mws slide

Upload: panoot-pattapongse

Post on 04-Feb-2018

223 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/21/2019 Mws Slide

    1/116

    1 Introduction

    what is a control system?

    what are the basic components of a control system?

    some examples of control system applications

    open-loop and closed-loop systems

    effects of feedback

    1 Introduction 2

    Control systems and models

    Acontrol systemis a [combinationorsetorgroup] of [componentsorelementsordevices] [connectedoracting] together under a set of[regulationsorrules] to perform a [certain objectiveoruseful task].

    Fig. 1-1: Basic components of acontrol system.

    System model

    inputs

    outputs

    boundarysystem

    {law rule}

    environment = components

    inputs outputssystem

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    2/116

    1 Introduction 3

    A Control System

    contains a component called controller whose role is to control thesystem output so that an objective is achieved.

    Example: A simple control system

    r controllingcomponent

    controller

    A control system

    componentcontrolled

    plant

    u y

    This type of control systems is called open-loop.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    1 Introduction 4

    An Example: Open-Loop Temperature Control

    Consider as an example

    r

    component

    air-conditioner

    A room with air-conditioning system

    componentcontrolled

    a.c. room

    u ycontrolling

    where r fan speed control (LO, MED, HI)u amount of cool airy room temperature

    Open-loop controller neglects information about room-size, numberof persons in room, inside- and outside-temp., time of day, etc.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    3/116

    1 Introduction 5

    Closed-Loop Temperature Control

    Consider theclosed-loopconfiguration

    room

    u

    a.c.thermostat

    componentcomparing e

    ry

    d

    wheredrepresents the information being neglected. The controller(block a.c.) called closed-loop controller can bedesignedto makeyless insensitive tod.

    Benefits

    more comfortable

    save energy, etc.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    1 Introduction 6

    Feedback Control

    Also known as Automatic ControlCan reduce the effect of disturbancedonybecause

    d affectsyin the feedforward manner,

    andyisfed backto the feedforward path for making decision.

    In this way, the closed-loop control makes the cause-and-effectrelationship a complete loop.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    4/116

    1 Introduction 7

    Control system example: Sun-tracking solar collectors

    Fig. 1-5: Solar collector field.

    the collector dish must trackthe sun accurately

    a predetermined desired rateis modified or trim by actualposition errors determined bythe sun sensor

    the controller constantlycalculates the suns rate forazimuth and elevation

    Fig. 1-7: Important components of

    the sun-tracking control systems.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    1 Introduction 8

    Open-loop and closed-loop control systems

    Open-loop or nonfeedback control

    good for plant without disturbances

    good for plant with accurate model

    require less equipment (e.g. sensor, comparing device)

    no stability problem

    Closed-loop or feedback control

    reduce the effect of unknown disturbances

    tolerate model inaccuracy

    require more equipment

    need good design to avoidinstability

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    5/116

    1 Introduction 9

    Effect of feedback

    complete the cause-and-effect relationships

    increase the gain of a system in one frequency range but decrease itin another

    can improve stability or be harmful to stability

    can increase or decrease the sensitivity of a system

    can reduce the effect of noise

    can affect bandwidth, impedance, transient response, steady-stateresponse, and frequency response

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    6/116

    2 Mathematical Foundation

    Laplace transform

    inverse Laplace transform by partial-fraction expansion

    application of Laplace transform to the solution of linear ODE

    impulse response and transfer functions of linear systems

    MATL AB tools and case studies

    2 Mathematical Foundation 2

    Laplace transform

    Definition Given a real functionf(t)satisfying

    0 |f(t)et| dt

  • 7/21/2019 Mws Slide

    7/116

    2 Mathematical Foundation 3

    Impulse function

    An impulse (or Dirac delta) function is defined as the limit of arectangular pulse function:

    (t) =

    1, 0 t

  • 7/21/2019 Mws Slide

    8/116

    2 Mathematical Foundation 5

    Important theorems of Laplace transform

    Linearity

    L[f1(t) +f2(t)] =F1(s) +F2(s)

    DifferentiationL[f(t)] =sF(s) f(0)

    L[f(t)] = sL[f

    (t)] f(0)

    = s2F(s) sf(0) f(0)Integration

    L

    t0

    f() d

    =

    1

    sL[f(t)] =

    F(s)

    s

    L t

    0

    1

    0

    f(2) d2 d1=F(s)

    s

    2

    Shift in time

    L[f(t T)us(t T)] =eT sF(s)

    Initial value theoremlimt0

    f(t) = lims

    sF(s)

    if the limit exists.

    Final value theoremlim

    tf(t) = lim

    s0sF(s)

    ifsF(s)is analytic inRe s 0.

    Complex shifting

    L[etf(t)] =F(s )

    Real convolution

    F1(s)F2(s) =L[f1(t) f2(t)]=L

    t0

    f1()f2(t) d

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 6

    0+ and 0

    If we define

    F+(s) =L+[f(t)] =

    0+

    f(t)est dt and F(s) =L[f(t)] =

    0

    f(t)est dt

    we must replace0in the theorems by0+or0 accordingly, e.g.,

    L+[f(t)] =sF+(s) f(0+), L+

    t0+

    f() d

    =

    F+(s)

    s , f(0+) = lim

    ssF+(s)

    Then 0+ or 0? Both work fine. For example,consider du(t)

    dt =(t)

    and the differentiation formula,

    1

    0 1 2

    u(t)

    u(0) = 0u(0+) = 1

    L+[u(t)] = sL+[u(t)] u(0+)= s

    1

    s 1 =L+[(t)] = 0

    L[u(t)] = sL[u(t)] u(0)= s

    1

    s 0 =L[(t)] = 1

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    9/116

    2 Mathematical Foundation 7

    Inverse Laplace transform

    The inverse Laplace transform ofF(s)can be found from

    f(t) =L1[F(s)] = 1

    2j c+j

    cjF(s)est ds

    wherecis a real constant chosen so thatF(s)is analytic in regionRe s c.

    The method of partial-fraction expansion can also be used to findL1[]ofrational functions.

    Rational function Ifan= 0we call the function

    P(s) =ansn +an1sn1 + +a1s+a0

    apolynomialofdegreenwitha1,a2, . . . ,anas itscoefficients.

    Rational functionsare fractions of polynomials (c.f. rational numbersare fractions of integers).

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 8

    Partial-fraction expansion: Real roots

    Ifis a real root ofP(s)withmultiplicitym, i.e.,P(s) = (s )mR(s),then we write

    Q(s)

    P(s)=

    A1s

    + A2

    (s )2+ +

    Am(s )m

    + {terms ofR(s)}

    Multiplying both sides by(s )mQ(s)

    R(s)=Am+ +A2(s )

    m2 +A1(s )m1 + {terms ofR(s)}(s )m

    Letting(s) = (s )mQ(s)

    P(s)=

    Q(s)

    R(s), we can findAkfrom

    Ak = 1

    (m k)!(mk)(), k= 1, 2, . . . , m

    and its correspondingL1[]term is

    L1

    Ak(s )k

    = et

    tk1

    (k 1)!

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    10/116

    2 Mathematical Foundation 9

    Partial-fraction expansion: Complex conjugate roots

    If+jis a complex root (multiplicitym) ofP(s)with , i.e.P(s) = (s j)mR(s), then we write

    Q(s)P(s)

    = A1+jB1s j

    + + Am+jBm(s j)m

    + A1 jB1s +j

    + + Am jBm(s +j)m

    + {terms ofR(s)}

    Letting(s) = (s j)mQ(s)

    P(s), we have

    Ak+jBk = 1

    (m k)!(mk)(+j), k= 1, 2, . . . , m

    and its L

    1

    [](with its conjugate pair) isL1

    Ak+jBk

    (s j)k+

    Ak jBk(s +j)k

    = et

    tk1

    (k 1)!

    (Ak+jBk)e

    jt + (Ak jBk)ejt

    = et tk1

    (k 1)!(2Akcos t 2Bksin t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 10

    Partial-fraction expansion: Examples

    Example 1 5s+ 3

    (s+ 1)(s+ 2)(s+ 3)= 1

    s+ 1+

    7

    s+ 2+ 6

    s+ 3

    Example 2 2

    s(s+ 1)3(s+ 2)=

    1

    s+

    1

    s+ 2

    2

    s+ 1

    2

    (s+ 1)3

    Example 3 2(s + 1)

    s2(s2 + 2s+ 2)=

    1

    s2+

    j/2

    s+ 1 +j+

    j/2

    s+ 1 j

    Example 4 G(s) = 2n

    s2 + 2ns+2n=

    n1 2

    j/2

    s++j

    j/2

    s+ j

    where = nand =n1 2 or

    g(t) =L1[G(s)] = n1 2

    ent sin nt

    1 2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    11/116

    2 Mathematical Foundation 11

    Application of Laplace transform to solution of linear ODE

    Example Solvey(t) + 3y(t) + 2y(t) = 5us(t)wherey(0) =1,y(0) = 2.

