Lecture-5 - Nonlinear Systems - 1

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 17

5 LECTURE 5: NONLINEAR SYSTSEMS

At the end of this lecture a learner should be able to:


i) Differentiate between linear and nonlinear systems
ii) Explain the characteristics of nonlinear systems
iii) Characterize a non-linear system using the phase plane method

Concepts
Phase plane methods

5.1 Definitions
All systems are nonlinear systems since they have elements that are nonlinear in nature. Sometimes,
some nonlinear elements are deliberately introduced in control systems such as the ON-OFF operation
of a fridge. It is , therefore, necessary to understand how they affect the operation of a plant and how the
nonlinearities affect the performance of a plant.
Non-linear control systems are represented by plants and devices which are non linear and usually are
represented as differential equations. The following are examples of nonlinear equations:
x 2  y 2  2 xy  2 x  0

d 2x dx x 1  x 2 ,
 3x  2 x  0 Define x1=x then
dt 2
dt x 2  2 x1  3 x1 x 2
A non-linear equation has points of minimum, inflexion and maximum where the slope is zero, dy/dx=0.
5.1.1 Linear systems and Nonlinear Systems
Linear systems are characterized by linear differential equations with constant coefficients. These are
easily analyzed using classical and state space methods. Linear systems are characterized by a set of
properties:
i) Linearity and superposition:
The superposition principle, states that, for all linear systems, the net response caused by two or
more stimuli is the sum of the responses that would have been caused by each stimulus individually.
y  f ( x, u )
f ( a * x )  a * f ( x, u ); a a scalar
f ( a1 x1  a 2 x2 )  a1 * f ( x1 , u1 )  a 2 * f ( x2 , u2 )
A function, f(x),  that satisfies the superposition principle is called a linear function. 

ii) Existence and uniqueness of solution to differential equations: Equation and initial conditions
dx
 3x 2 ; x(0)  0
dt
In practice all systems exhibit some form of nonlinearity caused by such effects as hysteresis and
saturation.

Linear Systems Nonlinear Systems

1
Linear equation forms a straight line: y=mx+c, y , Equation forms a curve: ax2+by2+xy=c or
x are variables, a, b,c are constants differential equations

Is an equation with a maximum of one degree Equation with a degree of two or more

Superposition principle is applicable to a system Superposition principle does not apply to the
characterized by a linear equation. systems characterized by non-linear equations.

Nonlinear systems
i) Linearity
Non-linear system refers to the type of system where the output from the system does not vary
directly with respect to input to the system. Therefore the principle of linearity and superposition
does not apply to them. Sometimes it may be necessary to introduce nonlinearities deliberately
to improve the performance of a system and make its operation safer.
ii) Analysis:
Non-linear systems cannot be analyzed using classical or state space methods. However, there
are some specific methods for analyzing some specific types of nonlinearities.
iii) Stability:
Linear systems stability depend on the location of poles defined by the characteristic equation.
The stability of a nonlinear system depends on:
o Physical properties of the system\
o Magnitude of input
o Initial conditions

5.1.2 Characteristics of nonlinear systems:


Some of the nonlinearities that characterize nonlinear systems are:
i) Dead band
In a dead zone, the output is zero for small values of input less than a certain value, after which increase
in input the output is proportional to the input

i) Relay dead band ii) Motor dead band


Fig. 5. 1 a) Dead band,
2
ii) Limit cycles:
A limit cycle is an isolated closed trajectory that sustains oscillation of fixed amplitude and
frequency; the neighboring trajectories may spiral either towards or away from the limit cycle.
Thus, limit cycles can only occur in nonlinear systems.
X’

Fig. 5.1 b) Limit cycles


The limit cycle may be a decreasing or increasing amplitude, but reaches a point where there are
sustained oscillations at a given frequency.
iii) Hysteresis
Is a phenomenon associated with magnetization curves, depending on whether the magnetization force,
H, is increasing or decreasing.

