tessler 1993 qualitative structural analysis diagrammic reasoning

Upload: unknown

Post on 24-Feb-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    1/10

    Qual i t at i ve Structural

    Anal ysi s

    Usi ng Dagrammati c

    Reasoni ng

    Shi r l ey Tessl er Yum

    I wasaki

    and

    Ki ncho

    Law

    Knowedge Syst ems

    Labor at or y

    Stanf or d Uni versi t y

    7 1Wel chRoad Bl dgC

    Stanfor d Cal i f orni a

    943 5

    i wasaki @ksl

    . stanf ord

    ed u

    1

    I nt r oduct i on

    Ci vi l

    Engi neer i ngDepartment

    Stanf or d

    Uni versi t y

    Terman

    Engi neer i ng

    Cent er

    Stanf or d Cal i forni a

    943 5

    l aw@i ve. stanf ord

    . ed

    u

    Abst r act

    Di agrammat i c r easoni ng i s

    a

    type of

    r easoni ng i n

    whi ch t he

    pr i mary means of i nf erence

    i s

    t he d i r e c t

    mani pul at i on

    and

    i nspecti on of a

    di agr am

    Di agrammat i c r easoni ng i s

    preval ent

    i n

    human

    probl em

    sol vi ng behavi or

    especi al l y

    f or

    pr obl ems

    i nvol vi ng spat i al

    rel at i onshi ps

    among physi cal

    obj ect s

    Our

    research examnes t he

    rel at i onshi p

    bet ween

    di agrammat i c

    r easoni ng

    and

    symbol i c

    r easoni ng

    i n a

    comput at i onal

    f r amework

    Wehave

    b u i l t

    a

    system cal l edREDRAW t h a t

    emul at es t he human capabi l i t y

    f or reasoni ng w t h

    pi ctures i n

    c i v i l

    engi neer i ng

    The

    c l a s s

    of

    s t r u c t u r a l

    anal ysi s

    pr obl ems

    chosen

    provi des a

    r e a l i s t i c

    domai n

    whose sol ut i on

    process

    requi res

    domai n- speci f i c

    know edge

    as

    wel l

    as

    pi c t or i a l r easoni ng s k i l l s Wehypot hesi ze

    t h a t

    di agrammat i c

    r epresent ati ons such as

    t hose used by s t r u c t u r a l engi neer s

    provi de

    an envi r onment

    wher e

    i nf er ences about t he

    physi cal r e s u l t s of pr oposed s t r u c t u r a l

    conf i gur at i ons

    can t ake

    pl ace

    i n a more i n t u i t i v e

    manner than t h a t possi bl e

    t hrough

    pur el y symbol i c

    r epresent ati ons

    Humans often use

    di agr ams

    t o f a c i l i t a t e pr obl em

    sol vi ng I n many t ypes of

    pr obl ems

    i ncl udi ng

    but not l i m t ed t o

    pr obl ems

    i nvol vi ng behavi or s of physi cal

    o bj e ct s dr aw ng a

    di agram

    i s a cruci al st ep i n t he

    sol ut i on

    process

    Drawng

    can

    reveal i mpor t ant

    i nf ormati on that may not

    be

    e x p l i c i t i n a wri t t en descri pt i on

    and can

    hel p one gai n

    i nsi ghts i nto t he nature of the

    pr obl em Though

    such use

    of

    di agrams

    i s

    an

    i n t e gr a l

    part

    of human

    pr obl em

    sol vi ng

    behavi or

    has not recei ved nearl y

    as much

    at t ent i on

    i n

    I as

    symbol i c r easoni ng has

    One

    i mpor t ant

    advant age of di agr ammat i c

    r epr esent at i on

    i n some

    t ypes

    of

    pr obl ems

    i s

    that

    makes

    expl i c i t t he s pa t i a l

    rel at i onshi ps

    t h a t

    mght requi re ext ensi ve

    search and

    numerous

    i nf er ence steps t o determne usi ng a

    symbol i c

    r epr esent at i on

    Lar ki n and

    Simon

    have

    shown t h a t even when t he

    i nf ormati on cont ent s

    of

    symbol i c

    and

    di agrammat i c r epr esent at i ons ar e equi val ent a

    di agr ammat i c

    r epr esent at i on

    can offer

    computat i onal

    advant age

    i n pr obl ems

    where s pa t i a l

    rel at i onshi ps pl ay

    a pr omnent

    rol e

    [ Lark i n

    Simon

    1987]

