57
Passive Network Synthesis Revisited Malcolm C. Smith Department of Engineering University of Cambridge U.K. Mathematical Theory of Networks and Systems (MTNS) 17th International Symposium Kyoto International Conference Hall Kyoto, Japan, July 24-28, 2006 Semi-Plenary Lecture 1

Passive Network Synthesis Revisited Malcolm C. Smith ... › ~helton › MTNSHISTORY › CONTENTS › 2006K… · Passive Network Synthesis Revisited Malcolm C. Smith The Ideal Inerter

  • Upload
    others

  • View
    10

  • Download
    0

Embed Size (px)

Citation preview

  • Passive Network Synthesis Revisited

    Malcolm C. Smith

    Department of Engineering

    University of Cambridge

    U.K.

    Mathematical Theory of Networks and Systems (MTNS)

    17th International Symposium

    Kyoto International Conference Hall

    Kyoto, Japan, July 24-28, 2006

    Semi-Plenary Lecture

    1

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Outline of Talk

    1. Motivating example (vehicle suspension).

    2. A new mechanical element.

    3. Positive-real functions and Brune synthesis.

    4. Bott-Duffin method.

    5. Darlington synthesis.

    6. Minimum reactance synthesis.

    7. Synthesis of resistive n-ports.

    8. Vehicle suspension.

    9. Synthesis with restricted complexity.

    10. Motorcycle steering instabilities.

    MTNS 2006, Kyoto, 24 July, 2006 2

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Motivating Example – Vehicle Suspension

    Performance objectives

    1. Control vehicle body in the face of variable loads.

    2. Insulate effect of road undulations (ride).

    3. Minimise roll, pitch under braking, acceleration and

    cornering (handling).

    MTNS 2006, Kyoto, 24 July, 2006 3

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Quarter-car Vehicle Model (conventional suspension)

    PSfrag replacements

    tyre

    load

    disturbances

    road

    disturbances

    mu

    ms

    damperspring

    kt

    MTNS 2006, Kyoto, 24 July, 2006 4

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Most General Passive Vehicle Suspension

    PSfrag replacements

    mu

    Z(s)

    ms

    zrkt

    Replace the spring and damper with a

    general positive-real impedance Z(s).

    But is Z(s) physically realisable?

    MTNS 2006, Kyoto, 24 July, 2006 5

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Electrical-Mechanical Analogies

    1. Force-Voltage Analogy.

    voltage ↔ forcecurrent ↔ velocity

    Oldest analogy historically, cf. electromotive force.

    2. Force-Current Analogy.

    current ↔ forcevoltage ↔ velocity

    electrical ground ↔ mechanical ground

    Independently proposed by: Darrieus (1929), Hähnle (1932), Firestone (1933).

    Respects circuit “topology”, e.g. terminals, through- and across-variables.

    MTNS 2006, Kyoto, 24 July, 2006 6

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Standard Element Correspondences (Force-Current Analogy)

    v = Ri resistor ↔ damper cv = Fv = L didt inductor ↔ spring kv = dFdt

    C dvdt = i capacitor ↔ mass m dvdt = F

    PSfrag replacements v1

    v1

    v2

    v2

    ii

    FF

    Electrical

    Mechanical

    What are the terminals of the mass element?

    MTNS 2006, Kyoto, 24 July, 2006 7

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Exceptional Nature of the Mass Element

    Newton’s Second Law gives the following network interpretation of the mass

    element:

    • One terminal is the centre of mass,• Other terminal is a fixed point in the inertial frame.

    Hence, the mass element is analogous to a grounded capacitor.

    Standard network symbol

    for the mass element:PSfrag replacements

    v1 = 0v2

    F

    MTNS 2006, Kyoto, 24 July, 2006 8

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Table of usual correspondences

    PSfrag replacements

    ElectricalMechanical

    spring

    mass

    damper

    inductor

    capacitor

    resistor

    i

    i

    ii

    i

    i

    v1

    v1

    v1

    v1

    v1

    v1 = 0v2

    v2

    v2

    v2

    v2

    v2

    F

    FF

    FF

    MTNS 2006, Kyoto, 24 July, 2006 9

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Consequences for network synthesis

    Two major problems with the use of the mass element for synthesis of

    “black-box” mechanical impedances:

    • An electrical circuit with ungrounded capacitors will not have a directmechanical analogue,

    • Possibility of unreasonably large masses being required.

