Describing Function analysis-v1 - people.unica.it · Describing Function analysis of nonlinear...

Describing Function analysisDescribing Function analysisof of

nonlinear systemsnonlinear systems

University of CagliariUniversity of CagliariDept of Electrical and Electronic Eng.Dept of Electrical and Electronic Eng.

Prof. Elio USAIeusai@diee.unica.it

University of LeicesterDepartment of Engineering

Describing Function analysis of nonlinear systems – Prof Elio USAI – March 2008

System definition and problem statement

r: reference variablee: error signalu: control variablem: manipulated inputy: controlled variable

ym: measure of the outputz: feedback signald: disturbancen: measurement noise

Aactuator

Pprocess

Ccontroller

Ttrasmitter

++r(t) e(t) u(t) m(t)

Ffilter

Controller

In many cases the system presents a nonlinear phenomenon which is fully characterised by its static characteristics, i.e., its dynamics can be neglected

Saturated actuators

Relay control

Gears backlash

Hysteresis in magnetic materials

Dead zone in electro-mechanical systems

Saturated actuators

Relay control

Gears backlash

Saturated actuators

Relay control

Gears backlash

u (driving gear)

m (driven gear)

Saturated actuators

Relay control

Gears backlash

Saturated actuators

Relay control

Gears backlash

Nonlinearities are not always a drawback, they can also a have a“stabilising” effect

+r(t) e(t) u(t) m(t) y(t)_

( ) ( )24.111

( ) ( )ssH

1.015.0

( ) 10=sC

( )( )2

0 2 4 6 8 10 12 14 16 18 20-20

20Step response

time [s]

NOT saturated actuatorSaturated actuator

The saturated system oscillates but does not diverge

The NOT saturated system is not stable since its stability margin are negatives

• mg = -12 dB

• mϕ = - 47 deg

Bode Diagram of the difference transfer function

Frequency (rad/sec)

System: sys Frequency (rad/sec): 1.68

Phase (deg): -227

System: sys Frequency (rad/sec): 0.941 Phase (deg): -180

50 System: sys Frequency (rad/sec): 1.68 Magnitude (dB): -0.0708

System: sys Frequency (rad/sec): 0.94

Magnitude (dB): 12

Many systems can be reduced to a simplified form in which the all linear dynamics is concentrated in a unique block and the static non linear characteristics is represented by a separate block

f(u) P(jω)C(jω)

H(jω)

+r(t) e(t) u(t) m(t) y(t)_

f(u) P(jω)H (jω)C (jω)C(jω)+r’(t) u(t) m(t) w(t)

If constant (or very slowly varying) reference signals are considered, under some conditions it is possible to separate the low-frequency, almost static, behaviour defined by a working nominal condition and a high-frequency behaviour due to small variations around such a working point

f(u) P(jω)H (jω)C (jω)C(jω)+r’(t) u(t) m(t) w(t)

r(t)=R

f(u) G (jω)+r’(t)=kCR u(t) m(t) w(t)

( ) ( )( ) ( )( ) ( )

GC −=

The non linear characteristics is translated so that the origin of the new reference Cartesian system is the point (u0, m0) in the original one

f’(u) G (jω)+Δr’(t)=0 Δu(t) Δm(t) Δw(t)

( )⎪⎩

⎪⎨⎧

( )( ) ( ) 00

muufuf

−Δ+=Δ′

Δ′=Δ

The aim of the system analysis is to define which is the steady-state behaviour of the system defined by the variations around the nominal working point

If f’(u) is a passive sector function absolute stability tools allow for sufficientconditions for global asymptotic stability of the variation system, i.e, the steady state is characterised by constant values of the system variables

If the origin of the variation system is not stable does the varia-bles diverge to infinity or some periodic motion can appear?