    Si nce

    humans

    r eason

    w t h so much

    apparent

    ease

    i n

    some

    pr obl ems

    a program

    ha t coul d

    reason d i r e c t l y w t h a

    di agrammat i c

    r epresent ati on woul d

    be

    moreunderst andabl e t o

    t he

    user than a

    program

    t h a t

    r easons

    excl usi vel y w t h

    a

    pur el y

    symbol i c

    r epr esent at i on of

    t he

    same

    i nf ormati on

    I n

    ad di t i o n

    a

    di agrammat i c

    r easoni ng

    program

    shoul d

    offer i nsi ght

    into

    t he

    rel ati onshi p

    bet ween di agr ammat i c

    r easoni ng

    and

    symbol i c

    r easoni ng

    Such

    a

    pr ogr am

    may

    al so

    be usef ul

    i n

    i mpar t i ng vi sual i zati on

    s k i l l s

    t o

    st udent s of

    di sci pl i nes wher e such a f a c i l i t y i s c r uc i a l

    such

    as i n c i v i l or

    mechani cal

    engi neer i ng

    and

    desi gn

    I n t h i s paper we present our work

    ai med

    t owar ds

    unders t andi ng

    t he rol e of

    di agr ammat i c

    r easoni ng i n

    probl emsol vi ng

    The

    pr obl em

    we

    chose

    f or

    st udyi ng

    di agr ammat i c

    r easoni ng i s t h a t

    of

    det ermni ng t he

    def l ecti on

    shape

    of a bui l di ng

    f r ame

    s t r u c t u r e under

    240

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    2/10

    l oad

    W

    have

    const r uct ed a

    computer

    programcal l ed

    REDR W

    Reasoni ng

    w t h

    Daw

    ngs

    that

    sol ves t h i s

    probl emqual i tati vel y

    usi ng a

    di agrami n

    away s im lar

    to

    humanengi neers

    1

    Rol es of

    di agram

    i n

    Probl em

    Sol vi ng

    Some research has been

    done on

    t he

    rol es

    t h a t diagrammati c

    r easoni ng pl ay

    i n human

    probl em

    sol vi ng

    Novak

    and

    Bul ko,

    [Novak

    Bul ko

    1992] , f or

    exampl e,

    have

    asserted

    t hat

    a di agram

    and

    i t s annotat i ons

    serve

    as a short-termmemory devi ce i n t he probl em

    sol vi ng

    process

    Such a

    devi ce al l ows t empor ar i l y- needed i nf ormat i on to be retr i eved

    l ater i n t he

    same manner that wri t i ng

    down i nt ermedi at e

    resul t s i n

    mul ti pl i cati on

    probl em

    f rees

    t he per son to performf urt her cal cul ati ons

    They

    al so post ul ate that a

    di agrammayact

    as

    a substrate

    or

    concept

    anchor t hat al l ows t he

    new

    part of a

    probl em o

    be descr i bed rel ati ve to

    wel l - under st ood probl em

    base

    Larki n and S mon di scuss

    extensi vel y t he

    advantages

    of di agram f or

    f aci l i tat i ng i nf erence about t opol ogi cal

    or

    geometri c r el at i onshi ps [Lark i nmon

    98 ]