Fig. 5.1 c) Hysteresis loop in Magnetization curve


iv) Saturation
When one increases the input to a physical device, the following phenomenon is often observed: when the input is
small, its increase leads to a corresponding (often proportional) increase of output: but when the input reaches a
certain level, its further increase does produce little or no increase of the output. The output simply stays around its
maximum value. The device is said to be saturation when this happen. A typical saturation nonlinearity is
represented in Figure 2, where the thick line is the real nonlinearity and the thin line is an idealized saturation
nonlinearity.

3
Fig. 5.1 e) Saturation nonlinearity

Most actuators display saturation characteristics. For example, the output torque of a two-phase servo motor cannot
increase infinitely and tends to saturate, due to the properties of magnetic material.

v) Relay Nonlinearity

As a special case, one can obtain the describing function for the relay-type (on-off) nonlinearity shown in Figure 5.
w
on
M

x
0

-M
off
Fig. 5.1 f) Relay nonlinearity

vi) Backlash

Backlash often occurs in transmission systems. It is caused by the small gaps which exist in transmission
mechanism. In gear trains, there always exists small gaps between a pair of mating gears as shown in Figure 9.

output angle

C
B

slope 1
A
O
Fig. 7.1d) Backlash nonlinearity
-b b input angle
D
E

Fig. 5.1 g) Gear Backlash nonlinearity

The backlash occurs as result of the unavoidable errors in manufacturing and assembly. As a results of the gaps,
when the driving gear rotates a smaller angle than the gap b, the driven gear does not move at all, which
corresponds to the dead zone (OA segment); after contact has been established between the two gears, the driven
gear follows the rotation of the driving gear in a liner fashion (AB segment). When the driving gear rotates in the
reverse direction by a distance of 2b, the driven gear again does not move, corresponding the BC segment.

After the contact between the two gears is re-established, the driven gear follows the rotation of the driving gear in
the reverse direction (CD segment). Therefore, if the driving gear is in periodic motion, the driven gear will move in
the fashion represented by the closed loop EBCD. Note that the height of B, C, D, E in the figure depends on the
amplitude of the input sinusoidal.

vii) Jump Resonance


In jump resonance, the output jumps in magnitude and phase as the frequency is changed near
resonance. Jump resonance causes a nonlinear system to behave differently with different inputs in
some closed loop systems with saturation. Typical response is shown in Fig. 7.1 b).

4
Fig. 5.1h) Jump Resonance phenomenon
viii) Subharmonic generation
Harmonics are frequencies that are multiple of the fundamental frequency. The power line has a
fundamental frequency of 60 Hz. The examples of harmonics in power system are frequencies of 120,
180 and 240 Hz.
Subharmonics are harmonics having frequencies less than the fundamental frequency. When a series
capacitor is used in a transmission line, sub-harmonic currents are created. Power
systems can exhibit subharmonic currents under some conditions. Subharmonics can create induction
generator effects, torsional interactions, torque amplification, sub synchronous resonance, and
transformer saturation. In particular, the third harmonic causes a sharp increase in the current in the
neutral conductor.
The main effects of voltage and current harmonics in a power system are usually:
i) The potential amplification of some harmonics due to parallel or series resonance*. This
amplification leads to higher energy consumption.
ii) Reduced performance of energy generation, transport and usage systems.
iii) The premature ageing of insulation on grid components, leading to energy reduction.
iv) Poor functioning of the system or any of its components.

ix) Soft and hard self-excitation


The system stability depends on excitation and initial conditions. If gain for large inputs is
sufficiently high, the overall system is unstable but stable for small input. If system is unstable for small
inputs or disturbance, the resulting excitation is “soft”. “Hard” excitation must be caused by signals or
disturbance above a certain minimum amplitude as in the case of a dead zone.

5.1.3 Why study nonlinear systems?


Most real systems are nonlinear, however the dynamics of linear systems are not rich enough to
describe many commonly observed phenomena. A lot of techniques that are used for nonlinear systems
come from linear systems, because:
 Nonlinear systems can (sometime) be approximated by linear systems.
 Nonlinear systems can (sometime) be “transformed” into linear systems.
 The tools are generalized and extended.