Limit cycles

Limit cycle: a periodic oscillation around a constant working point

f(u) G (jω)+r’=5δ−1(τ) u(t) m(t) w(t)

( ) ( )( )

( ) ( )1116

14.1110

-6 -4 -2 0 2 4 6 8-1

+−= um

Limit cycles

Limit cycle: a periodic oscillation around a constant working point

0 5 10 15 20 25 300

time [s]

(t) w0

545.41150

Limit cycles

Limit cycle can be a drawback in control systems:

Instability of the equilibrium point

Wear and failure in mechanical systems

Loss of accuracy in regulation

Parameters of the limit cicle can be used to discriminate between acceptable and dangerous oscillations

oscillation frequency

oscillation magnitude

Electronic oscillators can be based on limit cycles

Describing Function - Assumptions

The describing Function approach to the analysis of steady-state oscillations in non linear systems is an approximate tool to estimate the limit cycle parameters.

It is based on the following assumptions

There is only one single nonlinear component

The nonlinear component is not dynamical and time invariant

The linear component has low-pass filter properties

The nonlinear characteristic is symmetric with respect to the origin

There is only one single nonlinear component

The system can be represented by a lumped parameters system with two main blocks:

•The linear part

•The nonlinear part

N.L. G (jω)+r’=const. u(t) m(t) w(t)

The nonlinear component is not dynamical and time invariant

The system is autonomous. All the system dynamics is concentrated in the linear part so that classical analysis tools such as Nyquist and Bode plots can be applied.

The linear component has low-pass filter properties

This is the main assumption that allows for neglecting the higher frequency harmonics that can appear when a nonlinear system is driven by a harmonic signal

( ) ( ) K ,3 ,2=>> njnGjG ωω

The more the low-pass filter assumption is verified the more the estimation error affecting the limit cycle parameters is small

The nonlinear characteristic is symmetric with respect to the origin

This guarantees that the static term in the Fourier expansion of the output of the nonlinearity, subjected to an harmonic signal, can be neglected

Such an assumption is usually taken for the sake of simplicity, and it can be relaxed

Fourier expansion - Recall

Consider a periodic function

( ) constant real a is ),()(,)( TTtytytfty −==

( ) ( )( )∑∞

220 cossin2

TkTk tkbtkaaty ππ

( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( )tdtktfdttktfT

tdtktfdttktfT

πππ

sin1sin20,2,1,0

cos1cos2

∫∫

−−

Describing Function – Harmonic balance

( )tUtu ωsin)( =Δ

( ) ( )( )∑∞

sincos)(k

kk tkbtkatm ωω

( ) ( )( )( ) ( )ωϕω

ϕωϕω

jkGjkGG

tkbtkaGtw

kkkkkk

+++=Δ ∑∞

=1sincos)(

tjUetu ω=Δ )(

eeMtm k

arctan)(

+== ∑

= ϑωϑ

Consider the polar representation of a complex number associatedwith the exponential form of harmonic signals

Taking into account the low-pass property of the linear part of the system

jk eeMeGeeMeGtw kk ωϑϕωϑϕ 11

)( ≅=Δ ∑∞

A permanent oscillation in the loop appears if Δu(t)=-Δw(t)

tjjjtj eeMeGUe ωϑϕω 1111−=

N.L. G (jω)+Δr’(t)=0 Δu(t) Δm(t) Δw(t)

01 11 11 =+ ϑϕ jj e

( ) ( ) 0,1 =+ ωω UNjG

( ) ( )111, jabU

UN +=ω is the Describing Function of the nonlinear term

Harmonic balance equation

The armonic balance equation is a necessary condition for the existence of limit cycles in the nonlinear system

The approximate analysis gives good estimates if the low-pass filter hypothesis is strongly verified. It is a good tools for engineers

The harmonic balance equation is similar to the characteristic polynomial function, i.e. it leads to the Nyquist condition for closed-loop stability

The Describing Function is a linear approximation of the static nonlinearity limited to the first harmonic

( ) ( )ωω

UNjG −=

In most cases the Describing Function is not a function of the frequency and this simplifies the verification of the harmonic balance equation by means of the Nyquist plot of the transfer function

Nyquist Diagram

Real Axis

-1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2-10

ω=+∞

G(jω)

U=0+∞← U (U0,ω0)

-\1/N(U)

