13
Instructions for use Title Role of Tensorial Formulations in Characterizing the Pointwise Multi-Dimensional Performance of Dynamic Systems Author(s) Takizawa, Haruo Citation 北海道大學工學部研究報告, 145, 1-12 Issue Date 1988-12-27 Doc URL http://hdl.handle.net/2115/42177 Type bulletin (article) File Information 145_1-12.pdf Hokkaido University Collection of Scholarly and Academic Papers : HUSCAP

Instructions for use - HUSCAP...Instructions for use Title Role of Tensorial Formulations in Characterizing the Pointwise Multi-Dimensional Performance of Dynamic Systems Author(s)

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

  • Instructions for use

    Title Role of Tensorial Formulations in Characterizing the Pointwise Multi-Dimensional Performance of Dynamic Systems

    Author(s) Takizawa, Haruo

    Citation 北海道大學工學部研究報告, 145, 1-12

    Issue Date 1988-12-27

    Doc URL http://hdl.handle.net/2115/42177

    Type bulletin (article)

    File Information 145_1-12.pdf

    Hokkaido University Collection of Scholarly and Academic Papers : HUSCAP

    https://eprints.lib.hokudai.ac.jp/dspace/about.en.jsp

  • 北海道大学工学部研究報告

    第145号(昭和63年)Bulletin of the Faculty of Engineering,

     Kokkaido University, No. 145 (1988)

      ROLE OF TENSORIAL FORMULATIONSEN CffARACTERIZING THE POKNTWISE  MULTI-DgMENSIONAL PERFORMANCE            OF DYNAMIC SYSTEMS

      Haruo TAKIzAwA*

    (Received September IO, 1988)

    Abstract

         A family of mathematical formulations of a strictly tensorial nature is shown to

    be advantageousiy used for obtaiRing better insights into the pointwise multi-

    dimensional features of earthqual〈e shal〈ing and structural response. This can be

    rendered sufficiently general and resorts primarily to the tensors of root-square values

    or root-mean-square amplitudes including aiso the tensors of their time-domain and/or

    frequency-domain decoinposition. By permitting to interpret a}i these tensors consis-

    teRtly by meaRs of the energy fed into simpie oscillators of muiti-dimensional isotropy,

    th・憤・dl・id・・1・・1・・a・e t・be ass{9・・d w{thi・th・f・am・w・・k・f・t・u伽,ai dy。。mics.

    Spectral decompositioR accompanies an Herinite and non-negative definite tensor,

    rnechanical implication of its complex cross-correlatioR components being noted with

    a specific interest. With the mechanics of structure related straightforwardiy to the

    multi-dimensioRal characteristics of earthquake shakings, findings in the present study

    ・・eth・・highlight・曲y g・・m・t・lc simpii・itie・whi・h・h・・act。,lze th。 te。s。, fi。lds。f

    response developed over the constitutive points of system. They prove useful for

    understanding important trends behind the quite messy appeareRce of spatial distribu-

    tion. .Examined in addition are practically important but non-tensorial properties in

    the pointwise directionai depeRdence of peak amplitudes.

    Instructive Equalities for a MechaRieal Understanding

             of Mathematieally Defined TeRsors

         For a mu}ti-dimensional time history, {a(t)},

    rotating coordinate axes, two matrices of

    which exhibits its vectorial nature upon

    [P(t, Q))] = Re({F(t, (v)}Mt(t bl)}i”)

    w…{F(1,・)}イ{・(・、〉}・x,(一}ωt1)・・、

                      co

    ’AuSiegEiaFtaeceiOyfeoSfSOErnig’:.nCehearriigtE.Of the Chair of Structural Engineering 1 (Department of Architec.