5
5.2 Analysis methods for nonlinear systems:
In practice, all physical systems have some form of nonlinearity. Sometimes it may be desirable to
introduce a nonlinearity deliberately in order to improve the performance of a system or make its
operation safer. One of the simple system with an introduced nonlinearity is a relay controlled or
ON/OFF system for home heating system where an heater is turned ON when the temperature falls
below a certain specified value and OFF when the temperature exceeds another given value.
Conventional analysis methods such as the Nyquist stability criterion cannot be used to analyze these
nonlinear systems. The two methods used analyze nonlinear systems are:
i) Describing function method in control system.
ii) Phase plane method in control system
iii) Linearization
iv) Lyapunov method

5.2.1 Describing Function


Harmonics
Nonlinear systems have nonlinearities that give rise to harmonics. Harmonics are multiples of the
fundamental frequency. For example a square wave has a fundamental frequency with large amplitude
and other harmonics as shown in Fig. 5.3.

Fig. 5.6 Harmonics


A complex waveform can be broken down mathematically into its individual components called the
fundamental frequency and a number of harmonic frequencies. In power systems
harmonics in power systems are generated by non-linear loads, such as transistors, IGBTs, MOSFETS
and diodes. Other nonlinear devices include office equipment such as computers and printers,
fluorescent lighting. In control systems, nonlinear loads include gears, motors and on-off switches.
Harmonics are higher frequency waveforms superimposed onto the fundamental frequency which may
cause distortion or nonlinearity in a function. The amount of distortion applied to the fundamental
wave will depend entirely on the type, quantity and shape of the harmonics present.
In nonlinear systems, the presence of harmonics may cause instability. Therefore a nonlinear element
has to be linearized to remove the unwanted signals.
Fourier Series

6
The Fourier series is described mathematically by letting x(t),t=nT, be a periodic function with period T
given by:
x (t )  x (t  nT ) ,
n , is an integer and T is the sampling period.
Then, x(t) can be written as a Fourier series:

a0 nt nt
x (t )    a n cos(2 )  bn sin(2 )
2 n 1 T T
where an and bn are the coefficients of the Fourier series:
T
2 nt
a n   x (t ) cos(2 )dt
T 0 T
T
2 nt
bn 
T  x(t ) sin(2
0
T
)dt

Note that for a function with period TT, the frequencies of the sines and cosines are 1T, 2T, 3T, are
multiples of the fundamental frequency, 1T. In approximation, only the first order terms are used. The
frequency, f=1/T, and the frequency, n/T, is called the nth harmonic.  This shows how the harmonics are
related to the Fourier series.

Describing function
Describing Function (DF) is used to analyze the existence of limit cycles in nonlinear systems based in
the frequency-domain approach. It is a form of linearization in the frequency domain. DF is used to
determine the existence of limit cycles, the amplitude and at what frequency. It is given as a function
of frequency and amplitude, N(A,ω).
DF analyzes the steady-state oscillations in non-linear systems to estimate the limit cycle parameters. It
assumes that:
i) There is only one single nonlinear component; e.g. saturation and dead zone.
ii) The nonlinear component is not dynamical and time invariant
iii) The linear component has low-pass filter properties
iv) The nonlinear characteristic is symmetric with respect to the origin
The system may be represented as a lamped parameters as a nonlinear element, NL, and the linear
plant, G(jω), with unity feedback as shown in Fig. 5.7

Fig. 5.7 Describing Function representation


DF gives the complex fundamental-harmonic gain of a nonlinearity in the presence of a driving
sinusoidal signal, represented by its Fourier representation:
a0 
y (t )  y (t  nT )    (a k cos t  bk sin t )
2 k 1
where the Fourier coefficients, ak and bk are to be determined.