    Take the Laplace transform of both sides of the ODE to get

    s2Y(s) sy(0) y(0) + 3sY(s) 3y(0) + 2Y(s) =5

    s

    Substituting initial conditions and solving forY(s), we get

    Y(s) = s2 s+ 5

    s(s+ 1)(s+ 2)

    which can be expanded by partial-fraction expansions to give

    Y(s) =

    5/2

    s

    5

    s+ 1+

    3/2

    s+ 2

    Taking the inverse Laplace transform, we get the complete solution

    y(t) =5

    2 5et +

    3

    2e2t, t 0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 12

    Impulse response

    Consider alinear time-invariantwith inputu(t)and outputy(t). Thesystem can be characterized by itsimpulse responseg(t)which is definedas the output when the input is a unit-impulse function(t).

    1

    0

    (t)

    (t) =

    1, 0 t

  • 7/21/2019 Mws Slide

    12/116

    2 Mathematical Foundation 13

    Transfer function

    Thetransfer functionof a linear time-invariant system isdefinedas theLaplace transform of the impulse response, withall the initial conditions

    set to zero. G(s) =L[g(t)]

    Letnandmbe the order of the denominator and the numeratorpolynomials ofG(s), respectively. The transfer functionG(s)is said to be[strictly]properwhen [n > m]n m. Ifn < m, we call itimproper.

    We can show that the transfer function is also theratiobetween theLaplace transform of the output and the Laplace transform of the input

    G(s) =Y(s)

    U(s)

    by the use of convolution, linearity, and time-invariance properties.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 14

    Approximation of input signal

    We can depict the approximation of input signal by

    aweighted sumof sequence of pulse functions(t n)

    as follows:

    u(t) = lim0

    n=0

    u(n)(tn)

    1

    2

    2 3 410 n

    u(t)

    u(n)(tn)

    . . .. . .

    u(n)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    13/116

    2 Mathematical Foundation 15

    Interpretation of convolution

    Defineg(t)to be the systems response to(t).

    input output(t) g(t) by definition

    (tn) g(tn) time invarianceu(n)(tn) u(n)g(tn) scaling

    n=0

    u(n)(tn)

    n=0

    u(n)g(tn) superposition 0

    u()(t ) d

    0

    u()g(t ) d

    dn ,

    By sifting property of the impulse function

    LHS=

    0

    u()(t ) d=u(t)

    andg(t)on the RHS is the response to(t)or impulse response.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    2 Mathematical Foundation 16

    MATLABtools and case studies

    >> [r,p]=residue([5 3],conv([1 1],conv([1 2],[1 3])))

    r = -6.0000 7.0000 -1.0000

    p = -3.0000 -2.0000 -1.0000

    >> [r,p]=residue([2],conv([1 2],[1 3 3 1 0]))

    r = 1.0000 -2.0000 -0.0000 -2.0000 1.0000

    p = -2.0000 -1.0000 -1.0000 -1.0000 0

    >> syms s>> G=2/s/(s+1)^3/(s+2)

    G =2/(s*(s + 1)^3*(s + 2))

    >> ilaplace(G)

    ans =1/exp(2*t) - 2/exp(t) - t^2/exp(t) + 1

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    14/116

    3 Block Diagrams and Signal-Flow Graphs

    block diagramsblock diagrams of control systems

    block diagrams of multivariable systems

    signal-flow graphs (SFGs)basic properties of SFG

    SFG algebra

    gain formula for SFG

    MATL AB tools and case studies

    3 Block Diagrams and Signal-Flow Graphs 2

    Block diagrams

    is a simple pictorial representation of a system basic language of control engineers describescompositionandinterconnectionof a system

    describes the cause-and-effect relationships throughout the system

    Fig. 3-1: (a) Block

    diagram of a dc-motor

    control system.

    (b) Block diagram with

    transfer functions and

    amplifier characteristics.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    15/116

    3 Block Diagrams and Signal-Flow Graphs 3

    Block diagrams of control systems

    Basic block-diagram elements are

    blocksrepresent system components, and lines with arrowrepresent signals and showhow signals flow from one block to another.

    Y(s) =G(s)U(s)

    G(s) Y(s)U(s)

    Some special blocks are usedto represent sensing devices

    that perform simplemathematical operations suchasadditionandsubtraction.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 4

    Drawing a block diagram

    Consider a voltage divider as shown in the circuit below.

    + +

    R1

    Vi Vo

    R2I

    Vo= R2I R2VoI

    I=Vi Vo

    R1

    + 1R1

    Vi I

    Vo

    Combining both equations, we have

    +Vi I

    R2Vo1

    R1 or Vo

    Vi=

    R2R1+R2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    16/116

    3 Block Diagrams and Signal-Flow Graphs 5

    Basic block diagram of a feedback control system

    r(t), R(s) =reference input (command)

    y(t), Y(s) =output (controlled variable)

    b(t), B(s) =feedback signalu(t), U(s) =actuating signal

    =error signale(t), E(s), ifH(s) = 1

    H(s) =feedback transfer function

    From the figure above, we can write

    Y(s) =G(s)U(s), B(s) =H(s)Y(s), and U(s) =R(s) B(s)Solving these forY(s)in terms ofR(s), we have

    Y(s)

    R(s)=

    G(s)

    1 +G(s)H(s)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 6

    Block diagrams of multivariable systems

    Y(s) = M(s)R(s)

    M(s) = [I + G(s)H(s)]1G(s)= G(s)[I + H(s)G(s)]1

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    17/116

    3 Block Diagrams and Signal-Flow Graphs 7

    Block diagram algebra: Combining two blocks

    original diagram equivalent diagram

    uG1

    yG2

    u yG2G1

    +u y

    G2

    G1u y

    G1 G2

    +u y

    G1

    G2

    u yG1

    1 G1G2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 8

    Block diagram algebra: Moving one block

    +u1

    Gy

    u2

    +

    u1 y

    G

    u2G

    +u1 y

    G

    u2

    +u1 y

    G

    u21

    G

    u yG

    y

    u yG

    Gy

    u yG

    u

    u yG

    1

    G

    u

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    18/116

    3 Block Diagrams and Signal-Flow Graphs 9

    Block diagram reduction

    Example Consider a multi-loop feedback control system as shownbelow

    +

    + + +

    G3 G4G2G1

    H2

    H1

    H3

    YR

    Answer:Y

    R=

    G4G3G2G11 G4G3H1+G3G2H2+G4G3G2G1H3

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 10

    Reduction difficulty?

    Wrong way: move the summing point behind the branching point

    +

    + +

    +Y

    G2

    G4

    G1

    G3

    R

    Right way:

    +

    + +

    +Y

    G2

    G4

    G3

    R

    G1

    G1

    Answer: Y

    R=G2G1+G4G1+G2G3

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    19/116

    3 Block Diagrams and Signal-Flow Graphs 11

    Signal-flow graphs (SFGs)

    may be regarded as a simplified version of a block diagram

    developed by S. J. Mason in 1950s

    can be used to determine relationship between two signals incomplex systems

    a small circle callednodedepicts a signal (line in block diagram) and a line calledbranchwhose arrow and number show direction

    and gain (multiplier) for signal traveling from one node to another

    G(s)

    Y(s)U(s)

    G(s)

    Y(s)U(s)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 12

    Signal-flow graph: Examples

    Example 1 The voltage divider can be viewed as aunity negativefeedback whose SFG graph is shown below.

    VoI

    R21

    R1

    11

    Vi

    Example 2

    R 1 G1 G2 G3 G4

    H2

    H3

    H1

    Y

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    20/116

    3 Block Diagrams and Signal-Flow Graphs 13

    Definition of SFG terms

    Aninput node (source)is a node that has only outgoing branches. Anoutput node (sink)is a node that has only incoming branches. Apathis any collection of a continuous succession of branches

    traversed in the same direction.

    Aforward pathis a path that starts at an input node and ends at anoutput node, and along which no node is traversed more than once.

    Aloopis a path that originates and terminates on the same node andalong which no node is traversed more than once.

    Apath gainor aloop gainis the product of the branch gainsencountered in traversing that path or loop.

    Two parts of an SFG arenontouchingif they do not share a commonnode.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 14

    Gain formula for SFG

    The gain between the input nodeyinand output nodeyoutis

    M=yout

    yin=

    1

    Nk=1

    Mkk

    where = 1 (all individualloops)

    +

    (gain products of all combinations oftwonontouching loops)

    (gain products of all combinations ofthreenontouching loops)+

    (. . . fournontouching loops) (. . . fivenontouching loops) + N = total number of forward paths betweenyinandyout

    Mk = the gain products of thekth forward path

    k = thefor that part of the SFG that is nontouching with thekth forward path

    Theis also known as thedeterminantof SFG.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    21/116

    3 Block Diagrams and Signal-Flow Graphs 15

    Gain formula: An example

    r 1 a b c d

    e

    gh

    f

    y

    k

    l

    has 7 loops and 4 paths:

    Loop: L1= l, L2= bh, L3= dg, L4= abck, L5= eck, L6= abfgk, L7= efgkPath: P1=abcd, P2= abf, P3= ecd, P4=efNontouching loops: L1L3, L2L3, L1L5, L1L7

    = 1 l bh dg abck eck abfgk efgk+ldg +bhdg+leck+lefgk

    y

    r =

    1

    (abcd+abf+ecd(1 l) +ef(1 l))

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 16

    Gain formula: Another example

    Y(s)

    R(s)=

    G1G2G3+G1G41 +G1G2H1+G2G3H2+G1G2G3+G4H2+G1G4

    , Y(s)

    E(s)=?