Nyquist and DF Diagrams

Real Axis

-8 -7 -6 -5 -4 -3 -2 -1 0 1-3.5

ω=100

ω=17.7828

ω=3.1623

ω=0.56234

ω=0.1

System: sys Real: -1.14 Imag: -2.02

Freq (rad/sec): 0.651

System: sys Real: -2.01

Imag: -0.021 Freq (rad/sec): 0.997

Describing Function – Computation

The evaluation of coefficients a1 and b1 can be performed by means of both analytical calculation and numerical integration, depending on the type of nonlinearity involved

f’(u)Usin(ωt) Δm(t)

( ) ( )111, jabU

UN +=ω( ) ( )

( ) ( )dtttmT

dtttmT

The DF computation can be performed by means of its definition

MUN 4, =( ) ( ) ( )π

ϑϑπ

MdMdtttmT

4sin4sin2 22

2 01 ==Δ= ∫∫

Ideal relay

-3 -2 -1 0 1 2 3-3

Δ u(t)

0 0.5 1 1.5 2 2.5 3-2

Δ uΔ m

( ) ( ) 0cos2 2

1 =Δ= ∫−

dtttmT

ω Because of the odd symmetry of the Δm(t) signal

Ideal relay

A limit cycle can exist if the relative degree of G(jω) is greater than two

Real Axis

-1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2-4

-1/N(U)

G(jω)

(ωc;4M/π)

The negative reciprocal of the DF is the negative real axis in backward direction

The oscillation frequency is the critical frequency ωc of the linear system and the oscillation magnitude is proportional to the relay gain M

MjUN 4, −=( ) ( ) ( )

πϑϑ

MdMdtttmT

4cos2cos2 22

2 01 −=−=Δ= ∫∫

Pure hysteresis

( ) ( ) 0cos2 2

1 =Δ= ∫−

dtttmT

ω Because of the even symmetry of the Δm(t) signal

0 0.5 1 1.5 2 2.5 3-2

Δ uΔ m

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5-2

A limit cycle can exist if the relative degree of G(jω) is greater than one and G(jω) is a type-0 system

The negative reciprocal of the DF is the negative imaginary axis in backward direction

The oscillation frequency is lower than the critical frequency ωc of the linear system and the oscillation’s magnitude is proportional to the relay gain M and to the modulus of the transfer function at phase -π/4

Real Axis

-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4-1.6

G(jω)

-1/N(U)

Pure hysteresis

⎪⎪⎩

⎪⎪⎨

>−⎟⎠⎞

⎜⎝⎛−

≤−=

ββπ

,( ) ( )∫ ∫ =+−=γ

cos8cos8

Hysteretic relay

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5-2

β -β

If U≤β the hysteretic relay behaves as pure hysteresis

0 0.5 1 1.5 2 2.5 3-2.5

Δ uΔ m

( ) ( )2

01 14sin8sin8 4

⎟⎠⎞

⎜⎝⎛−=+−= ∫ ∫ U

ωωγ

The imaginary part of N(U) is proportional to the hysteresis area

Hysteretic relay

If β is larger than U the relay could not behave as a purehysteresis

Real Axis

-1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2-4

G(jω )

U -1/N(U)

The oscillation frequency is lower than the critical frequency ωc of the linear system and the oscillation’s magnitude is proportional to the relay gain M

The negative reciprocal of the DF is the parallel to the negative real axis and with constant negative imaginary part

⎪⎪⎩

⎪⎪⎨

>⎥⎥⎦

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛+⎟

⎠⎞

⎜⎝⎛

1arcsin2π

Saturation

If U≤Μ the saturation behaves as pure gain

( ) ( )⎥⎥⎦

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛+⎟

⎠⎞

⎜⎝⎛=+= ∫ ∫

21 1arcsin2sin8sin8 4

kUMkUdtt

TMdttkU

πωω

0 0.5 1 1.5 2 2.5 3-1

Δ uΔ m

( ) ( ) 0cos2 2

1 =Δ= ∫−

dtttmT

-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5-4

A limit cycle can exist if the relative degree of G(jω) is greater than two and gain k is sufficiently high