7
The frequency domain analysis cannot be directly applied to nonlinear systems since frequency
response functions may not be defined for nonlinear systems. Therefore DF is a form of linearization in
the frequency domain.
5.2.2 DF for Single Relay
The output is estimation for a relay function, for sinusoidal input and clipped output, is shown in Fig.
5.8. The output is a square or rectangular waveform which consist of harmonics of the fundamental
input sinewave. The output higher order harmonics are filtered out and is normally represented by the
first harmonic. This is a form of linearization in the frequency domain.

u(t) y(t)

Fig. 5.8 Describing Function For single Relay function


The approximation of the output, y(t), using Fourier analysis is done with a0=0 and taking only
the first harmonic, n = 1. Then the Fourier series of transfer signal ratio in time domain is:
y (t ) a1 cos t  b1 sin t 2
 ,   2f 
u (t ) A sin t T
Consequently, most of the information of u(t) is contained in the first harmonic term. Therefore
the linear element must have low-pass properties.
Using the trigonometric identity, a cos t  b sin t  a 2  b 2 (sin t   ),   tan 1 (b / a ) , then

y (t ) a12  b12 sin(t   )



u(t ) A sin t
The describing function, N(A,ω) is defined as the ratio of the phasor representation of output
component at frequency, ω. This may be written as:
a12  b12 * e  j (t  ) 1
N ( A,  )   (b1  ja1 )
Ae  jt A
Where a1 and b1 are the first harmonic Fourier coefficients:
2 2
 
 
a1 
2A 0 u(t ) costdt and b1  2 A  u(t ) sin tdt
0

And
2 2
 
 
N ( A,  ) 
A 0 u ( t ) sin tdt  j
A  u(t ) costdt
0

The output for the single relay , shown in Fig. 5.7 is:
 
  h, for  t  0

m( t )  

  h, 
 for 0  t 
 

And the DF is

8
4h
N ( A,  ) 
A
where A and ω are the amplitude and frequency of the output m(t) respectively.
The design is now carried out using the linearized non-linear function, N(A,ω) and the plant, G(jω).

5.3 Phase Plane


Given a function of time, x(t), the phase plane gives the plot of dx/dt versus x(t). The plot shows the
trajectories or path taken by a function around some point and whether the slope is increasing,
decreasing or zero around some points.
The purpose of the Phase plots is to qualitatively analyse the trajectory about an equilibrium point so
that information about stability may be obtained. However this method is limited to second order
systems.
Consider the system is represented as:
x  f ( x , x )  0

Is represented in matrix form as


x 1  x 2  f 1 ( x1 , x 2 )
x 2  f 2 ( x1 , x 2 )

For a second order system, the phase plane, gives the plot of x
 1 and x 2 .

Equilibrium and Singular Points


The points of analysis are the critical points, where, dx/dt=0. The critical points are also called
equilibrium points. Critical points correspond to points in a graph where there is a maximum, minimum
or a point of inflexion point. An equilibrium point is where a system can stay for ever since there is no
accelaration and corresponds, for a second order system, to:
x 1  x 2  f 1 ( x1 , x 2 )  0
x 2  f 2 ( x1 , x 2 )  0

An equilbrium point is also called a singular point is in the phase plane where the time derivative is
zero. It may appear as a point where trajectories meet at or diverge from it. A singular point may also
be a point where the trajectory approaches the point or the value becomes infinite.
Example
Determine the singular points of the equation:
x 1  x 2  0  x2  0
x  0.5 x  3x  x 2 :
x 2  0.5 x 2  3x1  x  0 2
1  x 1 ( x1  3)  0; x 1  0,3
Checking the slope dx2/dx1=+3+2x about (0,0) and (-3,0). By putting values of the points, trajectories
move towards from(0,0) and move away from (-3,0).
Consider the second order differential equation:
ax  bx  cx  f

The the roots of the quadratic equation are: s1,2=-b/2 +sprt(-b 2-4ac). The position of the roots is an
equilibrium point: when b=0, there is only one singular point at the oring and
i) λ1 and λ2 with the same sign(positive or negative, λ 1* λ2 >0 ): - Node- stable or unstable

9
ii) λ1 and λ2 both real and opposite sign, λ1* λ2 <0 : - saddle point(maximum and minimum
point
iii) λ1 and λ2 both complex conjugate with non-zero real parts, Re(λ 1)=Re(λ2) ǂ0: - focus –
stable
iv) λ1 and λ2 both complex conjugate with zero real parts, Re(λ 1)= Re(λ2) =0: - centre
(harmonic oscillator)

Fig. 5.9 Phase plots of different types of equilibrium points


Consider a pendulum without air resistance and constant aplitude would be represented by a circle or
ellipse. If the amplitude is changed, the path becomes a circle with a smaller amplitude as shown in Fig,.
5.6. However if there is air resistance, the amplirude decreases until it becomes zero.