  • 2 Haruo TAI〈lzAWA 2

       [P’(t,ω,ζ)]=Re({F’(t,ω,ζ)}{F’(t,ω,ζ)}T)

           w・th{・,(・,ω,ζ)}一IL({・(t1)}・xp(一ζω(・一t1)))…←jfi g bl・1)d・,

    are盒rst introduced。 Obviously之hey prove real, symmetric and Ron-negative definite ten-

    sors. Physlcal meaniRg of each te登sor is also clear therein;[P(t,ω)]112 stands for the

    or(iinary Fourier amplitude of{a(t1)}tmncated at ti :t, while [P(t,ω,ζ)]1’2 corres-

    pondsωits evolutionary(instantaReous)Fourier amplltude weighted toward t under a

    suitable choice of the parameterζ. With relation to the former tensor, a multi-dimensional

    version of Husid piot:

        〔E(・)]一△{・(t1)}{・(t1)}・d・1

             【一÷九..〔・(・,ω)]・・】

    is addidonally used. Thls incl“des, as its particular case of[E(○○〉], the squared RS

    (root-sq縦are)or RMS(root-mean-square)i跳ensity tensors proposed by Arias or PenzieR.

         Then let another set o負wo matrices(real, symmetric, non-negative definite a登d

    tensor1al)be def盈ed witho級t resorting to the Fourier integral tra捻sform but by means of the

    motion of isotropically multi-dimensional simple oscillat6rs subjected to{a(t)};

        [W(・,ω,ζ)]一一∫1({・(el)}{a(・・)}・+{a(・1)}{・(・1)}・)・・1

                  【一{a(・)}{a(・)}・+・ζω f.1{a(t1)}{a(・1)}・d・・+・・{・(・)}{・(の円

        [W,(・,・,ζ)Ha(・)}{a(・)}・+ζω金({・(・)}{・(・)}・)+ω・{・(・)}{・(・)}・

                    コ                                 サ

           wiもh {d(t)}÷2ζの{d(t)}十ω2{d(t)}m一{a(t)}

               (notation of{d(t)}used instead of{d(t,(i),ζ)}for brevity)

    Of these,(1/2)[W(t,ω,ζ)l provides a necessary a識d sufficle織t tool to represent the work

    transmitted into the oscillator until t under the action of{a(t)}。 Its second term of e登ergy

    dissipation is replaced in[W’(t,ω,ζ)]by a totally differe就appearence of the damping

    actio登having no immediate mechanical mea鍛1簸g. With its i簸signiflcant contrib鷺tlon,

    (1/2)[W’(t,ω,ζ)]equals essentially to the te鍛sor of kinetic plus strain energles stored ln

    the OSCi至latOr at t。

         It can be shown thatthe two groups of tensors are mathematically related oRe another

    by、the following strict equali乞ies.

    f。..w(ω・ω,ζ)[・(・,ω1)]…1一:[W(・,・,ζ)]

                                  (4/π)(油ω12

           wlth w(ω1,ω,ζ)=                          (ωユ2一ω2)2→一4ζ2ω2ω12

        [P’(t,ω,ζ)]:[W’(t,ω,ζ)]

    ・踏

  • Holi

    day 

    l nn

     re

    cord

    . Sa

    n Fe

    rnan

    do 

    EQ.

     Fe

    b. 

    9. 

    1971

                  

    NOOW 

    component 

    (peak 

    acceleration 

    = 

    250.

    0 gal}

    horizontal plane

     s 

    gew

      _こE(

    ・・)]ユ/2

    {a(

    t)}

    period 

    C$ec)

    s.o

    3.0

    2.0

    1.0

    O.7

    O.5

    O.3

    O.2

    O。葦

    … 

     二

    二亀

    e9(

    ≠〉鉛《2

      

      

      

            雷紹魅も魅⑳魅蟻蟻馬亀馬9曾

    ◎も

    。。

    @98 

     豊

    亀噂θθ◎09

    9・噺・・…  。

    ・ 9 

    ・ θρ§i) φ葛

    も魁

    。の

     …

      

     。

     。

     …

      

     。

     ・

     ・

    . ・ e 

    “ cae

    ee e 

    g . 

    . 

    . 

    . 

    . 

    ・ 

    . 

    . 

    ・ 

    ・ 

    . 

    ・ 

    . 

    . 