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    22/116

    3 Block Diagrams and Signal-Flow Graphs 17

    MATL ABtools and case studies (1)

    Transfer functions can be easily handled using Control System Toolbox.

    >> s=tf(s);

    >> G1=1/(s+1);>> G2=tf(1,[1 2])

    Transfer function:1

    -----s + 2

    >> G1*G2

    Transfer function:1

    -------------

    s ^ 2 + 3 s + 2

    >> G3=feedback(G1,G2)

    Transfer function:s + 2

    -------------s ^ 2 + 3 s + 3

    >> G4=G1/(1+G2*G1)

    Transfer function:s ^ 2 + 3 s + 2

    ---------------------s ^ 3 + 4 s ^ 2 + 6 s + 3

    The commandtfis used to specify the model in transfer function form.Then normally arithmetics +, -, *, and / can be used with transferfunctions.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 18

    MATL ABtools and case studies (2)

    Commandsseries,parallel,feedbackcan be used to make morecomplicated models.

    >> zpk(G4)

    Zero/pole/gain:(s+2) (s+1)

    ---------------------(s+1) (s^2 + 3s + 3)

    >> minreal(G4)

    Transfer function:s + 2

    -------------s ^ 2 + 3 s + 3

    The commandzpkis used to specify the model in zero-pole-gain formwhile the commandminrealcan be used to simplify model by pole-zerocancellation.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    23/116

    3 Block Diagrams and Signal-Flow Graphs 19

    MATL ABtools and case studies (3)

    More complicated connection can be done using commandsappendandconnect.

    >> G5=append(G1,G2);>> connect(G5,[1 -2;2 1],1,1)

    Transfer function:s + 2

    -------------s ^ 2 + 3 s + 3

    Theappend(G1,G2)command creates the structure y(1) = G1 u(1)andy(2) = G2 u(2)while theconnectcommand has the syntax

    sys = connect(blksys,Q,input,output)

    Theindex-based interconnection matrixQ=[1 -2;2 1]feeds-y(2)intou(1)andy(1)intou(2)andinput=1andoutput=2indicate thatrdrivesu(1)andyisy(1).

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    3 Block Diagrams and Signal-Flow Graphs 20

    MATL ABtools and case studies (4)

    Multivariable systems can also be handled.

    >> G=[1/(s+1),-1/s;2,1/(s+2)]; H=[1 0; 0 1];>> feedback(G,H)

    Transfer function from input 1 to output...

    3 s^2 + 9 s + 4#1: ----------------------s ^ 3 + 7 s ^ 2 + 1 2 s + 4

    2 s ^ 3 + 6 s ^ 2 + 4 s#2: ----------------------

    s ^ 3 + 7 s ^ 2 + 1 2 s + 4

    Transfer function from input 2 to output...- s ^ 2 - 3 s - 2

    #1: ----------------------s ^ 3 + 7 s ^ 2 + 1 2 s + 4

    3 s^2 + 8 s + 4#2: ----------------------

    s ^ 3 + 7 s ^ 2 + 1 2 s + 4

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    24/116

    4 Modeling of Physical Systems

    modeling of electrical networks modeling of mechanical systems elements

    translational motion & rotational motion

    friction, backlash and dead zone (nonlinear characteristics)

    equations of mechanical systems sensors and encoders in control systems DC motors in control systems linearization of nonlinear systems systems with transportation lags (time delays) MATL AB tools and case studies

    4 Modeling of Physical Systems 2

    Introduction

    two most common models: transfer function and state-variable transfer functions are valid only for linear time-invariant systems

    state equations can be applied to linear as well as nonlinear systems analysis and design for nonlinear systems are usually quite complex linear models are only approximations must determine how to accurately describe a system mathematically

    and how to make proper assumptions so that the system may berealistically characterized by a linear mathematical model

    the objective of this chapter is to demonstrate mathematicalmodeling of control systems and components

    the coverage here is by no means exhaustive

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    25/116

    4 Modeling of Physical Systems 3

    Ideal v.s. real

    Example 1 A resistor V

    I

    +

    ideal: Ohms law V =I R real: over voltage and current, parasitic capacitance and inductance

    Example 2 A mechanical spring

    ideal: F =kxmasslessfrictionless

    real:

    F

    F

    x

    slope= k

    free length

    solidheight

    What is real is that everything is subject to certainconditions.To describe it, we need someassumptions.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 4

    Modeling of electrical systems elements

    component voltage-current current-voltageC

    capacitore(t) =

    1

    C

    t0

    i() d i(t) =Cd

    dte(t)

    R

    resistore(t) =Ri(t) i(t) =

    1

    Re(t)

    L

    inductore(t) =L

    d

    dti(t) i(t) =

    1

    L

    t0

    e() d

    Capacitor current-charge: i(t) = d

    dtq(t)

    Inductor voltage-flux: e(t) = d

    dt(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    26/116

    4 Modeling of Physical Systems 5

    Modeling of electrical networks

    Use Kirchhoffs voltage and current laws: loop and node methods.

    Usingi1(t),i2(t), andec(t)as states, we can write

    L1di1dt

    =R1i1 ec+e

    L2di2dt

    =R2i2+ec

    Cdecdt

    =i1 i2

    =L1L2Cs3 + (R1L2+R2L1)Cs

    2 + (L1+L2+R1R2C)s+R1R2

    I1(s)

    E(s)=

    L2Cs2 +R2Cs+ 1

    ,

    I2(s)

    E(s)=

    1

    ,

    Ec(s)

    E(s) =

    L2s+R2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 6

    Operational amplifiers

    Operational amplifiers or OpAmps areactivecomponents which can beused inrealizationof wider class of transfer functions.

    +

    +

    C R1

    R2

    R3

    R

    +

    EO(s)

    +

    EI(s)

    EO(s)

    EI(s) =

    R3R1R2R

    (RCs+ 1)

    Note that this transfer function isimproper.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    27/116

    4 Modeling of Physical Systems 7

    Modeling of mechanical systems elements: Translational

    component force-displacement force-velocityk

    springf(t) =kx(t) f(t) =k

    t0

    v() d

    b

    damperf(t) =b

    d

    dtx(t) f(t) =bv(t)

    m

    massf(t) =m

    d2

    dt2x(t) f(t) =m

    d

    dtv(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 8

    Modeling of mechanical systems elements: Rotational

    component torque-angulardisplacementtorque-

    angular velocityK

    springT(t) =K (t) T(t) =K

    t0

    () d

    B

    damper

    T(t) =Bd

    dt(t) T(t) =B(t)

    J

    inertia

    T(t) =Jd2

    dt2(t) T(t) =J

    d

    dt(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    28/116

    4 Modeling of Physical Systems 9

    Modeling of friction

    Viscous frictionis a linear relationship between force and velocity.

    Static frictiontends to prevent motion from the beginning. Once the

    motion begins, static friction vanishes and other frictions take over.Coulomb frictionis a retarding force having a constant amplitude with

    respect to the direction of velocity.

    f(t) =Bdx(t)

    dt f(t) =(Fs)|x=0 f(t) =Fc

    dx(t)

    dt

    dx(t)dt

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 10

    Conversion between translational and rotational motions

    The mechanisms ofbelt-and-pulleyand rack-and-pinioncan be used tochange between translational and rotational motions. LetW =M g.

    Belt and pulley Ifris the

    radius of the pulley. Then theequivalent inertiathat the motorsees is

    Jeq= M r2

    Rack and pinion IfLis thescrew lead, i.e., the lineardistance thatMtravels perrevolution of the screw. Then

    Jeq= M L

    2

    2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    29/116

    4 Modeling of Physical Systems 11

    Gear trains

    IfTi,i, andNi(i= 1,2) are torque applied,angular displacement, and teeth number of

    the geari, then we haver1r2

    =N1N2

    , 1r1= 2r2, T11= T22

    by considering the teeth ratio, distance alongsurface, and work done, respectively.

