Nonlinear Systems:

While linear control systems are effective and widely used,

many real-world systems exhibit nonlinear behavior.
This nonlinearity can arise from various sources, such as:

Saturation Effects: Actuators and sensors can saturate

Friction and Hysteresis: Mechanical systems often experience friction that varies with speed,
Hysteresis effects can also introduce nonlinearity

Variable System Parameters: The parameters of a system may change depending on operating conditions

Designing controllers for the nonlinear system using linear system theory has adverse effects as

- demands high quality components and hence higher cost

- design will be limited to linear operating range

Linear system obey the principle of superposition.

- stability is internal to the system and does not depend on external input

For nonlinear systems

- Response is sensitive to the magnitute of the input
- Laplace and z-Transform are not applicable because the system does not obey the principle of superposition
- Stability of the system depends on the input and also the initial state
- May exhibit limit cycles

F Cos wt
Consider a spring mass damper system
Friction coefficient f
system equation with components assumed linear

Let us now consider the spring as nonlinear

(a) (b)

Frequency response for linear dynamics Frequency response for non-linear dynamics

(a) Hard spring (b) Soft Spring

- Jumps in magnitude and phase near resonance

- Response function is double valued for some range of frequency
- This phenominon is called jump resonance
 If the forcing function is removed the response of the system

Linear Nonlinear system

- damped sinusoidal oscillations - damped oscillations

- frequncy of oscillations remain same - frequncy of oscillations increases with decrese in amplitude

- frequncy of oscillations decreases with decrese in amplitude

For nonlinear system the response (frequency) is

sensitive of amplitude

Frequency response for nonlinear system may not be defined

It will have harmonics or may sub harmonics in the response

Limit Cycle:
The nonlinear system enen when staying within within its tolerance limits, may exhibit a closed trajectory or limit cycle.

The limit cycle describes the oscillation of nonlinear systems. The existance of limit cycle corresponds to an ocillation of

fixed amplitude and period, which may not be sinusoidal.

Consider nonlinear oscillator Vander Pol's Oscillator

with the choice of phase varibales as state variable dynamcis becomes

The characterisitc equation becomes

Damping coefficient is

i.e. overdamped

system is unstable and hence x increased and hence damping factor becomes
+ve and thus system enters limit cycle.
 No tool like z or s - transform exists to analyze the performance of nonlinear system.

- Linearization in the operating range

f ′ (x0 ) f ”(x0 )
f (x) ≈ f (x0 ) + (x − x0 ) + (x − x0 )2 + · · ·
1! 2!

This linearization may not be possible for some nonlinear systems, like dry friction

- Piecewise Linearization:
Different linear differential equation model
in different regions

End condition of one model will be initial

condition of the other model

Nonlinear differential equations are solved

this way numerically

- Phase plane method:

restricted to second order system

- Describing function methos:

Based on harmonic linearization,

Because of the filtering property of the linear part the response is represented by the fundamental frequency
 Common Physical Nonlinearities:

Nonlinearities can be classified as incidental and intentional

incidental: these are inherently present in the system e.g. Saturation, dead-zone, coulomb friction,

stiction, backlash

intentional: These are deliberately inserted in the system to modify the system characteristics

e.g. relay

Saturation: In the limited range i/p - o/p ralation is linear,

However as the input exceeds the limit o/p saturates

e.g. Amplifiers

Friction:

Frictional forces exist between two sliding surfaces.

Stiction is the force required to initiate motion.

Stiction is greater than the column friction.

Backlash:

It is the play between the teeth of drive gear

and driven gear.

For a given i/p, o/p is multivalued

(Double Valued Nonlinearity) O/p depends on the

history of the input.

- In servo system backlash causes oscillations/chattering

- unstable for large backlash

Dead Zone

Occurs in system which are insensitive to small inputs

e.g. Gear with backlash drive the torsional spring load

Relay:

May be called as nonlinear power amplifier

Dead zone - relay coil requires finite amout of current to actuate the relay

Hysteresis - Large coil current is needed to close the relay than the current at which the relay drops out.

Methods for the analysis of Nonlinear Systems:

a. Phase Plane Analysis

b. Describing Function Method

c. Liapunov's Stability Analysis

The Phase Plane Method:

Consider the unforced spring, mass, damper system

in the standard form

Let the state variables be

 The solution for the state vector can be obtained analytically and plotted for different intial conditions.