    .ee 

    a」 

    s eeei2)

     es

    b e’

    ’”‘’’’’’’’’”

    ・ 

    ④8 

    亀㊥

    働{

    多 

    ◎ 

    ・ ● 馬

    ● ρ ∈珍柳 動 ㊧ 8 

    0 ◎ . 。

    ・ o馬

    留亀0 

    9 ・ 。 ・

     ・

    . . eqp 

    e e 

    e . . . . .

    . 

     一

      

    t  

    e  

    e  

    “ 

     .

      

    . 

     .  .  .

    g  

    o  

    ● 

     ◎

      

     ■

      

     ●

      

     ●

      

     ●

      

     幽

      

     響

      

     o

    ● 

      

    9  

     ●    ら    o 

       ●    の    ■   

     ,    ,    ●

    一 

      

     一

      

      

    i    

    一 

      

     一    

    一    

    一    

    一    

    十 

      

     十

      

      

    i    一    一    i 

       一    一    t

    [ec

     i〈

    t ,

     ee

     ,

     O.

    05)

    ]i/

    2

    [w(

    co 

    , 

    as 

    , 

    o.os

    )]

    i/2 

    (li)

    hori

    zo-n.ta. .

    1.

    .plane 

    e

      S90W

     ’

    f、

     、

    ㌔100

      

    Neow 

    Eli)

    Kv-loo 

    kine $

     e

    time (sec)

    ee te

    F韮gure

     1

    o5

    IO 15

     20

     25

     30

     35

     4e

     45

     5e

     55

    co

    臣 塁 1罎

    /gz.

    覇 :,g/

    羅 i,/1/

    蒙 轟 IZ’5

    N

  • 4 Haruo TAKIZAwA 4

    With definite-integral and limiting properties of

    f, co w(bli, bl, g) dbl, i 1

       lim w(bli, (h), e) = 6(bli-bl)

        g一.o

    the role of w(bli, bl, g) is understood as a weighting function centered on bli = (h) in the

    above integral operation. Actually [W(t, en, e)] coincides with [P(t, bl)] averaged around

    bl, the smoothing in which becomes more notable according to a larger damping factor of g.

    When noting another definite integral of

    f, co w(blb to, g) dw = 一1}c(g)

    where c(g) =

    Cos-i

    l

    coshnvig

    for O S一 e〈1

    for g == 1

    for g>1

    a remaining link of

        2 c(g) [E (t) ] = f, co [w (t, fa) , e) ] dcke)

    caR be obtained.

         As evidenced in the foregoing formulas, the energy fed into simpie oscillators of

    multi-dirnensional isotropy permits to iRterpret consistently all the tensors of Arias integra-

    ted intensity, Husid time-axis growth, Fourier spectral modulus, time-dependent frequency

    content and the likes of them. Thus the mathematically defined tensors can be assigned

    their iRdividual roles within the framework of structural dynamics. An illustration in Figure

    1 is intended to demonstrate such advantages wheR applied to the analysis of 2-dimensional

    earthquake ground motions. This includes [W’(t, o, e)]”2 and [W(oo, bl, g)]’i2 under

    g =O.05 as well as [E(oo)]i’2. A closed curve at sampled points along time and frequency

    axes stands for the locus obtained by tracing the tip of vectors that specify directional

    components of tensor. Appealingly the 2-dimensional characteristics of available energy are

    seen to differ in complicated features over the combined domajns of time and frequency.

    These findings highlight the fact that the overall tensorial rneasure of [E(oo)] becomes of

    little use for describing specific sit囎tions of multi-dimensional dynamics.

                    A Note on the Cross-Correlation Components

                            of Complex-Valued Tensors

         The previous definitions of [P(t, ao)] aRd [P’(t, ch), g)] were based on deletion of

    imagiRary part in complex-valued matrices. Then a question may arise concerning what tke

    deleted part implies. Such complex matrices can be conditioned, in general, to be Hermitian

  • 5        ROLE OF TENSORIAL FORMULATIONS IN CHARACTERIZING

    THE POINTWISE MULTI-DIMENSIONAL PERFORMANCE OF DYNAMIC SYSTEMS 5

    and non-negative definite tensors. Furthermore they were composed of tensor product of

    vectors in the deterministic instances. The latter condition of tensor-product decomposition

    is however not necessarily assumed in the following, which allows to include stochastic

    problems.