    In practice, considering inertia loads andfrictions at both sides, the equivalenttorque seeing from gear 1 is given as

    T =J1ed21dt2

    +B1ed1dt

    +Fc11|1| +

    N1N2

    Fc12|2|

    where J1e= J1+

    N1N2

    2J2, B1e= B1+

    N1N2

    2B2.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 12

    Backlash, dead zone, and saturation

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    30/116

    4 Modeling of Physical Systems 13

    Equations of mechanical systems: Translational

    x(t)

    forcef(t)

    bfriction

    k

    mmass x(t)

    f(t)

    bx kx

    m

    Applying Newtons law of motionto the free-body diagram, we have

    f(t) =md2

    dt2x(t) +b

    d

    dtx(t) +kx(t)

    Assuming zero initial conditions

    (ms2 +bs+k)X(s) =F(s) = X(s)F(s)

    = 1

    ms2 +bs+k

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 14

    What about the gravity?

    b

    k

    x(t)

    f(t)

    y(t)

    x0

    m

    Whenf(t) = 0the spring stretches by theweight of the mass (mg) alone to the length

    x0=mg

    k

    Let y(t)be the displacement measured fromthe free length of the spring, and

    x(t)be the displacement measured fromx0= mg/k

    Now we havey(t) =x0+x(t). With reference from the free length

    f(t) +mg =md2

    dt2y(t) +b

    d

    dty(t) +ky(t)

    However, fromy(t) =x0+x(t)

    y(t) = x(t) and y(t) = x(t) = f(t) =md2

    dt2x(t) +b

    d

    dtx(t) +kx(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    31/116

    4 Modeling of Physical Systems 15

    Equations of mechanical systems: Rotational

    K BJ

    (t)T(t)

    From Newtons law of motion,

    T(t) =Jd2

    dt2(t) +B

    d

    dt(t) +K(t)

    Assuming zero initial conditions,

    T(s) = (Js2 +Bs+K)(s)(s)

    T(s) =

    1

    Js2 +Bs+K

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 16

    Sensors and encoders in control systems (1)

    Potentiometer mechanical displacement electrical voltage

    e(t) =Ksc(t) or e(t) =Ks[1(t) 2(t)]c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    32/116

    4 Modeling of Physical Systems 17

    Sensors and encoders in control systems (2)

    Tachometer mechanical velocity electrical voltage

    et(t) =Ktd(t)

    dt =Kt(t) = Et(s)

    (s) =Kts

    Incremental encoder mechanical movement electrical pulse

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 18

    DC motors in control systems

    The dc motor is a torquetransducerconverting electrical energy tomechanical energy.

    The relationship among the developedtorqueTm, the flux, and the armaturecurrentiais

    Tm= KmiawhereKmis a proportional constant.

    When motor rotates, the voltage acrossits terminals calledback emf ebisdeveloped

    eb=Kmm

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    33/116

    4 Modeling of Physical Systems 19

    Mathematical modeling of PM DC motors (1)

    Consider a dc motor system with the following variables andparameters:

    = magnetic flux in the air gapJm = rotor inertia

    Bm = viscous-friction coefficientm(t) = rotor displacementm(t) = rotor angular velocity

    Ra = armature resistanceLa = armature inductanceKi = torque constantKb = back-emf constant

    ia(t) = armature currenteb(t) = back emf

    TL(t) = load torqueTm(t) = motor torqueea(t) = applied voltage

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 20

    Mathematical modeling of PM DC motors (2)

    ea(t) eb(t) =Ladia(t)dt

    +Raia(t)

    Tm(t) =Kiia(t) [ :=Kmia(t)], eb(t) =Kbdm(t)

    dt =Kbm(t)

    Tm(t) TL(t) =Jmd2m(t)dt2

    +Bmdm(t)dt

    m(s)

    Ea(s)

    TL=0

    = Ki

    LaJms3 + (RaJm+BmLa)s2 + (KbKi+RaBm)s

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    34/116

    4 Modeling of Physical Systems 21

    Linearization of nonlinear systems

    Consider a nonlinear system

    x(t) = f(x(t), r(t)) where x(t) =

    x1(t)

    x2(t)...xn(t)

    , r(t) = r1(t)

    r2(t)...rp(t)

    , f(, ) = f1(, )f2(, )...fn(, )

    The linearization of the system about an operating condition(x0, r0),e.g., f(x0, r0) = 0, is given by

    x(t) = Ax(t) + Br(t) where x(t) = x(t) x0, r(t) = r(t) r0and

    A= [aij] where aij =

    fi

    xjx0,r0 and

    B= [bij] where bij =

    fi

    rjx0,r0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 22

    Magnetic-ball-suspension system: Nonlinear model

    Consider a magnetic-ball-suspension system

    e(t) =input voltagei(t) =winding currenty(t) =ball position

    L=winding inductanceR=winding resistance

    M=mass of ball

    Md2y(t)

    dt2 =M g i2(t)

    y(t) , e(t) =Ri(t) +Ldi(t)

    dt

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    35/116

    4 Modeling of Physical Systems 23

    Magnetic-ball-suspension system: Linearization

    Define the state variables as

    x1(t) =y(t)

    y0, x2(t) = y(t), and x3(t) =i(t)

    i0

    around the nominal positiony0= constant, we havei0=

    M gx01andthe linearized model is

    d

    dt

    x1(t)x2(t)

    i(t)

    =

    0 1 0g

    y002

    g

    M y0

    0 0 RL

    x1(t)x2(t)

    i(t)

    +

    001

    L

    e(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    4 Modeling of Physical Systems 24

    Systems with transportation lags (time delays)

    There are many ways to approximate time delays using Maclaurin series

    eTds = 1 Tds+12

    T2d s2

    or

    eTds = 11 +Tds+T2d s2/2

    A better approximation is to use the Pade approximation

    eTds =1 Tds/21 +Tds/2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    36/116

    5 Poles and Zeros of Transfer Functions

    poles and zeros residue and system response first- and second-order systems pole-zero cancellation

    MATL AB tools and case studies

    5 Poles and Zeros of Transfer Functions 2

    Linear time-invariant systems

    Consider a linear constant coefficient ODE

    y(n)(t) +1y(n1)(t) +2y(n2)(t) + +n1y(t) +ny(t) =f(t)

    whose solution can be found substitutingy(t) =emt.

    If the coefficientsiare not constant, but depend on time asi(t), wehave alinear time-varyingsystem.

    If the coefficientsidepend onyasi(y, y,...,y(n1)), we have anonlinear time-invariantsystem.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    37/116

    5 Poles and Zeros of Transfer Functions 3

    Time-domain response

    The time-domain response is given by

    y(t) = k=1

    ck

    emkt +Yp

    (t) =c1em1t +c

    2em2t +

    +c

    nemnt +Y

    p(t)

    wherem1,m2, . . . ,mnare (real or complex) roots of the characteristics(orauxiliary) equation

    mn +1mn1 +2mn2 + +n1m+n= 0

    andYp(t)is the particular solution depending onf(t).

    If the initial conditionsy(0),y(0), . . . ,y(n1)(0)are given, we candeterminec1,c2, . . . ,cn.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    5 Poles and Zeros of Transfer Functions 4

    Poles and zeros

    In more general cases, the forcing functionf(t)is related to the inputu(t)by

    f(t) =1u(n1)(t) +2u(n2)(t) + +n1u(t) +nu(t)

    or

    y(n) +1y(n1) + +n1y+ny=1u(n1) + +n1u+nu

    and the transfer function is given by

    G(s) = 1s

    n1 +2sn2 + +n1s+nsn +1sn1 +2sn2 + +n1s+n =:

    n(s)

    d(s)

    We call the roots ofn(s) = 0zeros andd(s) = 0poles ofG(s).

    Example G(s) = 26.25(s+ 4)

    s(s+ 3.5)(s + 5)(s+ 6)

    has a zero at 4and four poles at0, 3.5, 5, and 6.c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    38/116

    5 Poles and Zeros of Transfer Functions 5

    Residues and system response

    The system response can be found fromG(s)by taking the inverseLaplace transform. To do that, we need to find the partial fraction

    G(s) = Ai

    s+pi+ Bk

    s+k+jk+

    Bks+k jk

    In complex analysis,AiandBkare known as theresidueofG(s)ats= piand ats= k jk, respectively.In this case, the system response is given by

    g(t) =

    Aiepit +

    ekt [2Re(Bk)cos kt+ 2Im(Bk)sin kt]

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    5 Poles and Zeros of Transfer Functions 6

    First- and second-order systems

    Hence, many system responses can be considered as summation of thefirst-order system

    G1(s) = A

    s+p

    and the second-order system

    G2(s) = K2

    s2 + 2s+2

    We will consider the behavior of their system responses in more detail.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    39/116

    5 Poles and Zeros of Transfer Functions 7

    Pole-zero cancellation

    Consider, for example, two transfer functions with their partial fractions

    C1(s) =

    26.25(s+ 4)

    s(s+ 3.5)(s + 5)(s+ 6) =

    1

    s 3.5

    s+ 5+

    3.5

    s+ 6 1

    s+ 3.5

    and

    C2(s) = 26.25(s+ 4)

    s(s+ 4.01)(s+ 5)(s+ 6)=

    0.87

    s 5.3

    s+ 5+

    4.4

    s+ 6+

    0.033

    s+ 4.01

    We can see that the residue of the pole at 4.01, which is closer to thezero at 4, is smaller. Hence we can approximateC2(s)by neglecting theresponse corresponding to the pole at 4.01with the second-order

    C2(s) 0.87

    s 5.3

    s+ 5+ 4.4

    s+ 6

    orc2(t) 0.87 5.3e5t + 4.4e6t.However, the unstable polecannotbe cancelled by the unstable zero.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

    5 Poles and Zeros of Transfer Functions 8

    MATLABtools and case studies

    Use commandsresidue,roots,pole, andzero.