But to obtain analytical solution is difficult for the nonlinear system like when spring is nonlinear

For nonlinear spring the dynamics is

Phase portrait for the undamped system with linear spring

ζ=0

Method of Constructing Trajectories:

- Analytically - Graphically

Method of Isocline (Graphical Method)

Consider a second order autonomous system

The slope m of the trajectory at any point in the phase plane is given as

Every point (x1, x2) of the phase plane, has with it the associated slope of trajectory except at Singular Point.

Isocline are the lines in the phase plane corresponding to slope of the phase portrait.

The direction of phase trajectory at any pont (x1, x2) can be obtained from the sign of

For a specific trajectory slope m1 we have

This equation gives the locus of all such points in the phase-plane of which the slope of the phase trajectory is m1.

Such a locus is called a Isocline.

 Undamped System:

Intigrating

Singular Point or Equilibrium Point:

Phase plane portraint of an autonomous system is a family of non-crossing trajectories which describe the response of the system
to all possible initial conditions.

The singular point or the equilibrium point is the point where the derivatives of the state variables is zero.

Consider the dynamics of the system of second order ẍ + a(x, ẋ)ẋ + b(x, ẋ)x = 0

Let us choose the state variables as x = x1

ẋ1 = x2 ≡ f1 (x1 , x2 ) ... (A)
ẋ2 = −bx1 − ax2 ≡ f2 (x1 , x2 )

Now origin is the equilibrium point for this system as the system is autonomous. Ẋ = f (X)

To find equilibrium pont we will evaluate Ẋ = 0 which gives origin as equilibrim point.

Using taylor series (A) can be written as

The system equations can be approximated by the linear terms only, provided they are dominant in the neighbourhood of the origin.
 Thus the approximated dynamics aroung the origin can be written as

ẋ1 = a1 x1 + b1 x2
ẋ2 = a2 x1 + b2 x2
The characteristics of the singular point can be studied using resultant characteristic equation.

Thus using Laplace trnasform and eliminating one of the varibale gives

and the characteristic equation of the system linearized about the orgiin becomes
rewritten as ... (B)

Now, if the roots of the characteristic equation have negative real part, all trajectories near the origin will approach the singular point.

However, if one root is zero, then the nature of trajectory depends on higher order terms.

The nature of solution of (B) depends on the location of two roots, λ1 , λ2

1. λ1 , λ2 are complex conjugate having negative real parts. The system is underdamped 0 < ζ < 1

Stable Focus: The phase trajectory is a logrithmic spiral

2. λ1 , λ2 are complex conjugate having positive real parts.

Ustable Focus: The phase trajectory is a logrithmic spiral expandint out.

Using transformation Z = N Y

where

Now the transformed system becomes

... (1)
 Now,

Let us define

So we get (Or)

This is the equation of the logrithmic spiral.

For σ<0 it is stable focus

For σ>0 it is unstable focus.

3. λ1 , λ2 are real and lie in the left half of s-plane.

The system is overdamped ζ > 1

Stable Node: the time response for x1 and x2 is monotonically decreasing

4. λ1 , λ2 are real, unequal and lie in the right s-plane

Unstable Node:
5. System with repeated roots:

System with critical damping ζ = 1

The response monotonically decreases to singular point.

The two e.v. coalesce into a single e.v. with slope determined
by the roots.

6. Real part of complex conjugate roots is zero:

Roots lie on the imaginary axis i.e. ζ = 0

Response is sustained oscillations From (1) above with σ = 0

7. System with one +ve real root and other -ve real root: Saddle Point

The e.v. due to the -ve root provide a trajectory that enters the singular point.

The trajectory due to +ve root leave the singular point.

 Phase portrait for a critically dmaped system

Phase plane method is restricted to second order system with constant parameters and constand and zero input.

Single tracjectory pass through a given point in phase plane except the singular point.

Singular Points:

Consider a general time invariant system described by the state equation

for an autonomous system

A point in the state space where derivative of the state variable becomes zero is called a singular point or

an equilibrium point.

The linearized model of the system

may be written as

The equilibrium point for this system

may be found as

Thus origin is the equilibrium point if A is nonsingular matrix.