         For a tensorial Hermitian matrix, [R], it goes without saying that its symmetric real

    part is tensorial. On the other hand, its skew-symmetric imagiRary part is under differeRt

    circumstaRces during rotation of coordinate axes. According to its component representa-

    tion of

    [R]= [p,,P一”jq,2 Pi2 p+,,jq’2], 1 pi2 P一,’i2q,, P’2 p+,,lq’2 Pp3,1 一+ /iqq3,l

                                   P3i十jq3i P23-jq23 P33

    in 2-dimensional and 3-dimensional instances, respectively, the single component of qi2 iR

    2-dimensional instances turns out to be invariant, while the three-component set of (q23, q3i,

    qi2) in 3-dimensional instances, when picked up and arrayed in the current manner, is

    subjected to a vectorial transforination. Discussiens are henceforth focused upon siinpler

    2-dimensional instances due to space limitation.

         Among the 4-dof parameters of [R], invariants consist of pii十p22, pnp22 rmpi22 and

    qi2. lts positive definite condition is given by pii十p22 〉 O and piip22-pi22 〉 qi22. Also

    this is not positive definite but noR-negative definite if and only if pii十p22 〉 O, piip22 rmpi22

    : qi22 and qi2# O, by disregarding trivial cases that reduce to 1-dimensional problems.

    Under the two situations combined, normalization by use of the principal axes identified for

    the real-part £ensor leads to

    [R] == psnajor

    = Pmajor

    where pmaior (>O),

    ([

    costh

    SiRlbn

     cosV 一sinth

      sinth costh

    pminor (>e) and

         ][

    1

    0

      1

    -jrσr

    エ]L:彫溜」+[

    ji9 ][COSψ si妙msin&th cosyb]

    一j,eメ@jr盾п@])

    V stand for major and minor principai values and

    orientation angle of major axis, respectively, and

       r = ./L121’!ll!}9Li’iOr (O 〈r$ 1)

             Prnajor

       cr m r==llll’L=2 =:一:==llli’l12 (ldl$1)

                         幽

    Without }oss of generality, a particular case of th x O is chosen below.

         Decomposition of the above tensor into a tensor product of mutually complex conju-

    gate vecters :

        [R] = {V}{V}T

  • 6 Haruo TAI〈lzAWA 6

           With {V} = Vi5filEi51r(一」, lg..) eXP(」ip)

    is strictly contingent upon 」61 = 1, and a single phase factor of di remains indeterminate

    there. On the contrary, its Choleski decomposition :

        [R] = [L][L]T

           w・・hこ・1乖[一喜。∴](・1・w・・…ang・1・・)

    becomes a}ways possible under the non-negative definite requirement. Associating this with

    a stochastic phase vector of L exp(jdii) exp(jdi2) J where qsi and q52 are random variables

    satisfying E lexp(」(dii-di2))1 = O, an aiternative expression of the latter decomposition is

    given by

        [R] == E{{V・}{V,}T]

           with {v’} = miiiEi6: (,(ri.,.p(jipesP+(jrlt(lil-gE一) ,.p(jip,))]

    Thjs indicates that even the tensor-product decomposjtion can be free from any requisite

    when considering the problem in a stochastic sense. The dimensionless factor of 6 is to

    describe therein the degree of statistical dependence or independence observed between

    major and minor components. Restricted to completely dependent situations of if =±1,

    only a single phase factor of dii is retained which corresponds directly to the deterministic

    cases. EveR though the role of d is totally inconceivable in deterministic problems save for

    the ambiguity of its sign (no more than the relation of  q12=±面『〉, such

    a stochastic extension will certainly merit some attentions.