Fig. 5.10 Phase plots or trajectories for y=dx/dt and x


Although many systems are reprented higher than second order differential equations, they can be
approximated to a second order system, by considering the pair of dominant poles. When solutions of
high accuracy are not required in nonlinear sysatems, a graphical method can be used to plot the

10
trajectories as the solution to a differential equation of low order, called the phase plane methods.
The trajectory about an equilibrium point may be used in design.
Phase plane plots are subject to the following limitations:
i) They are applicable to second order systems only
ii) The driving function, u(t) can be represented by initial conditions
The response of second order systems can be described by the variable , x, and its derivative, dx/dt. The
differential equation can be reduced to a form where time is a parameter. The system is plotted on a
phase plane, the vertical axis representing velocity, and the horizonal axis representingthe
displacement, x. The path which represent the motion is x’ versu x is called the phase trajectory.
Consider the system model:
x  ax  bx  f

dx
Denoting x1  x and  y  x 2 we can rewrite:
dt
dy
 x   ax  bx  f
dt
 -ay - bx  f
The slope near a point helps to map whether the slope near a point is increasing , zero or decreasing.
This information can be used to plot the trajectories and show regions where the slope is increasing,
zero or decreasing. The general form to denote the slope near a point as:

dy
dy dy dt  ay  bx  f N ( x, y )
 .  dt  
dx dt dx dx y D ( x, y )
dt
By using this form, it is possible to study the relative variations of y and x. The dy/dx represents the
slope on the phase plane and the trajectory taken by starting from a given point or initial condition.
Example
Consider the the system described by the first order system: y  x   x This is a first order equation
that represents the straight line , with a slope of -1, as shown in Fig. 10.1. The poles of the system is at
dx
( s  1)  0, s  1 and the x (t )  e 1t ; y   1e 1t  1x (t ) and are stable trajectories and
dt
decays and goes to zero as time increases, t→∞ ,with a slope of -1. When x is negative, the slope is
positive. All trajectories go towards the origin, the equilibrium point. Equilibrium point is defined as the
point where dx/dt=0. The critical point at this point is x’=0, which is at the origin. Critical points in
agraph correspond to maximum, maximum or inflexion point. The trajectories may be represented as
shown in Fig.5.8

Fig. 5.10 Phase plots or trajectories for y and x

11
5.3.1 Sketching Phase trajectories
Sketching phase trajectories is the basis of phase plane analysis. There are two methods of plotting the
phase portraits:a
i) Analytical method: Obtains a relationship between y=x’ and x by solving
the differential equation and and then construct the
ii) Graphical method: is used to consstruct the plots indirectly.

5.3.2 Analytical method


Consider the harmonioc oscilator system represented by the model:
x   2 x  0
then
dx dy
y ; x    2 x
dt dt
and
dy dy dx dx
 .  . x   2 x
dt dx dt dx
Solving
xf xf

x dx   xdx :  x dx    xdx


2 2

x0 x0

Solving by variable separation gives

x 2 x 02
x2   A2 ; A  x02 
 2
2
Which is the equation of an ellipse as shown below. It may also be solved using Lapace transform as:
( s 2   2 ) X ( s)  0 ;
and
s=±jω, x(t) = Ae-ωt

Fig. 5.11 Trajectory of a harmonic oscillator.