The negative reciprocal of the DF is part of the negative real axis in backward direction

The oscillation frequency is the critical frequency ωc of the linear system and the oscillation’s magnitude depends on the saturation parameters M and k

Saturation Nyquist and DF Diagrams

Real Axis

-1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2-4

G(jω) ω

-1/N(U)

( )⎪⎩

⎪⎨

>⎥⎥⎦

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛+⎟

⎠⎞

⎜⎝⎛−

= ββββπ

1arcsin2

Dead zone

If U≤β the dead zone has no output

( )γωβ tU sin=

( ) ( ) 0cos2 2

1 =Δ= ∫−

dtttmT

-3 -2 -1 0 1 2 3-3

0 0.5 1 1.5 2 2.5 3-2.5

Δ uΔ m

( )( ) ( )⎥⎥⎦

⎢⎢⎣

⎡⎟⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛−⎟

⎠⎞

⎜⎝⎛−=−= ∫

1 1arcsin2

2sinsin8 4

UUUkUdtttUk

βββππ

ωβωγ

A limit cycle can exist if the relative degree of G(jω) is greater than two and gain k is sufficiently high

The negative reciprocal of the DF is part of the negative real axis in forward direction

The oscillation frequency is the critical frequency ωc of the linear system and the oscillation magnitude depends on the dead zone parameters β and k

Dead zone Nyquist and DF Diagrams

Real Axis

-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0-2.5

U=+∞

-1/N(U)

G(jω)

The nonlinear characteristics of the Dead Zone can be computed by subtracting the Saturation characteristics from a linear one

Dead zone

-3 -2 -1 0 1 2 3-3

dead zone ψ(Δu)

saturation φ (Δu)

( ) ( )uku ΔΦ−=ΔΨ

( ) ( )UNkUN ΦΨ −=

The Describing function of a nonlinear characteristics can be computed as the combination of the Describing Functions of the elementary constituting nonlinear characteristics

( ) ( ) ( )UNUNUNt 21 +=

N2(U,ω)G (jω)

+Δr’(t)=0 Δu(t)Δm(t) Δw(t)

N1(U,ω)

Nt(U,ω)

A number of Describing Function can be computed by particularisation of the function

in which the parameter α defines a peculiar point of the nonlinear characteristics

( ) ( )[ ]21arcsin2 αααπ

α −+=Φ

-α ( )( )⎪⎩

⎪⎨⎧

<Φ⋅

( )( ) ( )⎪⎩

⎪⎨⎧

<Φ⋅−+

+α-α

( )( )( )⎪⎩

⎪⎨⎧

<Φ−

+α-α

( ) ( )( )( ) ( )( )⎪

⎪⎩

⎪⎪⎨

<Φ−Φ⋅

≤<Φ−⋅

βαβ

βαα

+β-α

α MkN 4+=

+α-β

α MN 4=

−α( )

⎪⎩

⎪⎨⎧

2 ααπ

αα MN

( ) ( )[ ] ( )⎪⎩

⎪⎨⎧

<−−−Φ−

1141212

ααπαα

αα kjkNu/U

+α-α

( ) ( )[ ] ( )⎪⎩

⎪⎨⎧

<−+−Φ−

1141212

ααπαα

αα kjkNu/U

+α-α

( )⎪⎩

⎪⎨⎧

141410

, 2 απ

ααπ

+α-α

( )⎪⎩

⎪⎨⎧

<−−

141410

, 2 απ

ααπ

UMUNu/U

+α-α

( )ωππ

UMUN 44, −=

( ) ( ) ( )ωj

UNUNUNt1

1/jωG (jω)

+Δr’=0 Δu Δm Δw_

Nt(U,ω)

( ) ( )( )jM

−=− ω

ωωπ

ω 214,1

-8 -7 -6 -5 -4 -3 -2 -1 0 1-3.5

ω=0.1

ω=0.56234

ω=3.1623

ω=17.7828

ω=100

Real axis

Negative inverse of the Describing Function

Stability of limit cycles

If the block N.L. is a pure constant k, the stability of the feedback system can be performed by means of the Nyquist criterion, which gives the number of roots with positive real roots of the Harmonic Balance Equation