                               GeOlne亡r量。 Simplic量ties

                     Evidenced in the Tensor Fields of Response

         Consider a multi-dimensional time history, {b(t) }, which is specified by a linearly

    combined set of multi-diinensional time histories :                                        ’

       {b(t)}” 2 [t(3u)]{id(t)}

                i

    These time historles can be either deterrrainistic or stationarily stochastic with constant

    rectangular matrices of [」(6u)]. Then form a matrix of squared RS or RMS intensity,

    [B], for {b(t)}.

    [B],.

    m(1一“ie,O,lb,il’,’,lb,il’,’Tdt ::,e:1:[1],’ll,d8.triM.:liYZtochastic

  • 7        ROLE OF TENSORIAL FORMULATIONS IN CHARACTERIZING

    THE POINTWISE MULTI-DIMENSIONAL PERFORMANCE OF DYNAMIC SYSTEMS 7

           =ΣΣ[1(βU)][ijR][j(βU)P『

              i l

           w…e〔・jR]迂二..{・d(・)}{・d(・)}・d…E【{・d(・)}{・d(・)}・】([・・R]一[・jR]・)

    Such relations of a practical impertance are found in multi-dimensional performance of

    systems for which the classical modal formulation becomes applicable under a multi-

    dimensional and vectoria1 ground acceleration of{a(t>}. In these instances,{b(t)}stands

    for a subset in the rnulti-dimensional response,{id(t)> and £i(βu)] coinciding with the

    motion of i・th order modal oscillator of multi-dimensional isotropy and the associated matrix

    of modal participation factors, respectively. Following the equation of motion of

           ロ                                          ゆ

       {歪d(t)} 十 2iζiω{韮d(t)} 十iω2{肇d(t)} = 一{a(t)}

    the matrices of [i∫R] (unsymmetric for i≠」)are characterized by their tensorial nature.

    Still the resuiting symmetric matrix of[B]may or may not be a teRsor.

         With a view to demonstrati捻g advantages in applying the above simple formula, taken

    up is a single-story rigid-floor system of two-way eccentricity subjected to a 2-dimensional

    ground acceleration. Response of 2-dimensional vectorial drift at an arbitrary point(x, y)

    on the horizontal plane is assigned to {b(t)}。 The x and y axes are, for specificity, oriented

    according to principal axes of its overall translational stiffness with the origin located on its

    gravity ce簸ter. Then

    [i(iCiiU)] = det.e.2-t一,E 2 (一i(S8..一nvYx) ILiEy 一]exJ (i =: 1 to 3)

    in which iex =iex/i., ie,=ie,/im, X =x/im and Y==y/im with (iex, ie,) and i. representing

    the position of i-th order modal center of twist and the radius of gyration, respectively.

    Substitution of this expression into the preceding formula yields

    [B](=r[BB.Xx, BBX,Y,]) == ,2.},,S, ,jA(一i(%..;1rmYx)ILje,一y 一(」ex-x)」

           W…ellA一(1+ie.2+ie,2)1(、塙、塙、)L…「・・」[・・R]{甑}

    Differing from [ijR], symmetry upon interchangiRg indices, jiA= i」A, features the current

    riioda} factors of ijA. A rnore compact formulation for three componeR£s of the symmetric

    tensor of [B] is

       Bxx = a(Y’By/a)2 十 (7..一rs,2/a)

       B., a 一cr (R-13k/cr) (Y’rsy/ev) 一 (yxymfgx/3y/a)

       Byy・:α(文一盈/α)2+(管ゾ底2/α)

            where ev = iiA 十 22A 十 33A 十 2(i2A 十 i3月目十 23A)

                   p. == ie. nA 一1一 2e. 22A 十 3e. 33A

                               十 (,e. 十 ,e.),,A 十 (,e. 十 ,e.),,A 十 (,e. 十 ,e.),,A

  • 8 Haruo TAKIZAwA8

    6, = ,e, ,,A 十 2e, ,,A 十 3e, ,,A

               十(16y十26y),2A十(16y十36y),,A十(26y十緯y)23A

    7.. == ley2 llA 十 2ey2 22A 十 3ey2 33A

               十2(,ey2ey l2A十潭y3ey13A十26y3ey23A)