If is the equilibrium point, the origin of the coordinate frame may be shifted to as

system dynamics in terms of

new phase variable
 Consider the linearized second order system

To examine the phase plane trajectory apply the transformation

such that the dynamics becomes

To find the trajectory traced by the representative point P

the velocity vector at P is shown in the figure

change of coordinate axes by
linear transformation

The slope of the velocity vector is given as

which upon integration gives where c is the constant of integration

The plot of this equation in z1-z2 plane gives the phase trajectory.

Nodal Point

Let us consider a linear second order system

with real, distinct and -ve eigenvalues

equilibrium point of this type is a stable node.

 Saddle Point:

Consider the e.v. as in figure

The origin which is the equilibrium

point is the saddle point

Phase portrait

The trajectories are the equilateral hyperbolas

Focus Point:

Consider the system with eighen values

The state model for this system may be written as

Z = P Y
using the transformation the dynamics becomes

Let us define
 Stable focus for -ve real part of Stable focus for -ve real part of
complex [Link] complex [Link]

Centre or Vortex Point

Consider the [Link] with zero real part

This has the solution

c - depends on initial conditions

Example: A regulator problem

where

defining state variables

On linearizing about origin which is the equilibrium point

Now [Link] of A will decide the nature of

equilibrium point
 If [Link] are real and -ve, so equilibrium point will be

stable node.

[Link] are complex conjugate with -ve real part,

If
so equilibrium point will be stable focus.

Stability of Nonlinear Systems

For Linear system

i. Free System: A system with zero input and arbitrary initial condition is stable if its trajectory

tends towards the equilibrium point.

ii. Forced System: System is stable if for a bounded input the output is also bounded.

For linear system there is only one equilibrium point in the whole state space.

For nonlinear system, local stability (small deviation about equilibrium point) does not imply stability in t

whole state space.

Definition of stability for nonlinear system:

i. Stability ii. Asymptotic stability iii. Asymptotic stability in the large

i. Stability: Consider an autonomous system

with origin as the equilibrim point

Mathematical definition of stability:

Limit Cycles

Limit cycle describes the oscillations of the nonlinear system.

Limit cycle in nonlinear system corresponds to an oscillation of fixed amplitude and period.

Let us consider the Vander Pol's differential equation model

and compare it to the linear second order system

damping factor of nonlinear system is

when x is large system is overdamped and thus x decreases

and when x is small system has a negative damping and

thus x increases

Unstable limit cycle

 Linear System Nonlinear System

oscillatory behavior depends on does not depend on initial condition

inittal condition
not sensitive to parameter values
sensitive to parameters

Construction of phase trajectories

Analytical Method:

Consider a second order autonomous system

Dividing we get

Consider the dynamics

upper half of the state-plane moves from left to right

lower half of the state-plane moves from right to left

The trajectory cross the x1 axis at 90 degree

 with the choice of state variables as

the system state space model may be written as

output of the controller is given as

separating the variables and integrating we get

solving we get

Region-I (e positive, relay output +M)

Region-II (e negative, relay output -M)

The trajectories in these two regions is parabola, symetrical about x_1 axis

Vertex of parabolic tracjectory in Region-I is at

Vertex of parabolic tracjectory in Region-II is at

Shifting the singular point to origin the dynamics becomes | by keeping r = 0

with the choice of state variables as the dynamics becomes

The slope equation now

becomes
 Suppose initially the system is at A, the system will follow

the path B -> c -> D

If relay is switched at D, the representative point will then

follow the parabolic trajectory to origin.

This is the shortest time path to origin.

Now let us consider the system with PI controller

and the relay

Modified system is described as

with the choice of state variables as

the dynamics can be written as

The slope equation is given as

On integration we get the solution as

The state-space is divided into two regions by the line

Suppose initially the system is at A, the system will follow the path B -> c

There will be frequent switching across the line. This fast switching is called chattering.

For ideal second order system, the switching frequency will be infinite and amplitude will be zero.
Compensation has thus converted the oscillating system into an asymptotically stable one.

If KD is small, the slope of the line is nearly vertical, the chattering of the relay can be reduced

though at the expense of high settling time.

Describing function method

Phase plane method is applicable to second order systems.