N.L. G (jω)+Δr’(t)=0 Δu(t) Δm(t) Δw(t)

( ) 011 =+k

The Nyquist criterion looks at the relative position of the transfer function G(jω) with respect to the point (-1/k, 0) in the complex plain.By extension the criterion can be applied to any point –α of the complex plain with reference to the Harmonic Balance Equation

( ) C∈=+ ααω ,01 jG

If the transfer function G(jω) represents a stable system, the reducedNyquist criterion can be applied, i.e. the closed loop stability can be stated if the reference point –α in the complex plain remains on the left-side when running along the Nyquist plot of G(jω) from ω=0+ to ω =+∞.

Nyquist plot of G(jω)

Real Axis

-3 -2.5 -2 -1.5 -1 -0.5 0-4

ω=+∞

( )( )( )15.0

=ωωω

The closed loop is not stable with respect to the point (-1/k,0).

The closed loop is stable with respect to the point -α

Can the generalization of the Nyquist criterion be used to analyse the stability of a limit cycle?

If a perturbation of the magnitude of the periodic oscillation occurs, it tends to the original value as time passes.

What does it mean that a limit cycle is stable?

YES, it can

It is a marginally stable condition.How can a limit cycle be considered from the Nyquist criterion point of view?

By a point in the Describing Function plot.How can the magnitude of a limit cycle be represented?

Nyquist and DF plots

Real Axis

-3 -2.5 -2 -1.5 -1 -0.5 0-3

G(jω) ω

Points B, B’, B”, A and A’represent possible magnitudes of the oscillation.

A and B re-present two possible limit cycles

Applying the reducedNyquist criterion with respect to:

point B’: oscillations tend to decrease (stable system)

point B”: oscillations tend to increase (unstable system)

point A’: oscillations tend to decrease (stable system)

B: Stable limit cycle A: Unstable limit cycle

Real Axis

-3 -2.5 -2 -1.5 -1 -0.5 0-3

G(jω ) ω

BB’ B”

The linear system G(jω) is assumed to be stable in order to apply the reduced Nyquist criterion

DF Analysis – Example of application

Consider a DC motor with permanent magnets

M d.c.

The position of the motor shaft is measured by means of a rotational variable resistance.

The voltage on the rotational resistance drives the position of a relay that switches the motor supply voltage between +/- 24 V d.c.

The linear approximate model of a DC motor is the following

( ) ( ) ( ) ( )( ) ( ) ( )

( ) ( )( ) ( )

tktedt

tBtCdt

tdiLtiRtv

( )⎪⎩

⎪⎨⎧

≥−

r tvϑϑ

Rr: rotor resistanceLr: rotor inductanceke: voltage feedback constantkt: torque constantJm: motor inertiaJl: load inertiaBm: motor friction coefficientBl: load friction coefficient

vr: rotor supply voltageir: rotor wound currentCem: electromagnetic torqueω: rotational speedθ: shatf angular position

The motor transfer functions are the following

( ) ( )( ) ( )( ) etaa

s kkBsJRsLk

( ) ( )( ) ( )( )( ) ( ) ( )

( )( ) ( )

( )( ) ( )[ ]2222

JLkkBRBLJR

JLkkBRkj

JLkkBRBLJR

jjjkkBJjRLjj

kjVjjW

aetaaa

ωωω

ωωωωωω

ωωϑ

−+++

−+−

−+++

ℑ+ℜ=+++

Taking into account the parameters of the motor and of the load

R=0.4; % rotor resistance

L=0.001; % rotor inductance

ke=0.3; % voltage feedback constant

kt=0.3; % torque constant

Jm=0.01; % motor inertia

Jl=0.09 % load inertia

Bm=0.05; % motor friction coefficient

Bl=0.05; %load friction coefficient

J=Jm+Jl;

B=Bm+Bl;