    Question

    Is it possible to construct a physical device such that

    the relative acceleration between its endpoints is pro-

    portional to the applied force?

    MTNS 2006, Kyoto, 24 July, 2006 10

  • Passive Network Synthesis Revisited Malcolm C. Smith

    One method of realisation

    PSfrag replacements

    terminal 2 terminal 1gear

    rack pinions

    flywheel

    Suppose the flywheel of mass m rotates by α radians per meter of relative

    displacement between the terminals. Then:

    F = (mα2) (v̇2 − v̇1)

    (Assumes mass of gears, housing etc is negligible.)

    MTNS 2006, Kyoto, 24 July, 2006 11

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Ideal Inerter

    We define the Ideal Inerter to be a mechanical one-port device such

    that the equal and opposite force applied at the nodes is

    proportional to the relative acceleration between the nodes, i.e.

    F = b(v̇2 − v̇1).

    We call the constant b the inertance and its units are kilograms.

    The ideal inerter can be approximated in the same sense that real springs,

    dampers, inductors, etc approximate their mathematical ideals.

    We can assume its mass is small.

    M.C. Smith, Synthesis of Mechanical Networks: The Inerter,

    IEEE Trans. on Automat. Contr., 47 (2002), pp. 1648–1662.

    MTNS 2006, Kyoto, 24 July, 2006 12

  • Passive Network Synthesis Revisited Malcolm C. Smith

    A new correspondence for network synthesis

    PSfrag replacements

    ••

    ElectricalMechanical

    spring

    inerter

    damper

    inductor

    capacitor

    resistor

    i

    i

    ii

    i

    i

    v1

    v1

    v1

    v1

    v1

    v1

    v2

    v2

    v2

    v2

    v2

    v2

    Y (s) = ks

    dFdt = k(v2 − v1)

    Y (s) = bs

    F = bd(v2−v1)dt

    Y (s) = c

    F = c(v2 − v1)

    Y (s) = 1Ls

    didt =

    1L (v2 − v1)

    Y (s) = Cs

    i = C d(v2−v1)dt

    Y (s) = 1R

    i = 1R (v2 − v1)

    FF

    FF

    FF

    Y (s) = admittance =1

    impedance

    MTNS 2006, Kyoto, 24 July, 2006 13

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Rack and pinion inerter

    made at

    Cambridge University

    Engineering Department

    mass ≈ 3.5 kginertance ≈ 725 kg

    stroke ≈ 80 mm

    MTNS 2006, Kyoto, 24 July, 2006 14

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Damper-inerter series arrangement

    with centring springs

    PSfrag replacements

    MTNS 2006, Kyoto, 24 July, 2006 15

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Alternative Realisation of the Inerter

    PSfrag replacements

    screw nut flywheel

    MTNS 2006, Kyoto, 24 July, 2006 16

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Ballscrew inerter made at Cambridge University Engineering Department

    Mass ≈ 1 kg, Inertance (adjustable) = 60–180 kg

    MTNS 2006, Kyoto, 24 July, 2006 17

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Electrical equivalent of quarter car model

    PSfrag replacements

    mu

    Fs

    Y (s)

    ms

    zs

    zu

    zr

    kt

    PSfrag replacements

    żr

    mu

    Y (s)

    ms Fs

    k−1t

    +−

    Y (s) = Admittance =Force

    Velocity=

    Current

    Voltage

    MTNS 2006, Kyoto, 24 July, 2006 18

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Positive-real functions

    Definition. A function Z(s) is defined to be positive-real if one of the

    following two equivalent conditions is satisfied:

    1. Z(s) is analytic and Z(s) + Z(s)∗ ≥ 0 in Re(s) > 0.

    2. Z(s) is analytic in Re(s) > 0, Z(jω) + Z(jω)∗ ≥ 0 for all ω at whichZ(jω) is finite, and any poles of Z(s) on the imaginary axis or at infinity

    are simple and have a positive residue.