    Chandrasekaran and

    Narayanan

    [Chandrasekaran

    Narayanan

    1992] , Novak

    and

    Bul ko

    [Novak

    ul ko 1992] , Borni ng

    [Borni ng

    1979]

    and

    others have al so poi nted out the usef ul ness of di agram to human

    probl em

    sol vers

    as a devi ce

    t o

    ai d i n vi sual i zati on, gedanken

    exper i ment s

    or predi ct i on

    Fi nal l y,

    Novak

    and

    Bul ko

    [Novak

    ul ko 1992] ,

    Koedi nger

    [Koedi nger 1992]

    and

    others

    have

    expl ored

    t he

    i dea

    that

    di agram

    maysometi mes

    be

    used

    not pri mari l y

    f or

    maki ng base- l evel

    i nf erence, but rather

    to

    hel p i n

    t he

    sel ecti on

    of an

    appropr i ate method

    t o sol ve a probl em t hat i s as an

    ai d

    i n

    t he

    organi zat i on

    of

    cogni t i ve

    acti vi ty

    [Chandrasekaran et

    al

    1993]

    sal i ent

    f eat ure

    of

    di agrammati c

    r easoni ng

    i n many si tuat i ons

    i s

    i t s

    qual i t at i veness

    Peopl e r eason wth

    di agram

    to

    get r ough, qual i tati ve answers

    I f a

    more preci se,

    quant i t at i ve answer

    i s

    needed,

    t hey

    must

    resort to

    more

    f ormal ,

    mathemat i cal

    t echni ques

    However ,

    qual i tati ve

    t echni ques

    are ext r emel y

    useful

    i n gai ni ng

    val uabl e

    i nsi ght i nto the

    range

    of possi bl e sol ut i ons ni n i t i a l

    qual i tati ve underst andi ng

    t hus

    obt ai ned

    can gui de

    t he l ater anal ysi s

    f or more detai l ed answers I n t he cont ext of

    structural

    anal ysi s,

    knowng t he

    qual i tati ve

    def l ected

    shape al l ows

    one

    to i denti f y c r i t i c a l f eatures of t he

    shape

    One

    can

    t hen

    set

    up

    rel evant equati ons

    i n

    order

    to

    obtai n

    more

    preci se

    i nf ormat i on

    such as actual

    magni tudes

    of f orces

    and

    di spl acement s

    at

    speci f i c

    poi nt s of

    i nterest

    How

    do

    di agram actual l y hel p c i v i l engi neer s

    t o

    make qual i tati ve i nf erences? From

    studyi ng t ext books on

    el ement ary

    structural

    anal ysi s, such

    as

    [Brohn 1984] , that ai mto

    develop a i ntui t i ve

    underst andi ng

    of

    t he r esponse of t he structure under a l oad,

    we f ind

    that di agram f u l f i l l

    many of

    t he

    same rol es as t hose art i cul ated by researchers i n

    other

    f i el ds

    First,

    di agram

    are used as a vi sual l anguage of structural behavi or that

    can

    be

    understoodw t h t hemnimumof

    textual

    comments [Brohn

    1984]

    The l anguage al l ows

    t he engi neer

    to express

    expl i ci t l y

    t he const rai nt or physi cal

    l aw

    that

    i s

    rel evant at each

    part of t he

    proposed

    structure,

    i n

    such a way that t he const rai nt s

    and

    some of t he

    consequences

    are i mmedi ately

    apparent to t he r eader wthout f ur t her

    reasoni ng

    Secondl y, t he di agramserves as a

    pl ace

    hol der or

    short-term

    memorydevi ce by

    al l ow ng

    t he desi gner

    to

    sketch

    out

    t he

    resul t

    of

    one

    def ormat i on

    and

    t hen

    go

    back

    to

    see

    i f

    there

    i s

    a

    f urt her ef f ect

    or

    i nt eracti on

    that needs to be addressed

    Fi nal l y, vi sual

    i nspecti on of

    di agram seem to gui de

    t he

    engi neer i n

    choosi ng t he next

    st ep,

    resul t i ng

    i n a

    more

    ef f i ci ent

    probl emsol vi ng process

    t han

    i t woul dbe

    ot herw se

    Havi ng studi ed t he use of di agram i n al l t hese capaci t i es i n t he context of