Real Axis

-2 -1.5 -1 -0.5 0 0.5 1-20

( )( ) ( )( ) rad 0.17594M-Urad/s 056.360

=ℜ⋅==+

===ℑ crp

etacrjW

ωωω

J=Jm+Jl;

B=Bm+Bl;

Real Axis

-20 -15 -10 -5 0

x 10-3

System: Wp Real: -0.00568

Imag: -8.73e-007 Freq (rad/sec): 36.8

( )( ) ( )( ) rad 0.17594M-Urad/s 056.360

=ℜ⋅==+

===ℑ crp

etacrjW

ωωω

J=Jm+Jl;

B=Bm+Bl;

Real Axis

-20 -15 -10 -5 0

x 10-3

System: Wp Real: -0.00568

Imag: -8.73e-007 Freq (rad/sec): 36.8

STABLE limit cycle!!!!

The system presents a periodic steady-state oscillation

0 1 2 3 4 5 6 7 8 9 10-80

anglespeedcurrent

0 1 2 3 4 5 6 7 8 9 100

8.4 8.45 8.5 8.55 8.6 8.65 8.7 8.75 8.8

Time [s]

Tc=2π/ωc

2U=8M|G(jωc)|/π

What does it happen if the same control law is applied to the speed control problem?

Step Sign

DC Motor

( ) ( )( ) etaa

kkBsJRsLksW

+++=ω

The linear plant is characterised by an all-pole transfer function with relative degree two, therefore there is no contact point between the Nyquist plot and the real negative axis, but the origin at ω=+∞ (corresponding to U=0 in the DF plot)

Real Axis

-1 -0.5 0 0.5 1 1.5 2 2.5 3

ω=0 ω=+∞U=0

The Nyquist plot of the linear system is tangentto the negative reciprocal of the DF at the origin, i.e., U=0 and ω=+∞

Maybe a sliding mode behaviour is established asymptotically!?

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

anglespeed

Possibly a sort of dead-bit control, or finite time stabilisation seams to appear

Can it be a sliding mode behaviour?

By reducing the integration step of the simulation it is apparent that the stabilisation is asymptotic

Can it be a sliding mode behaviour?

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.080

Time [s]

The rotor current tends to a constant value, i.e. an asymptotic (2nd

order) sliding sliding mode appears

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1-40

Time [s]

By reducing the integration step of the simulation it is apparent that the stabilisation is asymptotic

The DC motor is a second order system whose state variables are the rotor speed and current

( ) ( ) ( )( ) ( ) ( ) ( )

( )( )

( ) ( ) ( ) ( )( )( )⎥

⎥⎦

⎢⎢⎣

⎡⎥⎦⎤

⎢⎣⎡ +=++=

⎥⎥⎥

⎢⎢⎢

⎥⎥⎦

⎢⎢⎣

⎥⎥⎥⎥

⎢⎢⎢⎢

−−=

+−−=

ωωω

The system transfer function has a zero in –2, and if the system output is steared to zero in a finite time, than the system behaves as a first order system with time constant τ=0.5

A MATLAB-Simulink scheme is the following

ScopeRelay

Manual Switch

Derivative

DC Motor

Take care that because of the feedback, the new closed loop gain (assuming the rotor velocity as the output) will be 1/2

It is apparent that the shaft speed tends to the steady state value π as a first order system with τ = 0.5 s

0 0.5 1 1.5 2 2.5 3 3.50

Time [s]

0 0.5 1 1.5 2 2.5 3 3.50

Speed [rad/s]

]Apart from a very short transient, the state trajectory tends to the steady-state values sliding on a linear manifold of the state space

The Nyquist plot of the linear system cross the negative reciprocal of the DF at the origin, i.e., U=0 and ω=+∞

A sliding mode behaviour is established in a finite time

Real Axis

-2 -1 0 1 2 3 4 5 6 7 8-4

DF Analysis & Sliding Modes

Sliding modes are characterised by infinite frequency of proper ideal switching devices

Most sliding mode controllers use a sign function in the controller, i.e., an ideal relay, which can be approximately represented by its Describing Function