    MTNS 2006, Kyoto, 24 July, 2006 19

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Passivity Defined

    Definition. A network is passive if for all admissible v, i which are square

    integrable on (−∞, T ], ∫ T

    −∞v(t)i(t) dt ≥ 0.

    Proposition. Consider a one-port electrical network for which the impedance

    Z(s) exists and is real-rational. The network is passive if and only if Z(s) is

    positive-real.

    R.W. Newcomb, Linear Multiport Synthesis, McGraw-Hill, 1966.

    B.D.O. Anderson and S. Vongpanitlerd, Network Analysis and Synthesis, Prentice-Hall, 1973.

    MTNS 2006, Kyoto, 24 July, 2006 20

  • Passive Network Synthesis Revisited Malcolm C. Smith

    O. Brune showed that any (ratio-

    nal) positive-real function could

    be realised as the impedance

    or admittance of a network

    comprising resistors, capacitors,

    inductors and transformers.

    (1931)

    MTNS 2006, Kyoto, 24 July, 2006 21

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Minimum functions

    A minimum function Z(s) is a positive-real function with no poles or zeros

    on jR ∪ {∞} and with the real part of Z(jω) equal to 0 at one or morefrequencies.

    PSfrag replacements

    ReZ(jω)

    0 ωω1 ω2

    MTNS 2006, Kyoto, 24 July, 2006 22

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Foster preamble for a positive-real Z(s)

    Removal of poles/zeros on jR ∪ {∞}. e.g.

    s2 + s+ 1

    s+ 1= s+

    1

    s+ 1↑

    lossless

    ↓s2 + 1

    s2 + 2s+ 1=

    (2s

    s2 + 1+ 1

    )−1

    Can always reduce a positive-real Z(s) to a minimum function.

    MTNS 2006, Kyoto, 24 July, 2006 23

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle

    Let Z(s) be a minimum function with Z(jω1) = jX1 (ω1 > 0).

    Write L1 = X1/ω1 and Z1(s) = Z(s)− L1s.Case 1. (L1 < 0)

    PSfrag replacements

    Z(s) Z1(s)

    L1 < 0

    (negative inductor!)

    Z1(s) is positive-real. Let Y1(s) = 1/Z1(s). Therefore, we can write

    Y2(s) = Y1(s)−2K1s

    s2 + ω21

    for some K1 > 0.

    MTNS 2006, Kyoto, 24 July, 2006 24

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    Then:

    PSfrag replacements

    Z(s) Z2(s)

    L1 < 0

    L2 > 0

    C2 > 0

    where L2 = 1/2K1, C2 = 2K1/ω21 and Z2 = 1/Y2.

    Straightforward calculation shows that

    Z2(s) = sL3+ Z3(s)↑

    proper

    where L3 = −L1/(1 + 2K1L1). Since Z2(s) is positive-real, L3 > 0 and Z3(s)is positive-real.

    MTNS 2006, Kyoto, 24 July, 2006 25

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    Then:

    PSfrag replacements

    Z(s) Z3(s)

    L1 < 0

    L2 > 0

    L3 > 0

    C2 > 0

    To remove negative inductor:

    PSfrag replacements

    L1 L3M

    Lp L2LsLp = L1 + L2

    Ls = L2 + L3

    M = L2

    Some algebra shows that: Lp, Ls > 0 andM2

    LpLs= 1 (unity coupling

    coefficient).

    MTNS 2006, Kyoto, 24 July, 2006 26

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    Realisation for completed cycle:

    PSfrag replacements

    M

    Lp LsZ(s) Z3(s)

    C2

    MTNS 2006, Kyoto, 24 July, 2006 27

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    Case 2. (L1 > 0). As before Z1(s) = Z(s)− L1s

    PSfrag replacements

    Z(s) Z1(s)

    L1 > 0

    (no need for

    negative inductor!)

    Problem: Z1(s) is not positive-real!

    Let’s press on and hope for the best!!

    As before let Y1 = 1/Z1 and write

    Y2(s) = Y1(s)−2K1s

    s2 + ω21.

    Despite the fact that Y1 is not positive-real we can show that K1 > 0.