    >> [r,p,k]=residue(26.25*[1 4],poly([0,-4.01,-5,-6]))

    r = p = k =

    4.3970 -6.0000 []-5.3030 -5.0000

    0.0332 -4.01000.8728 0

    >> p=[1 7 9 23 10]>> roots(p)>> s=tf(s); G=(s^2-2*s+2)/(s^4+7*s^3+19*s^2+23*s+10)>> pole(G)>> zero(G)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi,Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    40/116

    6 Stability of Linear Control Systems

    bounded-input bounded-output (BIBO) stability

    relationship between charactersitic equation roots and stability

    zero-input and asymptotic stability

    Routh-Hurwitz criterion

    Rouths tabulation

    special cases when Rouths tabulation terminates prematurely

    MATL ABtools and case studies

    6 Stability of Linear Control Systems 2

    Introduction

    absolute stability v.s. relative stability

    Absolute stability refers to the condition whether the system isstable or unstable; it is ayesornoanswer.

    Once the system is found to be stable, to determine how stable it

    is, we need a measure of relative stability.

    zero-state v.s. zero-input responses

    total response= zero-state response+zero-input response

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    41/116

    6 Stability of Linear Control Systems 3

    Bounded-input bounded-output stability

    y(t)u(t)

    G(s)

    g(t)

    U(s) Y(s)

    Definition With zero initial conditions, thesystem is said to be(BIBO) stableif its output

    y(t)is bounded to any bounded inputu(t).

    From the convolution integral relatingu(t),y(t), andg(t)

    y(t) =

    0

    u(t )g() d

    we have

    |y(t)| =

    0

    u(t )g() d

    0

    |u(t )| |g()| d

    Suppose the input is bounded, i.e., |u(t)| Mfor some positiveM, thenthe output is bounded if

    0

    |g(t)| dt Q <

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 4

    Relationship between charactersitic eqn. roots & stability

    The transfer functionG(s), by the Laplace transform definition, is

    G(s) = L[g(t)] =

    0

    g(t)est dt

    For anys= s0= 0+j0, we have

    |G(s0)| =

    0

    g(t)es0t dt

    0

    |g(t)|es0t dt=

    0

    |g(t)|e0t dt

    since |es0t| = |e0t|. If there is a poles0=0+j0ofG(s)with0 0,using the fact that |e0t| 1for allt 0,

    = |G(s0)|

    0

    |g(t)|e0t dt

    0

    |g(t)| dt

    which violates the BIBO stability requirement. The system isunstable.

    Thusfor a stable systems, the roots of the characteristic equation, orthe poles ofG(s)cannot be located in the closed right half s-plane.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    42/116

    6 Stability of Linear Control Systems 5

    Zero-input and asymptotic stability

    Let the input of an nth-order system be zero, the output due to the initialcondition can be expressed as

    y(t) =n1k=0

    y(k)(t0)gk(t) =y(t0)g0(t) +y(t0)g1(t) +. . .+y

    (n1)(t0)gn1(t)

    If the zero-input responsey(t), subject to the finite initial conditionsy(k)(t0)reaches zero astapproaches infinity, the system is said to bezero-input stableorasymptotic stable.

    Let thencharacteristic equation roots be expressed assi=i+ji,i= 1, 2, . . . , n.

    y(t) =m

    i=1

    Kiesit +nm1

    i=0

    Litiesit

    Fory(t)to approach0ast , the real parts ofsimust be negative.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 6

    Stability for linear time-invariant systems

    Hence, for linear time-invariantsystems, BIBO stability, zero-input

    stability, and asymptotic stability allhave the same requirementthat theroots of the characteristic equationmust all be located in the left-halfs-plane.

    If the characteristic equation hassimple roots on thej-axis and nonein the right-half plane, we say that thesystem ismarginally stable.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    43/116

    6 Stability of Linear Control Systems 7

    Routh-Hurwitz criterion

    The Routh-Hurwitz criterion is a method of determining the location ofzeros of a polynomial whether they are in the left- or right-half planes.

    Consider that the characteristic equation of a linear time-invariant SISOsystem is of the form

    F(s) =ansn +an1s

    n1 + +a1s +a0= 0

    In order thatF(s)does not have roots with non-negative real parts, it isnecessary (but not sufficient)that

    1. All the coefficientsa0,a1, . . . ,an1,anhave the same sign.

    2. None of the coefficients vanishes.

    an(s z1)(s z2) (s zn) =ansn an(

    zi) s

    n1 + +an(1)n zi

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 8

    Rouths tabulation or Rouths array (1)

    sn p11= an p12= an2 p13= an4 p14= an6 sn1 p21= an1 p22= an3 p23= an5 p24= an7 sn2 p31 p32 p33 p34 sn3 p41 p42 p43 p44 ...

    ...

    ...s0 pn+1,1

    where

    p31= 1

    p21

    p11 p12p21 p22 , p32= 1p21

    p11 p13p21 p23 , p33= 1p21

    p11 p14p21 p24 ,

    p41= 1

    p31

    p21 p22p31 p32 , p42= 1p31

    p21 p23p31 p33 , p43= 1p31

    p21 p24p31 p34 ,

    ...

    The roots are all in the LHP ifp11,p21,p31, . . . ,pn+1,1have the same sign.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    44/116

    6 Stability of Linear Control Systems 9

    Rouths tabulation or Rouths array (2)

    The number of changes of signs in the elements of the first columnsequals the number of roots with positive real parts or in the RHP.

    Example 1 P(s) =s4 + 2s3 + 3s2 + 4s + 5 = 0

    s4 1 3 5s3 2 4 0s2 23142 = 1

    25102 = 5

    s1 14251 = 6 0s0 5

    There are two sign changes: 1 6 and 6 5. HenceP(s)has two

    RHP roots.

    Example 2 2s4 +s3 + 3s2 + 5s + 10 = 0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 10

    Proof of Rouths criterion (1)

    Supposenis even. The polynomialP(s) =ans

    n +an1sn1 + +a1s +a0can be divided into real and

    imaginary parts asP(s) =P1(s) +P2(s)where

    Even part: P1(s) = ansn +an2s

    n2 + +a2s2 +a0

    Odd part: P2(s) = an1sn1 +an3sn3 + +a3s3 +a1s

    Definition If two polynomials have the same numbers of roots in LHP,RHP, and onj-axis, we say that they areequivalent.

    Fact 1 Ifj0is a root ofP(s), then it is also a root ofP1(s)andP2(s).

    Fact 2 For real constantnear zero,P(s)is equivalent to (as long as ithasndegree)Q(s, ) :=P(s) sP2(s) = (an an1)s

    n +an1sn1 + .

    Proof: Note thatQ(s, 0) =P(s). Now from Fact 1,P(s)and

    Q(s, ) =P1(s) +P2(s) sP2(s) = [P1(s) sP2(s)] +P2(s)

    have the same number ofj-axis roots for small.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    45/116

    6 Stability of Linear Control Systems 11

    Proof of Rouths criterion (2)

    Fact 3 If anan1

    ,Q(s, )drops degree as one of its roots approaches

    an1an an1

    =

    an

    an1

    1

    Now we can prove the Rouths criterion by noting that

    sum of the first and second rows of Rouths array isP(s),

    sum of the second and third rows isQ(s, an/an1).

    Whenan1= 0, the construction of Rouths array fails. To work aroundthis problem, we use the following result.

    Fact 4 If(s)is a polynomial such that(j)> 0for all, anddeg P2< deg P1, thenP1(s) +P2(s)is equivalent toP1(s) +(s)P2(s).Proof: LetQ(s, ) =P1+ [(1 ) +]P2where0 1. We can seethatQ(s, 0) =P1+P2,Q(s, 1) =P1+P2, and(1 ) + >0for all.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 12

    Rouths array: Special case 1

    Case 1 Somep,1= 0but there is a nonzero memberp,kin the row.

    Let(s) = 1 + (s2)k1 and use Fact 4.

    Example 1 p(s) =s5 + 2s4 + 3s3 + 6s2 + 5s + 3

    s5

    1 3 5s4 2 6 3s3 0 7

    P2(s) = 7s P2(s) =s

    Let(s) = 1 s2.