    det erm ni ng

    deformati onshape of f ramestructures,

    we

    have

    constructed

    REDR W

    o

    use di agrams i n

    al l

    those capaci t i es i nways

    s im lar

    to humans

    W

    w l l f i r s t

    expl ai n

    t he

    def l ecti on shape

    24

    1

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    3/10

    probl em n Secti on The ar chi tecture of RE R W l l be descr i bed

    i n det ai l

    i n

    Secti on

    Def l ecti on ShapeProbl em

    Determni ng the qual i tat i ve def l ected shape of a f rame structure

    under a l oad i s a

    cr uci al

    step i n anal yzi ng t he

    behavi or

    of a structure Structural engi neer s

    f i r s t make

    asi mpl e, 2-

    drawng

    of

    t he

    shape of

    t he

    gi ven

    f rame

    structure

    Gven

    a

    l oad

    on

    t he

    structure,

    they

    modi fy

    t he shape of t he structural member under t he

    l oad

    They

    i nspect

    t he modi f i ed

    shapeto i denti f y t he pl aces where

    constr ai nts

    f or

    equi l i br i umof t he

    structure

    ar e

    vi ol ated

    Those

    const rai nt vi ol at i ons are

    cor r ectedby

    modi fyi ng t he shape of connected structural

    members

    pr opagat i ng

    def l ecti on t o other parts of t he structure Thi s pr ocess

    i s

    r epeat ed

    unti l al l t he const r ai nts ar e sati sf i ed

    The

    drawng thus produced

    shows t he

    f i nal

    def l ected shapeof t he f rameunder the gi ven l oad

    G ven a

    di agram

    of a f rame structure

    and

    a l oad,

    t he

    programproduces an under l yi ng

    symbol i c r epr esentat i on i n order to f i l i t t e

    r easoni ng about engi neer i ng concept s

    Then

    t heprogramw l l use i t s structural

    engi neer i ng

    knowedge t o pr opagat e

    constr ai nts

    on t he

    di agramof t he structure and

    w l l

    i nspect

    andmodi fy t h i s

    pi cture

    unti l a f i nal shape i s

    produced

    that r epr esents a stabl e

    def l ected structure under the gi ven l oad

    As w th t he qual i tat i ve nature of

    human vi sual r easoni ng, the r easoni ng

    carr i ed

    out by

    RE R W

    s

    al so

    qual i tat i ve

    The

    answer produces

    i s

    a

    pi cture

    of a

    def l ected

    shape

    A though

    t he resul t i ng pi cture i s

    qual i tat i vel y consonant wi th t he probl emsol uti on,

    i t

    i s

    not

    nor

    does

    i t

    need t o

    be

    mathemat i cal l y

    accurate

    or to scal e

    J 2

    Fi gur e

    1

    Steps i n det erm ni ng t he def l ected shape

    RE R Wsol ves t h i s

    t ype

    of def l ected shape probl em

    by

    di rect l y

    mani pul ati ng a

    r epr esentat i on

    of

    t he shape i n t he

    manner

    shown above A though t heprobl em

    coul d

    be

    sol ved by sett i ng

    up

    equati ons,

    vi sual i zat i on

    i s a i ndi spensabl e f i r s t step that pr ovi des

    an

    engi neer wi th an i ntui t i ve

    under st andi ng

    of

    t he

    behavi or

    of t he

    structure and

    enabl es

    her

    to r ecogni ze agood

    strategy

    f or further anal ysi s

    4

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    4/10

    Before descri bi ng howRE R W

    nal yzes

    structures, we expl ai n br ie f ly

    t he

    r easons

    f or

    our

    choi ce of t h i s def l ect ed shape probl em n

    advantage of t hi s c i v i l engi neeri ng

    probl em

    domai n

    f or st udyi ng t he r ol e of vi sual r easoni ng i n