    MTNS 2006, Kyoto, 24 July, 2006 28

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    Hence:

    PSfrag replacements

    Z(s) Z2(s)

    L1 > 0

    L2 > 0

    C2 > 0

    But still Z2(s) is not positive-real. Again we can check that

    Z2(s) = sL3+ Z3(s)↑

    proper

    where L3 = −L1/(1 + 2K1L1).This time L3 < 0 and Z3(s) is positive-real.

    MTNS 2006, Kyoto, 24 July, 2006 29

  • Passive Network Synthesis Revisited Malcolm C. Smith

    The Brune cycle (cont.)

    So:

    PSfrag replacements

    Z(s) Z3(s)

    L1 > 0

    L2 > 0

    L3 < 0

    C2 > 0

    As before we can transform to:

    PSfrag replacements

    M

    Lp LsZ(s) Z3(s)

    C2

    where Lp, Ls > 0 andM2

    LpLs= 1 (unity coupling coefficient).

    MTNS 2006, Kyoto, 24 July, 2006 30

  • Passive Network Synthesis Revisited Malcolm C. Smith

    R. Bott and R.J. Duffin showed

    that transformers were un-

    necessary in the synthesis of

    positive-real functions. (1949)

    MTNS 2006, Kyoto, 24 July, 2006 31

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Richards’s transformation

    Theorem. If Z(s) is positive-real then

    R(s) =kZ(s)− sZ(k)kZ(k)− sZ(s)

    is positive-real for any k > 0.

    Proof.

    Z(s) is p.r. ⇒ Y (s) = Z(s)− Z(k)Z(s) + Z(k)

    is b.r. and Y (k) = 0

    ⇒ Y ′(s) = k + sk − sY (s) is b.r.

    ⇒ Z ′(s) = 1 + Y′(s)

    1− Y ′(s) is p.r.

    R(s) = Z ′(s) after simplification.

    MTNS 2006, Kyoto, 24 July, 2006 32

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Bott-Duffin construction (cont.)

    Idea: use Richards’s transformation to eliminate transformers from Brune

    cycle.

    As before, let Z(s) be a minimum function with Z(jω1) = jX1 (ω1 > 0).

    Write L1 = X1/ω1.

    Case 1. (L1 > 0)

    Since Z(s) is a minimum function we can always find a k s.t. L1 = Z(k)/k.

    Therefore:

    R(s) =kZ(s)− sZ(k)kZ(k)− sZ(s)

    has a zero at s = jω1.

    MTNS 2006, Kyoto, 24 July, 2006 33

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Bott-Duffin construction (cont.)

    We now write:

    Z(s) =kZ(k)R(s) + Z(k)s

    k + sR(s)

    =kZ(k)R(s)

    k + sR(s)+

    Z(k)s

    k + sR(s)

    =1

    1Z(k)R(s) +

    skZ(k)

    +1

    kZ(k)s +

    R(s)Z(k)

    .

    MTNS 2006, Kyoto, 24 July, 2006 34

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Bott-Duffin construction (cont.)

    Z(s) =1

    1Z(k)R(s) +

    skZ(k)

    +1

    kZ(k)s +

    R(s)Z(k)

    PSfrag replacements

    Z(k)k

    Z(s)

    kZ(k) Z(k)R(s)

    Z(k)R(s)

    MTNS 2006, Kyoto, 24 July, 2006 35

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Bott-Duffin construction (cont.)

    We can write: 1Z(k)R(s) = const× ss2+ω21 +1

    R1(s)etc.

    PSfrag replacements

    R1

    R2

    Z(s)

    MTNS 2006, Kyoto, 24 July, 2006 36

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Example — restricted degree

    Proposition. Consider the real-rational function

    Yb(s) = ka0s

    2 + a1s+ 1

    s(d0s2 + d1s+ 1)

    where d0, d1 ≥ 0 and k > 0. Then Yb(s) is positive real if only if the followingthree inequalities hold:

    β1 := a0d1 − a1d0 ≥ 0,β2 := a0 − d0 ≥ 0,β3 := a1 − d1 ≥ 0.

    MTNS 2006, Kyoto, 24 July, 2006 37

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Brune Realisation Procedure for Yb(s)

    Foster preamble always sufficient to complete the realisation if β1, β2 > 0.

    (No Brune or Bott-Duffin cycle is required).