    (s)P2(s) = s3 +ss

    3 1 1s2 8 3s1 118s0 3

    There are two sign changes inthe first column sop(s)hastwo RHP roots.

    Example 2 s4 +s3 + 2s2 + 2s + 3 = 0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    46/116

    6 Stability of Linear Control Systems 13

    Rouths array: Special case 2 (1)

    Case 2 The entire row is zero.

    ThenP(s)is even or odd and is also a factor of the original polynomial.

    Fact 5 Even or odd polynomials have as many roots in LHP as roots inRHP. Proof: Ifsis a root, then sis also a root.

    To remedy the situation, considerP(s) =P(s+)for some small >0.

    P(s) = (P+2

    2!P + )

    even

    + (P +2

    3!P + )

    odd

    is equivalent to

    (P+22!

    P + ) + (P +23!

    P + ) P(s) +P(s) as 0

    In this case, substituteP(s)withP(s) +P(s)and count only RHP roots.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    6 Stability of Linear Control Systems 14

    Rouths array: Special case 2 (2)

    Example P(s) =s5 +s4 + 6s3 + 6s2 + 8s + 8

    s5 1 6 8s4 1 6 8s3 0 0

    P(s) =P1(s) =s4 + 6s2 + 8

    P(s) = 4s3 + 12s

    Use 14(4s3 + 12s) =s3 + 3s.s3

    1 3s2 3 8s1 13s1 1s0 8

    Hence,P(s)has one (1)LHP root and four (4)

    j-axis roots.

    Remark For design purposes, we can use the all-zero-row condition tosolve for the parameter that causes the systems marginal stability.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    47/116

    6 Stability of Linear Control Systems 15

    A design example and relative stability

    Consider a negative feedback system shown in the block diagram below.

    Given

    G(s) = s2 2s + 2

    s4 + 7s3 + 19s2 + 23s + 10

    Find ranges ofKsuch that

    1. the closed-loop system is stable, and

    +G(s)

    K

    yu

    2. all the closed-loop system poles have real part less than 1.

    Solution The characteristic equation is1 +KG(s) = 0or

    s4 + 7s3 + 19s2 + 23s + 10 +K(s2 2s + 2) = 0

    s

    4

    + 7s

    3

    + (19 +K)s

    2

    + (23 2K)s+ 10 + 2K = 0The regionRe s < 1is equivalent toRe (s + 1)< 0. So lets= s + 1ors= s 1and findKsuch that1 +KG(s) = 1 +KG(s 1) = 0is stable inthe sense thatRe s

  • 7/21/2019 Mws Slide

    48/116

    7 Time-Domain Analysis of Control Systems

    time response and test signals for continuous-data control systems the unit-step response and time-domain specification steady-state error: unity feedback v.s. nonunity feedback time response of a first-order system transient response of a prototype second-order system

    damping ratio and damping factor, natural undamped frequency

    maximum overshoot, delay time, rise time, and settling time

    time-domain analysis of a position-control system dominant poles of transfer functions MATL ABtools and case studies

    7 Time-Domain Analysis of Control Systems 2

    Time response and typical test signals

    Time response of a control system can be divided into two parts:

    transient response: yt(t) goes to zero ast , i.e., limt

    yt(t) = 0

    steady-state response: yss(t) is what remains afteryt(t)has died out

    Typical test signals

    step-function input:r(t) =Rus(t)

    ramp-function input:r(t) =Rtus(t)

    parabolic-function input:r(t) = 1

    2

    Rt2us(t)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    49/116

    7 Time-Domain Analysis of Control Systems 3

    Unit-step response and time-domain specification

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 4

    Time-domain specification: Definition

    1.Maximum overshoot=ymax yssis used to measure relative stability.

    percent maximum overshoot=maximum overshoot

    yss 100%

    2.Delay time,td, is defined as the time required for the step response toreach50%of its final value.

    3.Rise time,tr, is defined as the time required for the step response torise from10%to90%of its final value.

    4.Settling time,ts, is defined as the time required for the step responseto decrease andstay withina specified percentage of its final value.

    5.Steady-state error,ess, is defined as the difference between theoutput and the reference input when the steady state (t ) isreached.

    Here we assume that the system has unityd.c. gain, i.e.,Gc(0) = 1.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    50/116

    7 Time-Domain Analysis of Control Systems 5

    Steady state error: Unity feedback

    Consider a feedback system withclosed-loop transfer function

    M(s) =Y(s)

    R(s)=

    G(s)

    1 +G(s)H(s)

    + G(s)

    H(s)

    R(s) U(s) Y(s)

    IfH(s) = 1, we have aunity feedbacksystem and define the error of thesystem ase(t) =r(t) y(t). Hence the steady-state error is given by

    ess= limt

    e(t) = lims0

    sE(s) = lims0

    sR(s)

    1 +G(s)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 6

    System types and error constants

    An open-loop transfer function

    G(s) =K(Tas+ 1)(Tbs+ 1) sn(T1s+ 1)(T2s+ 1)

    is called oftypenwhich is the order of the pole ofG(s)ats= 0.

    The steady-state error due to various kinds of inputs is shown below:

    Step (position) error constant:Kp= lim

    s0G(s)

    Ramp (velocity) error constant:Kv = lim

    s0sG(s)

    Parabolic (acceleration) error constant:

    Ka= lims0 s2

    G(s)

    type step ramp parabolic

    0 R

    1 +Kp

    1 0 R

    Kv

    2 0 0

    R

    Ka

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    51/116

    7 Time-Domain Analysis of Control Systems 7

    Steady state error: Unity feedback examples

    GivenH(s) = 1, determine system type ofG(s)and findessfor three basictypes of inputs.

    G(s) = K(s+ 3.15)s(s+ 1.5)(s+ 0.5)

    G(s) = Ks2(s+ 12)

    G(s) = 5(s+ 1)s2(s+ 12)(s+ 5)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 8

    Steady state error: Nonunity feedback

    IfH(s)is nonunity, the outputY(s)and the reference signalR(s)havedifferent units. So we cannot define the error byr(t) y(t). Differentdefinitions oferrorare needed.

    Case H(0) = 0 Thed.c. gainof sensorH(s)is given byKH=H(0).

    E(s) = 1KH

    R(s) Y(s)

    Case H(0) = 0 orH(s)has zero(s) at the origin, i.e.,H(s) =sNH1(s).

    E(s) = 1

    KHsNR(s) Y(s) where KH= lim

    s0H(s)

    sN =H1(0)

    Examples G(s) = 1

    s2(s+ 12)and

    H(s) =5(s+ 1)s+ 5

    H(s) = 10ss+ 5

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    52/116

    7 Time-Domain Analysis of Control Systems 9

    Time response of a first-order system

    Consider a first-order system (T >0)

    G(s) =

    Y(s)

    R(s)=

    K

    T s+ 1

    Impulse response r(t) =(t) R= 1

    y(t) = L1[G(s)] = KT

    et/T

    whent 0 yss(t) = 00.0

    0.2

    0.4

    0.6

    0.8

    1.0

    0 1 2 3 4 5

    63.2%

    t/T

    y(t)/K

    Step response r(t) =u(t) R(s) = 1/s

    y(t) = L1 G(s)s

    = L1 K

    s(T s+ 1)

    = L1 K

    s KT

    T s+ 1

    = K(1 et/T)

    whent 0 yss(t) =K

    Notice that step response=

    impulse response

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 10

    A prototype second-order system

    Consider the closed-loop system shown inthe block diagram. We have

    Y(s)

    R(s)=

    K

    s2 +ps+K

    which is asecond-ordertransfer function.

    + K

    s(s+p)

    yr e

    We write this as acanonicalorprototypesecond-order transfer function

    n

    n

    n

    G(s) = 2n

    s2 + 2ns+2n

    where >0andn>0whose poles are

    s2 + 2ns+2n = s

    2 + 2ns+22n+ (1 2)2n

    = (s+n)2 +22

    n

    where2 = 1 2.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    53/116

    7 Time-Domain Analysis of Control Systems 11

    Transient response of a second-order system (1)

    Whenr(t) = 1 R(s) = 1/s, we have

    y(t) = L1 2n

    s(s2 + 2ns+2n)

    = L1 1

    s

    + 2 Re L1 B

    s+ (+ j)n

    where

    B = 2n

    s(s+ (j)n)

    s=(+j)n=

    j2(+j)

    Hence

    y(t) = 1 1

    ent sin(nt+)

    where= (+j) = cos1 andt 0. damping ratio natural (undamped) frequency n damping factor := n conditional (damped) frequency :=n

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 12

    Transient response of a second-order system (2)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    54/116

    7 Time-Domain Analysis of Control Systems 13

    Pole location and response

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 14

    Maximum overshoot

    peak time tmaxortptmax=

    n1 2

    =

    maximum overshoot Mp:=y(tmax) y()

    Mp= e /

    12

    percent (maximum) overshoot %Mpor P.O.