    probl em

    sol vi ng

    i s

    t he

    f act

    t hat i t

    i s r i ch

    w t h

    domai n- speci f i c

    knowedge t hat has si gni f i cant i mpl i cat i ons

    on

    how

    t he di agram s mani pul ated

    and

    i nt erpret ed

    One possi bl e

    domai n

    i n wh ch t o st udy

    pi ctori al r easoni ng

    i s

    geomet r y, where pi ctures are abst ract di agram w thout bei ng

    a

    r epr esent at i on of

    anyt hi ng

    i n

    t he

    worl d

    I n

    geomet r y,

    t he

    onl y

    pr oper t y

    one

    r easons

    about

    i s

    t he

    geometri c

    pr oper t y There

    are

    no ot her

    t ypes of

    i nf ormat i on,

    apart

    f rom

    t hat

    r epr esent ed i n t he di agr am

    t hat

    one must

    t ake i n t o account when mani pul at i ng

    and

    i nspect i ng t he di agram

    I n cont r ast ,

    pi ct ures

    used f or

    r easoni ng i n

    engi neer i ng desi gn are not si mpl y abst ract

    geometri c shapes but

    act ual l y

    r epr esent t hi ngs

    i n

    t he

    real worl d

    Fur t hermore, how

    a

    pi ct ure

    i s

    i nt erpret ed

    and

    mani pul ated

    depends

    si gni f i cant l y

    onwhat

    r epr esent s

    For

    exampl e a l i ne i n

    our

    domai n

    r epr esent s abeamor a

    co umn

    hangi ng

    t he

    l engt h

    of

    t he

    l i ne woul d change

    t he

    i nf or mat i on

    r epr esented

    by

    t he

    di agram

    I n

    a

    ci rcui t

    di agr am on

    t he ot her

    hand one

    coul d change t he l engt h or curvature of t he l i ne

    r epr esent i ng

    an

    el ectr i cal

    connect i on

    w t hout

    changi ng t he i nf ormat i onal cont ent of t he di agram

    For

    t he

    goal of bet t er

    unders t andi ng

    t he

    rol e of vi sual reasoni ng i n probl emsol vi ng and

    i t s

    rel at i on

    t o

    symbol i c

    r easoni ng,

    i t

    i s

    i mpor t ant

    f or

    us

    t o work

    w t h

    a

    probl em

    requi r i ng

    a

    weal t h

    of

    domai n knowedge

    t hat

    has si gni f i cant

    i nf l uence

    on t he waydi agram are

    used

    and

    i nt erpret ed

    3 Archi tecture

    of thesystem

    Fromexamni ng

    t he

    way def l ecti on shape probl em are sol ved by

    humans

    i s

    appar ent

    t hat sol v i ng

    t h i s

    t ype

    of probl em

    r equi r es

    not

    onl y

    an abi l i t y t o mani pul at e

    and i nspect

    di agram

    but

    al so

    subst ant i al structural

    engi neer i ng

    knowedge St ruct ural

    engi neer i ng

    knowedge

    about t he

    propert i es of vari ous

    t ypes

    of j oi nts

    and

    support s

    i s

    necessary

    t o

    i dent i f y

    const rai nt s on t he shape f or

    t he

    structure

    t o be i n

    equi l i br i um Suchknowedge

    i s best r epr esent ed

    and

    mani pul ated

    symbol i cal l y

    On t he

    ot her

    hand,

    i nf ormat i on

    about

    shapes

    i s

    best

    r epr esented

    as

    a

    pi ct ure

    Many

    t ypes

    of

    modi f i cat i on

    and

    i nspect i on

    of

    t he

    shape

    are al somore easi l y carr i ed out w t h api cture

    The requi rement f or bot h pi ctori al

    and

    non- pi ctori al represent at i on and reasoni ng

    suggest s

    a

    l ayer ed archi t ecture

    Thus,

    RE R W

    ncl udes

    both symbol i c r easoni ng and

    di agrammati c

    reasoni ng components