    A continued fraction expansion is obtained:

    Yb(s) = ka0s

    2 + a1s+ 1

    s(d0s2 + d1s+ 1)

    =k

    s+

    1s

    kb+

    1

    c3 +1

    1

    c4+

    1

    b2s

    where kb =kβ2d0

    , c3 = kβ3, c4 =kβ4β1

    , b2 =kβ4β2

    and

    β4 := β22 − β1β3.

    PSfrag replacementsk

    c3

    b2

    kb

    c4

    MTNS 2006, Kyoto, 24 July, 2006 38

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Darlington Synthesis

    Realisation in Darlington form:

    a lossless two-port terminated

    in a single resistor.

    PSfrag replacements

    I1 I2

    V1 V2R Ω

    Y (s) Losslessnetwork

    For a lossless two-port with impedance:

    Z =

    Z11 Z12

    Z12 Z22

    we find

    Z1(s) = Z11R−1(Z11Z22 − Z212)/Z11 + 1

    R−1Z22 + 1.

    MTNS 2006, Kyoto, 24 July, 2006 39

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Writing

    Z1 =m1 + n1m2 + n2

    =n1m2

    m1/n1 + 1

    n2/m2 + 1,

    where m1, m2 are polynomials of even powers of s and n1, n2 are polynomials

    of odd powers of s, suggests the identification:

    Z11 =n1m2

    , Z22 = Rn2m2

    , Z12 =√R

    √n1n2 −m1m2

    m2.

    Augmentation factors are necessary to ensure a rational square root.

    Once Z(s) has been found, we then write:

    Z(s) = sC1 +s

    s2 + α2C2 + · · ·

    where C1 and C2 are non-negative definite constant matrices.

    MTNS 2006, Kyoto, 24 July, 2006 40

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Darlington Synthesis (cont.)Each term in the sum is realised in the form of a T-circuit and a series

    connection of all the elementary two-ports is then made:

    PSfrag replacements

    X1(s) X2(s)

    X3(s)

    1 : n

    Ideal

    Network 1Network 2

    I = 0

    PSfrag replacements

    X1(s)

    X2(s)

    X3(s)

    1 : n

    Ideal

    Network 1

    Network 2

    I = 0

    MTNS 2006, Kyoto, 24 July, 2006 41

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Electrical and mechanical realisations of the admittance Yb(s)

    PSfrag replacements

    k2

    k3

    λ1

    λ2

    k1

    b

    c

    k−12 H

    k−13 H

    1: − ρ

    k−11 H b F

    R1 Ω

    PSfrag replacements

    k2 k3

    λ1 λ2

    k1b c

    k−12 H

    k−13 H

    1: − ρk−11 H

    b F

    R1 Ω

    MTNS 2006, Kyoto, 24 July, 2006 42

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Minimum reactance synthesis

    PSfrag replacementsZ(s)

    Nondynamic

    Network

    X

    sL1

    1sC1

    Let L1 = · · · = C1 = · · · = 1. If

    M =

    M11 M12

    M21 M22

    is the hybrid matrix of X, i.e. v1

    i2

    = M

    i1

    v2

    ,

    then

    Z(s) = M11−M12(sI+M22)−1M21.

    D.C. Youla and P. Tissi, “N-Port Synthesis via Reactance Extraction, Part I”, IEEEInternational Convention Record, 183–205, 1966.

    MTNS 2006, Kyoto, 24 July, 2006 43

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Minimum reactance synthesis

    Conversely, if we can find a state-space realisation Z(s) = C(sI −A)−1B +Dsuch that the constant matrix

    M =

    D −CB −A

    has the properties

    M +M ′ ≥ 0,diag{I,Σ}M = Mdiag{I,Σ}

    where Σ is a diagonal matrix with diagonal entries +1 or −1. Then M is thehybrid matrix of the nondynamic network terminated with inductors or

    capacitors, which realises Z(s).

    A construction is possible using the positive-real lemma and matrix

    factorisations.

    B.D.O. Anderson and S. Vongpanitlerd, Network Analysis and Synthesis, Prentice-Hall, 1973.