    P.O.=y(tmax) y()

    y() 100%

    P.O.= 100e /

    12

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    55/116

    7 Time-Domain Analysis of Control Systems 15

    Delay time and rise time

    Delay time 0<

  • 7/21/2019 Mws Slide

    56/116

    7 Time-Domain Analysis of Control Systems 17

    Effects of adding a pole to transfer function (1)

    Add a pole ats= 1/Tpto aforward-path transfer function of aunity-feedback system as

    G(s) = 2n

    s(s+ 2n)(1 +Tps)

    The closed-loop transfer functionM=G/(1 +G)becomes

    2n

    Tps3 + (1 + 2nTp)s2 + 2ns+2n

    = 1, n= 1 = 0.25, n= 1

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 18

    Effects of adding a pole to transfer function (2)

    Add a pole ats= 1/Tpto the closed-loop transfer function as

    M(s) =Y(s)

    R(s)=

    G(s)

    1 +G(s)=

    2n(s2 + 2ns+2n)(1 +Tps)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    57/116

    7 Time-Domain Analysis of Control Systems 19

    Effects of adding a zero to transfer function (1)

    Add a zero ats= 1/Tpto the closed-loop transfer function as

    M(s) = G(s)

    1 +G(s)= 2n(1 +Tps)

    (s2 + 2ns+2n) or y(t) =y1(t) +Tz

    dy1(t)

    dt

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 20

    Effects of adding a zero to transfer function (2)

    Consider, for example, adding a zero ats= 1/Tzto a forward-path of athird-order system as

    G(s) = 6(1 +Tzs)

    s(s+ 1)(s+ 2) so that

    Y(s)

    R(s)=

    6(1 +Tzs)

    s3 + 3s2 + (2 + 6Tz)s+ 6

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    58/116

    7 Time-Domain Analysis of Control Systems 21

    Non-mininum phase (NMP) zeros and undershoot

    IfG(s)has unity d.c. gain and a real NMP zero ats= z0, then we have

    e(t) = 1

    y(t)

    E(s) = (1

    G(s))1

    s E(z0) =

    1

    z0=

    0

    e(t)ez0t dt

    Assuming its step responsey(t)has a-settling timetsand splittingintegral into two parts[0, ts] (ts, ), we have ts

    0

    e(t)ez0t dt+

    ts

    |e(t)|ez0t dt 1z0

    IfMuis the minimum undershoot, we havemax

    t0e(t) =Emax= 1 +Mu> 0

    and from the definition of-settling time

    |e(t)

    | ,

    t

    ts, we have

    Emax1 ez0ts

    z0+

    ez0ts

    z0 1

    z0or Mu 1

    ez0ts 1If 1andz0ts 1, this meansMu> 1z0ts .

    fast settling time large undershoot

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 22

    Dominant poles

    For analysis and design, it isimportant to sort out polesthat have a dominant transient

    effect, calleddominant poles.

    Dominant poles controldynamic performance,

    whereasinsignificant polesensure realizable controller.

    It is widely accepted that if the magnitude of the real part of a pole isat least510times that of a dominant pole or a pair of complex

    dominant poles, then the pole may be regarded as insignificant.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    59/116

    7 Time-Domain Analysis of Control Systems 23

    Relative damping ratio

    When the dynamic of a system of higher-order can be accuratelyrepresented by a pair of complex-conjugate dominant poles, then we

    can still useandnto indicate the transient dynamics. The damping ratio in this case is referred to as the relative damping

    ratio.

    Example

    M(s) = 20

    (s+ 10)(s2 + 2s+ 2) 2

    s2 + 2s+ 2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    7 Time-Domain Analysis of Control Systems 24

    MATLABtools and case studies

    Use commandsstep,ltiviewandsisotool.SIMULINKis also useful.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    60/116

    8 Basic Control Actions

    introductiondesign specifications

    controller configurations

    PID controller

    proportional action

    integral action

    derivative action

    tuning of PID controllerclosed-loop method

    open-loop method

    8 Basic Control Actions 2

    Introduction

    Design specifications

    relative stability, steady-state accuracy (error), transient response:maximum overshoot, rise time, settling time

    frequency-response characteristics: gain margin, phase margin,Mr

    Controller configurations

    series (cascade), feedback, and feedforward compensations

    one degree-of-freedom, two degrees-of-freedom

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    61/116

    8 Basic Control Actions 3

    PID control

    PID controllers a.k.a. three-term or three-mode control

    one of the oldest controller types and the most widely used in

    industries

    pulp & paper 86%

    steel 93%

    oil refineries 93%

    only textbook version is introduced

    differs a lot from the industrial version

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 4

    Text-book PID controllers

    PID controller:

    Integral

    Proportional

    Derivative

    Plant

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    62/116

    8 Basic Control Actions 5

    PID controllers

    PID controller

    u(t) = KP

    e(t)

    P

    + KI

    t

    0

    e() d

    I

    + KD

    d

    dte(t)

    D

    U(s) =

    KPE(s) +

    KI

    s E(s) +

    KDsE(s)

    or

    Gc(s) = Kp+KI

    s +KDs

    = Kc 1 + 1TIs+ TDs

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 6

    P action

    LetKI= 0andKD= 0, i.e.

    Gc(s) =KP U(s) =KPE(s)

    u=

    umax e > e0KPe+u0 e0 < e < e0umin e < e0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    63/116

    8 Basic Control Actions 7

    P control

    0 5 10 15 20

    0

    0.5

    1.0

    1.5

    set point, measured variable

    0 5 10 15 20

    2

    0

    2

    4

    6control variable

    KP = 5

    KP = 2

    KP = 1

    KP = 5

    KP = 2

    KP = 1

    stationary error (offset) largeKP fast response,small offset,

    worse stability

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 8

    I action

    +

    e

    t

    u = KPe(t) +u0

    = KP

    e(t) +

    1

    TI

    t0

    e() d

    PI control

    offset exists e dincreases u0increases yincreases reduce offset

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    64/116

    8 Basic Control Actions 9

    PI control

    0 5 10 15 20

    0

    0.5

    1.0

    1.5

    set point, measured variable

    0 5 10 15 20

    0

    1

    2

    control variable

    TI = 1TI = 2

    TI = 5

    TI =

    TI = 1

    TI = 2

    TI = 5

    TI =

    removes stationary error

    largeKI fast offset removal,worse stability

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 10

    D action

    A PI-controller contains no prediction

    The same control signal is obtained for both these cases:

    time

    II

    P

    time

    P

    e

    t

    e

    t

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    65/116

    8 Basic Control Actions 11

    D actionPrediction

    error

    time

    e(t)e(t+TD)

    e(t) +TDde(t)

    dt

    P control:u(t) =KPe(t)

    PD control:u(t) = KP

    e(t) +TD

    de(t)

    dt

    KPe(t+TD)

    TD= prediction horizon (time)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 12

    PD control

    0 5 10 15 20

    0

    0.5

    1.0

    set point, measured variable

    0 5 10 15 20

    2

    0

    2

    4

    6 control variable

    TD = 0.1

    TD = 0.5

    TD = 2

    TD = 0.1

    TD = 0.5

    TD = 2

    TDis used to reduce oscillation

    TDtoo small, no influence

    TDtoo large, decrease performance

    In industrial practice, especially in noisy environment, D-control isoften turned off.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    66/116

    8 Basic Control Actions 13

    PID control

    In brief, we use

    P control to adjust overall dynamic (increase P to get faster response)

    I control to adjust steady-state part (remove offset)

    D control to adjust transient part (reduce oscillation)

    PID tuning How to tune? Ziegler & Nichols (1942) suggested twomethods:

    open-loop method (or step response method)

    closed-loop method (or ultimate cycle method)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    8 Basic Control Actions 14

    PID tuning: Open-loop method

    or step response method or reaction curve method: Draw a straight linetangent to the step response at thepoint of inflection. Suppose that linecross 0%-output and 100%-output at A and B respectively. LetLbe thetime measured from step input to the point A and Tbe the time

    measured from A to B.

    P: KP=T

    L

    PI: KP= 0.9T

    L, TI=

    L

    0.3

    PID: KP= 1.2T

    L, TI= 2L, TD= 0.5L

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    67/116

    8 Basic Control Actions 15

    PID tuning: Closed-loop method

    or ultimate cycle method: Apply P-control and gradually increase gainKuntil the system oscillate at marginal stability. Call thatKthe ultimate

    gainKuand the period of oscillationthe ultimate periodPu.