    The

    former

    cont ai ns t he

    knowedge base of

    structural engi neer i ng knowedge

    about var i ous

    t ypes

    of

    structural

    members j oi nt s,

    support s,

    and

    t he

    const r ai nt s

    t hey

    i mpose

    on t he structure

    t

    al so

    i ncl udes a

    const rai nt -

    based i nf er ence mechani sm

    o make use of t he knowedge

    The l a t t e r

    diagrammati c

    r easoni ng component i ncl udes

    an i nt ernal

    r epr esent at i on

    of t he

    t wo- di mensi onal

    shapeof

    t he f rame structure as wel l

    as

    a

    set of

    operat ors t o mani pul at e

    and

    i nspect t he shape

    These operat ors, some of wh ch are shown i n Fi gure

    2,

    corr espond t o

    t he

    mani pul at i on

    and

    i nspect i on operat i ons

    peopl e

    perform

    f r equent l y

    and

    easi l y w t h di agram wh l e

    sol vi ng def l ect ed- shape

    probl em

    TheStructure

    Layer

    cont ai ns

    a symbol i c

    r epresent at i on of domai n- speci f i c

    knowedge t

    r epr esent s

    non- vi sual

    i nf ormat i on such as

    hi nged

    j oi nt rotat i on ,

    var i ous

    t ypes

    of

    structural

    members equi l i bri umcondi t i ons, as wel l as heur i s t i c knowedge

    f or

    cont rol l i ng

    t he

    structural

    anal ysi s

    pr ocess

    TheDagram

    Layer r epr esent s t he two dimensi onal shape of a structure There are

    several operat or s t hat di rect l y act on t h i s

    r epr esent at i on

    t o

    al l ow i nspect i on

    as

    wel l

    as

    243

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    5/10

    t r ansf ormat i on of t he

    shape These operators

    correspond

    to the operat i ons people

    perform

    easi l y

    w t h

    di agram The

    i nternal

    represent at i on of a shape i s a combi nat i onof

    a

    bi tmap

    whose el ements

    corr espond

    t o each

    poi nt

    i n a pi cture, and a

    more

    symbol i c

    representat i on where

    each

    l i ne

    i s r epresent edby a set of

    coordi nate poi nt s

    The D agramLayer i s i ndependent of t he structural

    engi neer i ng domai n

    i n t he sense t h a t

    i t does not cont ai n

    any

    structural engi neer i ng concepts

    However

    t he types of

    both

    mani pul at i on

    and

    i nspecti on operators

    provi ded

    f or the

    l ayer

    refl ects

    t he

    r equi r ement s

    of

    t he

    domai n

    For exampl e t he assumpti on that t he f rames consi st of i ncompressi bl e

    members

    made apart i cul ar set of operators necessary e. g t he program equi res a

    bend

    operator

    but not a stretch or compress

    operator) , and

    al so

    by

    the speci f i c f uncti oni ng of

    t hose

    requi r ed

    operators

    f or exampl e, t he

    bend

    operator creates amoderate curve rather

    t han

    a

    compl ete

    bend

    t hat

    woul d

    cause t he

    l i ne

    endpoi nt s

    to

    touchor cross

    ; or,

    t he i nspect

    operator

    may

    l ook at components connected to

    t he component

    i n

    questi on, but w l l not

    compare

    that

    component

    to any other, as

    mght i n

    some

    other domai n

    Structure

    Layer

    Obj ect s :

    beams,

    col umns, connect i ons,

    supports,

    l o ad e t c

    Operators :

    generate- f orce-equi l i bri umcondi t i ons,

    generate- moment - equi l i br i umcondi t i ons,

    et c

    Dagram

    Layer

    Obj ect s :