By the example it is apparent that a sliding mode behaviour is established in a finite time if the Nyquist plot of cross the inverse negative of the describing function at the point (U=0,ω=+∞)

It can also be derived that a sliding mode behaviour is established asymptotically if the Nyquist plot of is tangent to the inverse negative of the describing function at the point (U=0,ω=+∞)

TAKE CARE: the Describing function is an approximate tool

ScopeRelay

Manual Switch

Derivative

DC Motor

Substitute the ideal relay with a hysteretic one with β=1

It is apparent that an ideal sliding mode cannot appear because of the switching delay

The limit cycle is apparent, with a periodTlc≅0.6 ms

The magnitude is very small because of the low-pass filter property of the motor transfer function

0.4 0.4001 0.4002 0.4003 0.4004 0.4005 0.4006 0.4007 0.4008 0.4009 0.401-1

1x 10-5

Time [s]

0.4 0.4001 0.4002 0.4003 0.4004 0.4005 0.4006 0.4007 0.4008 0.4009 0.401

The Nyquistplot of the linear system has no common point with the negative inverse of the Describing Function of the hysteretic relay

Describing function analysis can be useful in sliding mode systems when a common point is present

Real Axis

-2 -1 0 1 2 3 4 5 6 7 8-4

ω=+∞ ω=0

U=1 -1/N(U)

G(jω)

Most effective use of the Describing Function approach to sliding modes….

Analysis of the characteristics of a chattering behaviour due to unmodelleddynamics of sensors and/or actuators

What is chattering? It appears as oscillations of the system variables, whose magnitude is related to the influence of the neglected dynamics on the system

bandwidth

It is very close to a limit cycle

Consider the motor drive as a hysteretic switching device plus a time constant τa=0.1 s

r=2*pi

0.1s+1Transfer Fcn

SignScope3Scope2

ScopeRelay

Manual Switch

Derivative

DC Motor

Real Axis

-2 -1 0 1 2 3 4 5 6 7 8-4

-1/N(U)

G(jω)

Nyquist plot parametersωlc=662 rad/s

Ulm=1.1833 rad/s2

Uls represents the magnitude of the oscillation of the sliding variable

-1/N(U)

G(j ω)

System: untitled1 Real: -0.0507 Imag: -0.0328 Freq (rad/sec): 662

Ulm=1.1833 rad/s2

Simulation resultsTls=9.7 ms

ωlc=646 rad/sUlm=1.8544 rad/s2

2.9 2.91 2.92 2.93 2.94 2.95 2.96 2.97 2.98 2.99 3-2

Steady-state regime

Time [s]

Differences are due to:• Numerical solution of

the simulation• Approximation of the

DF approach

The rotor velocity tends to its steady-state value as a first order system

with time constant τ=0.5s

Rotor wound current presents large variations

An approximate sliding mode is established

0 0.5 1 1.5 2 2.5 3 3.50

Time [s]

VelocityCurrent

⎥⎥⎦

⎢⎢⎣

⎡⎥⎦⎤

⎢⎣⎡ +=

⎥⎥⎥

⎢⎢⎢

⎥⎥⎦

⎢⎢⎣

⎥⎥⎥⎥

⎢⎢⎢⎢

−−=

⎥⎥⎦

⎢⎢⎣

s+−= ωω 2&

Full system dynamics

r vJLk

+⎟⎟⎠

⎞⎜⎜⎝

⎛−⎟⎟

⎞⎜⎜⎝

⎛−⎟

⎠⎞

⎜⎝⎛ ++⎟⎟

⎞⎜⎜⎝

⎛−+= ωσσ 222&Input-output dynamics

Internal reduced-order dynamics

ωσ ⎟⎠⎞

⎜⎝⎛ +−=

σωω +−= 2&Internal reduced-order dynamics ( ) ( )ss

( ) ( )tUt lcωσσ sin0 +=

At steady-state( ) ⎟

⎟⎠

⎞⎜⎜⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

1argsin2

ωωω

2.9 2.91 2.92 2.93 2.94 2.95 2.96 2.97 2.98 2.99 33.07

Time [s]