    MTNS 2006, Kyoto, 24 July, 2006 44

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Synthesis of resistive n-ports

    Let R be a symmetric n× n matrix.A necessary and sufficient condition for R to be realisable as the driving-point

    impedance of a network comprising resistors and transformers only is that it

    is non-negative definite.

    No necessary and sufficient condition is known in the case that transformers

    are not available.

    A general necessary condition is known: that the matrix is paramount.1

    A matrix is defined to be the paramount if each principal minor of the matrix

    is not less than the absolute value of any minor built from the same rows.

    It is also known that paramountcy is sufficient for the case of n ≤ 3.21I. Cederbaum, “Conditions for the Impedance and Admittance Matrices of n-ports without

    Ideal Transformers”, IEE Monograph No. 276R, 245–251, 1958.2P. Slepian and L. Weinberg, ”Synthesis applications of paramount and dominant matrices”,

    Proc. National Electron. Conf., vol. 14, Chicago, Illinois, Oct. 611-630, 1958.

    MTNS 2006, Kyoto, 24 July, 2006 45

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Simple Suspension Struts

    PSfrag replacements

    b

    kb

    k c

    k1

    k2

    PSfrag replacements

    bkb

    k c

    k1

    k2

    PSfrag replacements

    b

    kb k

    c

    k1

    k2

    PSfrag replacements

    b

    kb k

    ck1

    k2

    Layout S1 Layout S2 Layout S3 Layout S4

    parallel series

    MTNS 2006, Kyoto, 24 July, 2006 46

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Performance Measures

    Assume:

    Road Profile Spectrum = κ|n|−2 (m3/cycle)where κ = 5× 10−7 m3cycle−1 = road roughness parameter. Define:

    J1 = E[z̈2s(t)

    ]ride comfort

    = r.m.s. body vertical acceleration

    MTNS 2006, Kyoto, 24 July, 2006 47

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Optimisation of J1 (ride comfort)

    1 2 3 4 5 6 7 8 9 10 11 120.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2

    2.2

    2.4

    2.6

    PSfrag replacements

    static stiffness

    J1

    % J11 2 3 4 5 6 7 8 9 10 11 12

    0

    2

    4

    6

    8

    10

    12

    14

    16

    18

    PSfrag replacements

    static stiffness

    J1

    %J

    1

    (a) Optimal J1 (b) Percentage improvement in J1

    Key: layout S1 (bold), layout S2 (dashed), layout S3 (dot-dashed), and

    layout S4 (solid).

    M.C. Smith and F-C. Wang, 2004, Performance Benefits in Passive Vehicle Sus-

    pensions Employing Inerters, Vehicle System Dynamics, 42, 235–257.

    MTNS 2006, Kyoto, 24 July, 2006 48

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Control synthesis formulation

    PSfrag replacements

    F

    F

    Fs

    K(s)

    zr

    zs

    zumu

    kt

    ms

    ks

    PSfrag replacements

    F

    FsK(s)

    zr

    zs

    zu

    mu

    kt

    ms

    ks zwG(s)

    K(s)

    F żs − żu

    Ride comfort: w = zr, z = żs

    Performance measure: J1 = const.× ‖Tẑr→sẑs‖2

    MTNS 2006, Kyoto, 24 July, 2006 49

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Bilinear Matrix Inequality (BMI) formulation

    Let K(s) = Ck(sI − Ak)−1Bk +Dk and Tẑr→sẑs = Ccl(sI −Acl)−1Bcl.

    Theorem. There exists a positive real controller K(s) such that‖Tẑr→sẑs‖2 < ν and Acl is stable, if and only if the following problem isfeasible for some Xcl > 0, Xk > 0, Q, ν

    2 and Ak, Bk, Ck, Dk of compatible

    dimensions:A

    TclXcl +XclAcl XclBcl

    BTclXcl −I

    < 0,

    Xcl C

    Tcl

    Ccl Q

    > 0, tr(Q) < ν2,

    A

    TkXk +XkAk XkBk − CTkBTk Xk − Ck −DTk −Dk

    < 0.

    C. Papageorgiou and M.C. Smith, 2006, Positive real synthesis using matrix inequalities for mechanical

    networks: application to vehicle suspension, IEEE Trans. on Contr. Syst. Tech., 14, 423–435.