For higher order systems, system needs to be approximated by second order dynamics

Describing function method is applicable to wider class of nonlinear systems to determine their behavior

Output of the nonlinearity will not be sinusoidal, which may be expressed using Fourier Series as

If nonlinearity is assumed to be symmetrical, then average value of y will be zero, thus

Because of the low pass filtering only the fundamental component of the output will be retained.

The fundamental component of the output is given as

under the assumption the nonlinearity can be replaced by a describing function

which embodies amplification and phase shift in the fundamental

component of the input x

- All linear theory frequency domain techniques now become applicable

- It is used for the stability analysis only

- no general correlation between time and frequecy response

Derivation of describing functions

Describing function of the nonlinear element is given by

The fundamental component of the output is given as

where A1 and B1 are the Fourier coefficients and may be calculated as

and further,

Dead Zone and Saturation:

Because of half wave symmetry B1 = 0 and

because of quarter wave symmetry

 A1
Therefore the describing function is given as

Case - I saturation nonlinearity

Case - II Dead zone nonlinearity

Stability analysis by Describing function analysis

The characteristic equation of the system can be written as

According to the Nyquist stability criterion, the system will exhibit sustained oscillations or limit cycle when
Frequency independent describing function

Stability of limit cycle:

At L1:

If the amplitude X of the i/p

to nonlinearity increases
The operating point shifts
to L1'
L2:
Since the closed loop
Stable
system is unstable in this
Limit Cycle
condition the amplitude of
the i/p increases further.
L1:
If the amplitude X of the i/p
Ustable
to nonlinearity decreases
Limit cycle
The operating point shifts
to L1"
Since the closed loop
system is stable in this
condition the amplitude of
the i/p decreases further.
and hence the system
deviates from L1
 Liapunov's Stability Criterion

The general state equation for a nonlinear system is given as

the analytical solution of this is generally not possible.

Numerical solution will fail to provide insite into the stability behavior.

Lyapunov's Results for stability

i. Linearization and local stability ii. Lyapunov's Direct Method

(Energy based method)

i. Linearization and local stability

It is concerned with the local stability of the nonlinear system.

It is a formalization of the intuition that a nonlinear system should behave similarly to its
linearized approximation for small range motions.

Consider the autonomous system and assume that is continuously differentiable, then

using Taylor series

h.o.t. => higher order terms

ignoring higher order terms, and retaining the jacobian of f at x = 0 only will learize the system at the

origin - the equilibrium point.

Let Then

is called the linear approximation of the original

nonlinear system at the equilibrium point 0.

Similarly for the non-autonomous system s.t.

linearization may be carried out as

ignoring the higher order terms,

linearizes the system as
Now the stability of the nonlinear system may be concluded in neighbourhood of the origin as follows:

Lyapunov's Direct Method:

If the total energy of the system is continuously dissipated, then the system, whether linear or nonlinear

must eventually settle down to an equilibrium point.

Consider an autonomus system described by

and let be a solution. Further let be the total energy associated with the system.

If the derivative is negative for all except the equilibrium point, then it follows

that energy of the system decreases at t increases and finally the system will reach the equilibrium point.

Consider the spring mass damper system with dynamics as

The state space model may be written as

At any instant, the total energy of the system is sum total of kinetic energy (K.E.) of the moving mass

and the potential energy (P.E.) stored in the spring.

Thus

(P.E.) + (K.E.)
 The rate of change of energy is given by

Thus is negative at all points except where

So the energy of the system will continue to decrease till

Even when we observe that

so for non zero x1; x2 can not be zero continuously

Hence origin (x1 = 0, x2 = 0) is the only equilibrium point, where system will come to rest.

Constant energy curves v1, v2 and v3 as a

function of states x1 and x2 are shown in Fig (a)

The trajectory of the system starting from

initial condition (x10, x20) and passing through

(b)
different energy levels v1, v2 and v3 at different

times from t=0 to t=infinity is shown in Fig (b)

Direct method of Lyapunov and the Linear Systems

Consider the linear autonomous system

Consider the scaler Lyapunov function

where is a symmetric positive definite matrix ( )

Note that
If matrix is +ve definite, is negative definite. This implies the system is asymptotically stable

in the large at the origin.

If matrix is -ve definite, is positive definite. This implies the system is unstable

However if does not come out to be +ve or -ve definite, then we have failed to conclude the stability

of the system using the specific choice of P, and we need to repeat the process with some different P which

is +ve definite.