    P: KP= 0.5Ku

    PI: KP= 0.45Ku, TI= Pu

    1.2

    PID: KP= 0.6Ku, TI=Pu

    2 , TD=

    Pu

    8

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    68/116

    9 Root-Locus Technique

    basic properties of the root loci

    properties of the root lociK= 0andK= points, number of branches, symmetryroot loci on the real axis

    angles and intersect of asymptotes: behavior at |s| angles of departure and angles of arrival

    intersection with imaginary axis

    breakaway points (saddle points)

    root contours: multiple-parameter variation design gain from root locus shaping the root locus lead and lag compensators

    9 Root-Locus Technique 2

    Introduction

    location of closed-loop poles (roots of characteristic equation) relatesclosely tostabilityandtime-domaincharacteristics

    root locus: a systemetic contruction of the trajectories of the rootsof the characteristic equation

    proposed by Walter R. Evans (1920-99) in 1948 use poles and zeros of open-loop (or equivalents) to determine thetrajectories of closed-loop poles whenoneparameter is changing

    when more than one parameter varies, we called themroot contours give rough and quick graphical feeling of how to compensate for

    simple SISO system

    for root contours and more accurate root loci, computer tool(MATL AB) can be used

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    69/116

    9 Root-Locus Technique 3

    Basic properties of the root loci: Changing parameter

    Suppose the closed-loop transfer function of a control system is

    Y(s)

    R(s)=

    G(s)

    1 +G(s)H(s)

    Then the roots of characteristic equation must satisfy1 +G(s)H(s) = 0.Suppose thatG(s)H(s)contains a real variable parameter in the form

    G(s)H(s) =KB(s)

    A(s) 1 + K B(s)

    A(s) =

    A(s) +KB(s)

    A(s) = 0

    Example Consider the characteristic equation

    s(s+ 1)(s + 2) + s2 + (3 + 2K)s+ 5 = 0

    which can be written as

    1 +K 2s

    s3 + 4s2 + 5s + 5 or

    B(s)

    A(s)=

    2s

    s3 + 4s2 + 5s + 5

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 4

    Basic properties of RL: Magnitude & angle conditions

    For simplicity, let us express the characteristic equation as

    1 +KG(s) = 0 G(s) = 1K

    where

    G(s) =B(s)

    A(s)=

    mi=1(s zi)n

    j=1(s pj)=

    sm +bm1sm1 + +b1s +b0sn +an1sn1 + +a1s +a0

    To satisfy1 +KG(s) = 0, the following conditions must be satisfiedsimultaneously:

    Magnitude condition |G(s)| = 1|K|, < K <

    Angle condition G(s) =

    (2i + 1) =odd multiples of, K 0

    2i =even multiples of, K

    0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    70/116

    9 Root-Locus Technique 5

    Graphical interpretation of magnitude & angle conditions

    Example

    G(s) = s z1

    (s p1)(s p2)The figure shows thesituation whens0is on theroot loci.

    Re-axis0

    j-axis

    2 13

    s0p1

    s0p2

    s0

    s0z1

    Onlythe angle condition is used to construct the root loci.

    Only magnitude condition hasKinvolved. So, to find Kfor a specificlocation on a root locus we must use magnitude condition.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 6

    Properties of the root loci

    the number of root loci equals the order of characteristic equation root loci begin (K= 0) from open-loop polespjand end (K= ) at

    open-loop zeros (zi)

    the poles and zeros include those at infinity, if any

    they are symmetrical with respect toRe-axis and with respect to theaxes of symmetry of the pole-zero configuration.

    the entire real axis is occupied by the root loci for all values ofK forK 0, the root loci are found on a given section onRe-axis, if the

    total number of (real) poles and zeros to the right of the section is odd

    Example G(s) = s + 1

    s2

    +s 2=

    s + 1

    (s 1)(s + 2)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    71/116

    9 Root-Locus Technique 7

    Multiple roots

    ForK 0, root loci for

    G(s) = 1s2

    can be found from

    s2 +K= 0

    Re-axis0

    j-axis

    and for

    G(s) = 1

    (s )3from

    (s )3 +K= 0j-axis

    Re-axis0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 8

    Asymptotic properties

    When the magnitude ofsis large (|s| ),1 +KG(s) = 1 +KB(s)A(s)

    can

    be approximated using long division

    A(s)

    B(s)

    =snm + (an1

    bm1)snm1 +

    as1 +KG(s) = 1 +K

    B(s)

    A(s) 1 + K

    (s )nmwherean1 bm1= (n m)or

    = 1

    nm

    n

    j=1

    pj m

    i=1

    zi

    where bm1=

    mi=1

    zi, an1= n

    j=1

    pj

    In summary, the large-scale behavior of root locus is asymptotic to the

    lines (akaasymptotes) with center (centroid) atand anglesigiven by

    i=(2i+ 1) 180

    n m for K 0

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    72/116

    9 Root-Locus Technique 9

    Examples

    Example 1

    G(s) =

    s + 1

    s(s+ 2)(s + 3)

    Example 2

    G(s) = 1

    s(s+ 3)2

    Example 3

    G(s) = 1

    s(s+ 1)(s + 3)(s + 4)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 10

    Angles of departure and angles of arrival

    Angle of the tangent of root locus at a points0can be found using theangle conditionas

    m

    i=1

    (s0 zi) n

    j=1

    (s0 pj) = 180

    with the desired angle replaced byDfor the angle of departure and byAfor the angle of arrival.

    Example 4 G(s) = s + 2

    s2 + 2s + 2

    In case of multiple roots, use2,3, ... for sum of angles of multipleroots.

    Example 5 G(s) = s + 2

    (s2 + 2s + 2)2

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    73/116

    9 Root-Locus Technique 11

    Breakaway points (saddle points)

    At the point where more than one roots meet, a break-away occurs. Wecan find that point from thenecessarycondition for multiple roots that

    dds

    (1 +KG(s)) = 0

    ord

    dsG(s) = 0 or

    d

    ds

    1

    G(s)= 0

    or

    A(s)d

    dsB(s) =B(s)

    d

    dsA(s)

    orn

    j=1

    1

    s pj =m

    i=1

    1

    s zi

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 12

    Intersection with imaginary axis

    of root loci should be found using Routh-Hurwitz test because of thestability of the closed-loop system.

    The values of

    Kthat causes thej-axis crossing and the crossing locationshould also be determined.

    Example 2 [continued] The root loci crossj-axis atK= 54and theauxiliary equation iss2 + 9 = 0or at = 3rad/sec.

    Example 3 [continued] The root loci crossj-axis atK= 26.25and

    the auxiliary equation iss2

    + 1.5 = 0or at = 1.225rad/sec.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    74/116

    9 Root-Locus Technique 13

    More examples

    Example 6

    G(s) =(s+ 1)(s + 2)

    s(s2 + 2s + 2)

    Example 7

    G(s) = 1

    s(s2 + 8s + 32)

    Example 8

    G(s) = 1

    s(s+ 2)(s2

    + 2s + 2)

    Example 9

    G(s) = s2 + 4

    s(s2 + 9)

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 14

    . . . and their root loci

    3

    2

    4

    1

    1

    1

    j

    2

    24

    4

    26 4

    j

    1

    1

    2

    j

    2 134

    2

    1

    12

    3

    3j

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    75/116

    9 Root-Locus Technique 15

    Multiple roots: An example

    G(s) = s + 1

    s2(s+ 9)

    j

    8 6 4 2

    2

    4

    2

    4

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 16

    Root locus construction: Summary

    1. mark poles () & zeros () ons-plane2.nbranches, symmetrical aboutRe-axis, start (K= 0) from and end

    (K= ) at 3. draw loci forK

    0onRe-axis where the total count of real

    &

    to

    the right is odd; the remaining parts ofRe-axis are the loci forK 04. draw asymptotes (angles and centroid)

    5. find breakaway points and multiple roots (and gain)

    6. findj-axis crossing points (and gain) by Routh-Hurwitz criterion

    7. compute angles of departure & arrival

    8. complete the root locus

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

  • 7/21/2019 Mws Slide

    76/116

    9 Root-Locus Technique 17

    How to solve f(s) = 0

    bisection search: Find range[a, b]such thatf(a)f(b)< 0, i.e.,f(a)andf(b)has different sign. Withc= a+b2

    testf(c)

    iff(a)f(c)< 0, thenb cor use[a, c]iff(b)f(c)< 0 , thena cor use[c, b]

    as next range. Repeat until |a b| < . Newton-Raphson method: x x+h

    h= f(x)f(x)

    or 1

    h= f

    (x)f(x)

    + f(x)2f(x)

    iteration: Substitute a good guesssinn

    j=1

    1s pj =

    mi=1

    1s zi

    leaving only two unknowns. Solve and repeat. Good for findingbreakaway points.

    c2005 Dept. of Electrical Engineering, Chulalongkorn Univ., based on B. C. Kuo & F. Golnaraghi, Automatic Control Systems, 7th ed.

    9 Root-Locus Technique 18

    Design aspect of the root loci

    + yr e

    K G(s)u

    where

    G(s) = 1

    s(s+ 2)

    Find the range ofKso that

    the closed-loop is stable, has overshoot less than10%, has settling time less than

    1.5s,

    etc.

    Effects of adding poles and zeros to G(s)

    adding a pole toG(s)has the effect of pushing the root loci towardthe right-halfs-plane

    adding left-half plane zeros to