    MTNS 2006, Kyoto, 24 July, 2006 50

  • Passive Network Synthesis Revisited Malcolm C. Smith

    A special problem

    What class of positive-real functions Z(s) can be realised using one damper,

    one inerter, any number of springs and no transformers?

    PSfrag replacements

    Z(s) X

    c

    b

    Leads to the question: when can X be realised as a network of springs?

    MTNS 2006, Kyoto, 24 July, 2006 51

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Theorem. Let

    Y (s) =(R2R3 −R26) s3 +R3 s2 +R2 s+ 1

    s(detRs3 + (R1R3 −R25) s2 + (R1R2 −R24) s+R1), (1)

    where R :=

    R1 R4 R5

    R4 R2 R6

    R5 R6 R3

    is non-negative definite.

    A positive-real function Y (s) can be realised as the driving-point admittance

    of a network comprising one damper, one inerter, any number of springs and

    no transformers if and only if Y (s) can be written in the form of (1) and there

    exists an invertible diagonal matrix D = diag{1, x, y} such that DRD isparamount.

    An explicit set of inequalities can be found which are necessary and sufficient

    for the existence of x and y.

    M.Z.Q. Chen and M.C. Smith, Mechanical networks comprising one damper and one inerter, in preparation.

    MTNS 2006, Kyoto, 24 July, 2006 52

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Collaboration with Imperial College

    Application to motorcycle stability.

    At high speed motorcycles can experience significant steering instabilities.

    Observe: Paul Orritt at the 1999 Manx Grand Prix

    MTNS 2006, Kyoto, 24 July, 2006 53

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Weave and Wobble oscillations

    Steering dampers improve wobble (2–4 Hz) and worsen weave (6–9 Hz).

    Simulations show that steering inerters have, roughly, the opposite effect to

    the damper. Root-loci (with speed the varied parameter):

    Steering damper root-locus Steering inerter root-locus

    −18 −16 −14 −12 −10 −8 −6 −4 −2 0 2 40

    10

    20

    30

    40

    50

    60

    PSfrag replacements

    Real Axis

    Imag

    inar

    yA

    xis

    wobble

    weave

    6

    −15 −10 −5 0 50

    10

    20

    30

    40

    50

    60

    PSfrag replacements

    Real AxisIm

    agin

    ary

    Axi

    s

    wobble

    weave

    6

    Can the advantages be combined?

    S. Evangelou, D.J.N. Limebeer, R.S. Sharp and M.C. Smith, 2006, Steering compensation for high-

    performance motorcycles, Transactions of ASME, J. of Applied Mechanics, 73, to appear.

    MTNS 2006, Kyoto, 24 July, 2006 54

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Solution — a steering compensator

    ... consisting of a network of dampers, inerters and springs.

    Needs to behave like an inerter at weave frequencies and like a damper at

    wobble frequencies.

    PSfrag replacements

    damper

    inerter

    realimaginary

    forbidden

    PSfrag replacements

    damper

    inerter

    realimaginary

    forbidden

    Prototype designed by N.E. Houghton and manufactured in the Cambridge University

    Engineering Department.

    MTNS 2006, Kyoto, 24 July, 2006 55

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Conclusion

    • A new mechanical element called the “inerter” was introduced which isthe true network dual of the spring.

    • The inerter allows classical electrical network synthesis to be mappedexactly onto mechanical networks.

    • Applications of the inerter: vehicle suspension, motorcycle steering andvibration absorption.

    • Economy of realisation is an important problem for mechanical networksynthesis.

    • The problem of minimal realisation of positive-real functions remainsunsolved.

    MTNS 2006, Kyoto, 24 July, 2006 56

  • Passive Network Synthesis Revisited Malcolm C. Smith

    Acknowledgements

    Research students Michael Chen, Frank Scheibe

    Design Engineer Neil Houghton

    Technicians Alistair Ross, Barry Puddifoot, John Beavis

    Collaborations Fu-Cheng Wang (National Taiwan University)

    Christos Papageorgiou (University of Cyprus)

    David Limebeer, Simos Evangelou, Robin Sharp

    (Imperial College)

    MTNS 2006, Kyoto, 24 July, 2006 57