    7.y= le. ley nA 十 2e. 2ey 22A 十 3e. 3ey 33A

               十 (,e. ,e, 十 ,e, ,e.),,A 一Y (,e. ,e, 十 ,e, ,e.),3A

               十 (,e. ,e, 十 2e, ,e,)23A

    7yy :lex211A十26x222A十二x233A           十 2(,e. ,e. ,,A 十 ,e. ,e. ,,A 十 ,e. ,e. ,,A)

    Therefore the distribution of [B] on the horizontal plane is seen to be quadratic indicating

    the existence of a center of response at (B./a, 6y/a). Upon shifting the origin of R and Y

    axes to this characteristic point, three components of Bxx, 一B., and B,, consist, in addition

    to separately different constants, of only the single 2nd-order term of Y2, xy and x2,

    respectively, under a common coefficient of a. Moreover the parameters of a, Px, 6y, yxx,

    7xy and 7yy are to be reiated simpiy to the response of modal oscillators reflected upon [ijR] .

         A closer mathematical examination concerning the tensor fieid of [B] developed on

    the horizontal plane leads to the following finding of interesting rules. When noting the

    contour lines drawn by the principal values of pointwise tensors, they form a family of

    confocal quadratic curves and, at the same time, a family of orthogonal trajectories. More

    specifically the curves for major-axis and minor-axis components are elliptic and hyperbolic,

    respectively. Another markedly simple feature becomes also apparent in the flow lines

    describing the orientation of the principal axes of pointwise tensors. Actually the latter can

    be shown to coincide strictly with the above family of quadratic curves.

         By use of the same ground motion as in Figure 1, results of an example study are given

    in Figure 2. The three systems examined therein have an identical relative stiffness for their

    coupled lateral and torsional motion, only the absolute stiffness being designed to provide

    different fuRdamental periods of O.3, 1.0 and 2.5 seconds. lts upper-half part is intended to

    illustrate the 2-dimensional nature of response drifts at sampled poiRts. lndividual closed

    figures represent directional properties of the RS tensor of[B]1/2 in a si面lar way as魚

    Figure 1. This includes [iiR]ii2 shown on the same scale for comparisons. More complete

    data are presented in the lower-half part which involves full iRformation on the tensor field

    by means of the above-noted contour and flow liRes. The geometric simplicities observed

    there are striking enough to highlight an advantageous role of the tensor formulation.

    Numerals discriminating each co就our line stand for the魚creasing or decreas魚g factors

    compared to the major-axis component in [iiR]i’2, thus clarifying the effects of torsion.

  • S90W

    SgOIV

    $.

    Eo.

    ew.

      [エエR]

    ユ!2

    ee Z1 

    ct” se

    cit2

     S 

    geW

    yβ吊

    蒙,

     3

    @2 @1 @0 |1 ?2

      

     ρ

    @ ’

    f

      

    、@ 

     、

    @   、.’xS 、も

    ・、

    3’G/

    C

    一2 

          -1 

          0 

           1 

           2 

           3

    嶺x/

    ¢

    fundamental 

    per

    iod = O.

    3sec

    NOO 

    VV

    yμm

    @3S90W 

      

      

      

      

      

      

      

      

      

     F

    、 

    、 _一

    G.3

    、隔

      

    .R

    D(》

     1Eら

    く》

      

    、 

      

     賎

      、

    ’F

    込・ A 

    ◎・ 亀♂

      

    wS 

    ♂ ◎.も◎

    @◎.