Uls=1.1833 rad/s2

Δωls=0.0018 rad/s

Simulation resultsTls=9.7 ms

ωlc=646 rad/sUls=1.8544 rad/s2

Δωls=0.0029 rad/s

0015.02

+lsjω

The Describing Function approach to the analysis of the chattering phenomenon in sliding mode control systems can be used to have an estimate of the chattering parameters, i.e., frequency and magnitude

The estimates are affected by an error which depends on the low-pass properties of the linear part of the plant

The sliding variable must be considered as the output of the nonlinear feedback system

The actual system output behaviour can be estimated by considering the reduced order dynamics

In the presence of a constant reference value, the nonlinear function can be not symmetric, therefore an equivalent gain of the nonlinearity has to be considered to estimate the constant mean steady-state value of the output

Describing Function analysis-v1 - people.unica.it · Describing Function analysis of nonlinear...

Documents

Transcript of Describing Function analysis-v1 - people.unica.it · Describing Function analysis of nonlinear...

Describing Function analysis-v1 - Diee · Describing Function analysis of nonlinear systems – Prof Elio USAI –March 2008 System definition and problem statement In many cases

The describing function

Example: Linearization of a nonlinear model involving a nonlinear function of two variables.

Nonlinear Function Approximation - K.N.Toosi University of Technology

7 SINUSOIDAL-INPUT DESCRIBING FUNCTION (DF)

Describing Function Recorjkjkjkjkding With Simulink and MATLAB

Describing Function Slides

DESCRIBING FUNCTION ANALYSIS:

Inclusion of the Memory Function in Describing the Flow of ...faculty.kfupm.edu.sa/PET/menamul/Artecal_publeshed/Art 14_IJE... · Inclusion of the Memory Function in Describing the

Global existence for fully nonlinear reaction-diffusion ... · Global existence for fully nonlinear reaction-diffusion systems describing multicomponent reactive ﬂow Martine MARION,

Describing Function Properties of a Magnetic Pulsewidth Modulator · 2016-04-07 · 1. Introduction Describing Function Properties of a Magnetic Pulsewidth Modulator RD.MIDDLEBROOK,Senior

2020) · 2021. 1. 2. · The linearization of nonlinear systems and the quasi-linearization method, namely the covariance analysis describing function method, are proposed by the

mitter/publications/46_ENCY_nonlinear_filt.pdf · 2013-03-12 · of nonlinear feedback systems is the describing-function method (Atherton 1975). Stability criteria can be based on

MULTIPLE-INPUT DESCRIBING FUNCTIONS AND NONLINEAR SYSTEM DESIGN

MULTIPLE-INPUT DESCRIBING FUNCTIONS AND NONLINEAR SYSTEM ... · PDF fileMULTIPLE-INPUT DESCRIBING FUNCTIONS AND NONLINEAR SYSTEM DESIGN ... GELB AND VANDER VELDE Multiple-input describing

Nonlinear Optimization: Introduction Linear vs. nonlinear ...Nonlinear Optimization: Introduction Linear vs. nonlinear objective functions C /LQHDU C 1RQOLQHDU When objective function

COURSE STRUCTURE AND SYALLABUS (PE... · 2013. 11. 8. · describing function analysis of nonlinear systems – Stability analysis of Non-Linear systems through describing functions

Splash Screen 7 - 10. Main Idea/Vocabulary nonlinear function Determine whether a function is linear or nonlinear. 7 - 10.

8IN NONLINEAR SYSTEMS NONOSCILLATORY TRANSIENTS...TRANSIENT-INPUT DESCRIBING FUNCTION 441 Figure 8.1-2 Testing nonlinearity by actual input. (Chen, Ref. 3.) been chosen to correct

NONLINEAR SYSTEMS - Tripod.comcsindulkar.tripod.com/nonlinearsystems.pdf · NONLINEAR SYSTEMS Describing function of a nonlinear device ( output is the cube of its input) Stability