The trajectories depend on the nature of the stability around the critical or equilibrium points.
5.3.3 Graphical method:
There are several methods of graphically plotting the phase potraits. The method of isoclines is widely
used and easy to sketch the plots.
12
The method of iscoclines sketches the plost from the equation:

dx
dx  f ( x, x ) N ( x, x )
 dx  
dx dx x D( x, x )
dt
dx  f ( x, x )
The slope of the trajeectory passing through the point is given by  .
dx x
 f ( x, x )
An isocline with the slope, α, is given by   and the points on the issocline have the same
x
slope.
By taking different values of α, the different isoclines can be drawn in the phase plane. The phase
potraits can be drawn by sketching the continuous lines along the field of directions.
Example 1.
Consider the system modeled by:
dx

x   x  x : x
  ( x  x )
dx
Let the slope be α, then
x
x (1   )   x :  
x ...........................( 5)
(1   )
Equation (5) is a straight line with a slope of -1/(1+α). We can now obtain the different isoclines for
different values of α. A table can be drawn with α and the corresponding angle for each point. For this
example the trajcetories are shonw in Figure below.

Fig. 5.12 Phase potrait.

By looking at the location of poles, gives the trajectory whether it is stable or unstable: In this case,
the poles are complex conjugate on the LHS of the s-plane and are therefore stable.
1
( s 2  s  1) X (( s )  0; s1,2  0.5  j 3
2
5.3.4 Singular points
The singular points of a nonlinear differential equation are important in determining the properties of
its solution. In the phase plane method, the fundamental equation is of the form:

13
dy N ( x, y )
 
dx D( x, y )
The singular points of this equation are values of x and y, for which N(x,y)=D(x,y)=0. If N anmd D are
nonlinear functions, there may be a number of singular points. A singular point is always a point of
equilibrium and are also called critical points, since both x  y  0 . The resultring equilbrium may be
stable or unstable.The nature of the solution near the singular points is governed by its characteristic
roots or eigenvalues, λ1 and λ2 for a 2D phase plane. These are characterized as:
i) Node
Unstable node: positive real roots, +λ1, +λ2,
Stable node: negative real roots, -λ1, -λ2,
ii) Saddle points: real roots with real but opposite in sign, -λ1, +λ2,
iii) Centre, Vortex point: closed curves, poles at the imaginary axis, λ 1,2=±jω
iv) Focus
Stable focus: complex conjugate poles, negative, λ1,2 = -σ±jω,
Unstable focus: complex conjugate poles, both positive λ1,2= +σ±jω,

Node: This is a point in the phase plane consisting of a set of trajectories which directly converge as
they approach the point (stable node) or diverge as they move away from the point (unstable
node).

Focus: This is a point in the phase plane consisting of a family of spiral trajectories, which either
converge (stable) or diverge (unstable) from it.

Center: A point in the phase plane consisting of closed curves enclosing it. Occurs when roots are
complex conjugate on the jw-axis.

Saddle Point: This is a point where trajectories around it are such that some approach the point and
others diverge from it. It corresponds to roots where one is on the LHS of the s-plane while
the other is on the RHS.

i) Proper Node, ; λ=-σ1, -σ2 ii) Degenerate Node, λ=+σ1, +σ2

14
iii) Improper node focus, ; λ=-σ±jω iv) stable focus point; λ=-σ±jω

v) Centre, vortex, λ= ±jω vi) Saddle point, , λ=-σ1, +σ2


Fig. 5.13 trajectories about the Critical points
The various types of equilibrium points are shown in table 10.1.
Table 5.1 Determining the types of points:

5.3.5 Trace-Determinant diagram and the Critical Points


a b
x  Ax   x; Det A  ad - bc
c d 

15
The characteristic equation is given by
 a b
I  A   (   a )(  d )  bc  0
c  d
2  ( a  d )  ( ad  bc )  0
2  T  D  0 : p  T  trace, q  D    determinat

This is a quadratic equation λ2 − pλ + q = 0 with coefficients p, q and discriminant Δ :


2  p  q  0
1 1
1  ( p   ); 2  ( p   );   p 2  4 q;   T 2  4 D
2 2
This may be represented in graphical form by plotting q Vs p as shown in Fig. 5.12 below

Fig. 5.14 trajectories about the Critical points


Trace-determinant diagram
The trace-determinant diagram divides the plane into regions corresponding to different types of equilibrium
point.

16
Fig 5.16

17

You might also like