    9

     2

    @! @o 齬ケ

    ヨ2 ス

    G 

      

    ,c

    、・ Aさ“、

    3 、3

    、.戟 A》、至湛

    )、

    “’

    @’2 

    ”1 

    O PI

      

      

      

      

      

    wfundamental

     pe

    riod 

    = e.3s

    ec

    O.25

    O.2

    q

      

     ・

    鮨る

    →2E

    2急

     3・

    ノi。

    L,i

    @NOO

    W

    S90W

    eNoevv

      [ilRl112

    ,’9

    @、、

    ノ 

      

     監

    乳、

      

     ノ

    へ 

      

      

     ほ

    ∠1。

    』、ecl,

    2

    yκ脇

    ・茎・E

     3

    @2 @i ,’

    f

      、

    wS 、

    ・、

    3EG,

     ’

    一2 

          -1 

           G 

           1 

           2 

           3

    袖xノ

    fundamental 

    period 

    = 

    1.esec

    NOOV

    V

    y/

    ムSgG響

                     F

    oφ 、、

    _ \

    、  、R喰

    IE

    A

    、、 A

    \3

    一F

    、働(

    y\\

    _  \\ へ

      

     、

    h” @もや

    @  “   、・

     ’

    f・

    ;1融

    o.5

    O.45

    O.4

    O5も

    ?βm

    G  、さ

    b/\

    、曳 A♪

    3E ・,

    \》 ψ 、

    ・1・

    ・㍉

    8.認

      

      

      

     ぎ

    一1 

    。窪

    1急2 

      

      

      

      

      

      

     切

    fundar罰

    ental 

    period=

     L 

    O sec

    ♪イし、3塾

    x NOOW

    s ge}

    v

    S90W

    命一

      

    [uR

    コ1!

    2

    ナグ

    @ らロ

    ㊤ 41,_。2

    y/名m

    十1三

    q

     3

    @2 @1 @0 黷P

    r2

      ,’

    @’ 

    、、

    wS\

    \3’

     、

     G/C  ’

    f

    →騨

    2  

    -1

    o1

    fundamental 

    period

     = 2.

    Ssec

    S90W

    yfi 3 2 1 o

    一1

    一2

    23 

    X/im

     NOOW

    γ、

       F 

    ◎’

    C  

     、

      

      

      

    、、

    十1E 

    、駅

    、、

    嶺F

    も● @\

    \も

    ・ら

    奄 も

     、

    ~ミ

    ミ♂

     ミ

    @♂

     O.35

    ミ.3

    A¢鳶

    。 

    ’鳶

    ¢

    @四か

    2

    Gl、

    3E、

    書ノ

    C

    島, フ へさ へ3 、♪!

    A♪

    .、

    置 フ “h 

    .2 

    一1 

    o.e. 1 92

     も

    ・ 

      

      

      

     a 

      

    渉fundamental 

    period

     = 2.

    5sec

    3るx/

    im

     Ne

    ow

    Fig

    ure 

    2

    G :

    grav

    ity 

    cent

    er 

    C;in

    erti

    a ci

    rcle

     S:

    shea

    r ce

    nter

     iE

    :i-

    th 

    orde

    r ei

    genc

    ente

    r (

    i =

    1 to

     3)

    R: ce

    nter

     of

     re

    spon

    se F:

     co

    mmon

     fo

    ci fo

    r th

    e co

    nfoc

    al etllpses and 

    hyperbolas

      

      

    : 

    contour 

    for 

    major-

    axis 

    compenent.

     and 

    elso 

    Tine 

    of 

    flow 

    for 

    orientation of major axis

    ..一 : co

    ntou

    r fo

    r mi

    nor-

    exis

     co

    mpon

    ent.

     an

    d al

    so li

    ne of flow for 

    orientetion 

    of 

    minor 

    axis

    o 臣’ 羨1

    躍 毒 糞 馨舅

    年 緊 華 箋 舅1

    蒙.z・ 諺 o

  • S 90

    W

    S90W

      

     ide(

    t)

    t’

    ls /1

    .’tt

    t

    1 cm

    yβ田

    ÷1E

     3

    @2 @1 

     ,

    ’  、、   、×s 

    、3ε

    G/C

    噛i2

    一1

    S90W

    S90W

       id

    e(t)

    t

    o

    ’’

    ミ Z.s

    1

    fun

    darne

    nta

    l pe

    riod 

    = 

    e.3sec

    23 

    Xfim

      

    Nee

    W

    5 cm

    ÷1E

    y/ゼ田

    @3 @ 2

    @ 1

    @ 0

      ,’

    @’ @’

      

    陶@ 

     、

    wS 

    、、

    、 3E

     G 

      

    ,’

    f

    C

    一1

    |2

    鴨2 

      

     -

    1  

      

    G  

      

     1 

      

      

    2  

      

     3

    →xκ

    sgew

    y/im

    3 2 1 o

     sgew

      

     id

    e(t>

    t

    ロも 乙

    s’

    t一

    fund

    amen

    tal 

    peri

    od = 1.

