension of x?

Figure 1: Feedback interconnection

2. Show that the difference equation

yk = h[yk−1 , yk−2 , . . . , yk−n , uk , k],

xk+1 = f (xk , uk , k).

Do the reduction for the system

yk = yk−1 yk−2 + k sin(yk−1 uk ).

ÿ(t) + [a + b cos y(t)]ẏ + c sin y(t) = 0

xk+1 = xk + r(1 − xk )xk

xx=[]; x = 0.5; r = 2.83;

8. Consider the following piecewise-linear system representing a non-linear PD-controller

Relay with Plant

Figure 2: Nonlinear PD controller

10. Consider the phase-locked loop of question 3.

3. Equilibrium points: y = nπ for n an integer. Stable equilibria: n is even or zero.

I. Lestas, January 2009

You might also like