    Osec

    NoeW

    一1

    一2

    ’t

    5 cm

    m

    牽IE

     ’

    @’ 

    、、

    @、、

    ~S 、

    \,『

    、、

    や一2

    一1

    o1

    fund

    amen

    tal 

    period 

    = 

    2.Ssec

    23 

    x/tm

      NOOW

    胃 ま 〉 峯 N > 芝 〉

    Fig

    ure 

    3

    6

  • 11        ROLE OF TENSORIAL FORMULATIONS IN CI{ARACTERIZING

    THE POINTWISE MULTI-DIMENSIONAL PERFORMANCE OF DYNAMIC SYSTEMS Il

                       Pointwise Directional Characteristics

                         in the Peak Amplitudes of Response

         The current tensorial approach is iRdeed quite useful for understanding important

    trends behind a messy appearence of the pointwise directional characteristics in multi-

    dimensional response and their spatial distribution. However this does not necessarily

    comply with the conventional notions in the practical design of earthquake-resistant struc-

    tures. From the latter standpoints, envelope of Lissajous’ locus including the associated

    rotational properties in orbital motion, for example, may be more preferable instead of the

    RS or RMS tensor. With such lines of extension in mind and by keeping still a straight-

    forward relation to the tensorial understanding, studied hereafter is the pointwise directional

    dependence of peak amplitudes.

         The examination follows the setup in Figure 2, and its immediate concern is directed

    toward peal〈 amplitude in a 1-dimensional time history, be(t), extracted from {b(t) } along

    a direction of e on the horizoRtal plane ;

       be(t) :Lcose sinOj{b(t)}

              (also, ide(t) = L cose sinO J {id(t)} supplementarily]

    Under complete lack of operationa} ease, clumsy repetition of ad hoc evaluations must be

    continued along each direction as well as at each point.

         Figure 3 summarizes results of the examination concerniRg the variation of peak

    amplitude depending upon 0 and the associated field developed on the horizontal plane.

    Employing a preseRtation form corresponding to Figure 2, direct comparisons are intended

    therein between RS values and peak amplitudes. For both pointwise directional and fieid

    characteristics, discrepancies are seen to be minor enough from practical points of view.

    Hence it is concluded that essential features in the non-tensorial properties of peak amplitude

    may be replaced by the tensorial formulation. Note tkat the directional distributioR of peak

    amplitudes is, in general, inconsistent with the envelope of Lissajous’ orbit.

    BIBLIOGRAPHY

    1. Takizawa, H., “Work done by earthquake ground shakings upon various types of

       dynamic systems,”Proceedings(ゾthe 6th/4釦%Eart勿襯舵Engineeri’㎎Symposizam,

       Dec. 1982, pp. 1065-1072.

    2 . Tal〈izawa, H., “Coupled lateral and torsional failure of buildings under two-dimensional

       earthquake shakings,” Proceedings of the 8th World Conference on Earthquake Engi-

       nee7img, July 1984, Vol. IV, pp. 195-202.

    3. Tal〈izawa, H., “Toward a better understanding of the torsionally-coupled dynamic

       performance of buildings,”Proceedings Of the 7th/4勿%.Eart勿uake En8ineering Sympo一

  • 12 Haruo TAKIzAwA 12

    4.

    sium, Dec. 1986, pp. 1789-1794.

    Takizawa, H., “Nature of the effects of torsional coupling examined on the tensor field

    of deformation response and compared to the current regulations of seismic design,”

    ノ∂urnal(ゾStructural Engineeri’ng, Arch. Inst. Jap., Vol.33B, March 1987, pp.111-124,