    Operators :

    l i n e s s pl i n es c i r c l e s

    Mani pul at i on r o t a t e bend, t r a n s l a t e smoot h,

    e t c

    I nspecti on

    get- angul ar - di spl acement , get- di spl acement ,

    symmet r i c

    a l p,

    et c

    Fi gure 2 Types of

    obj ects

    andoperat or s i nREDR Wrogram

    There i s a

    cl ose

    l i nk between

    t he i nf ormat i on i n t he two l ayers The systemrel ates t he

    represent at i on of

    a part i cul ar beami n t he

    Structure

    Layer to a spl i ne i n t he

    D agram

    Layer,

    and

    t he concept of

    def l ecti on of a

    beam

    to an operat i on on aspl i ne t o transformi t s

    shape Li kewse, t he system

    i s

    abl e to i denti f y f eat ures

    of a shape e g

    di recti on

    of

    bendi ng, exi st ence of an

    i nf l ecti on

    poi nt )

    and

    to communi cate

    them

    to

    t he

    Structure

    Layer

    Communi cat i on between

    t he two l ayers

    takes

    pl ace

    by sendi ng commands and posti ng

    constr ai nt s

    by

    t he Structure Layer,

    whi ch

    i s carr i ed

    out

    or checked

    by

    t heD agram

    Layer

    Fi gure

    5 shows

    t he t wo- l ayer ed

    archi tectur e schemat i cal l y There i s a

    t ransl ator

    between

    t he two

    l ayers

    t o

    medi ate t he communi cat i onbetween t he

    two

    l ayers

    Went he

    Structure

    Layer

    posts

    a constrai nt or a

    command

    t he Transl at or

    t ransl ates i t

    i nto a cal l to a

    D agram

    Layer operator

    t hat

    can di rectl y act on t he r epresent at i on of t he shape to

    mani pul at e

    or

    i nspect The

    resul t

    i s agai n

    t ransl ated back to concepts that the Structure

    Layer underst ands

    STRU TURE

    L YER

    Def l ect

    B1 Ai r down

    00

    D GR ML YER

    BendB pi c : y

    D agram

    Representati on

    Fi gure

    3 Two l ayered archi t ecture of t he

    REDR Wrogram

    24

    4

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    6/10

    The

    RE R W

    rogramhas

    been i mpl ement ed and has

    successful l y

    anal yzed

    s i x

    of t he

    3 basi c def l ected

    shape

    probl ems

    descri bed

    by Al l en [A l len

    1978]

    An i nf ormal

    eval uat i on by a c i v i l

    engi neer shows t h a t t he

    pr ogram

    r e f l e c t s t he

    q u a l i t a t i v e

    reasoni ng

    process used

    i n anal yzi ng

    f rame s t r u c t u r e s , and

    t h a t i t woul d be usef ul i n hel pi ng

    students

    and

    novi ce

    engi neers

    l earn

    t o

    sol ve

    t h i s

    type of

    probl em

    3. 1 Exampl e

    I n t h i s s e c t i o n,

    we

    i l l u s t r a t e

    t he probl emsol vi ng

    process byRE R W

    i t h t he exampl e

    present ed ear l i e r i n Fi gure

    3

    Wi l l u s t r a t e the type of

    communi cat i on t h a t

    takes pl ace

    between t he l ayers G ven t he

    f rame

    s t r u c t u r e of

    Fi gure

    4 a ,

    wi th a l o ad ,

    Load3,

    pl aced on

    i t ,

    t he Structure

    Layer, S

    sends a command,

    Def l ect

    Beam3 i n

    t he

    same di recti on as the l oad, whi ch

    t he

    Transl ator, T

    t r a ns l a t es

    into

    an operat i on

    Bend

    Beam3 p i c i n t he negat i ve

    di recti on

    of

    t he y- coordi nat e . Carryi ng

    out

    t h i s

    operat i on

    wi l l

    r es ul t i n t he shape shown i n

    Fi gure

    4 b)

    S i n f e r s t h a t si nce J oi nt3

    i s

    a

    r i gi d

    j o i n t , Beam3

    and Col umn3 must

    r emai n

    perpendi cul ar

    t o each other at J oi nt3

    S

    i s s u es a query t o t es t t hi s

    const rai nt The

    query

    i s tr ansl ated into

    get t he angl e

    between

    Beam3. p i c

    and

    Col umn3. p i c

    at t he ends

    connect ed

    by

    J oi nt3

    . p i c

    f o r

    t heD agram a y e r , The

    answer , t he actual angl e between

    t he

    two

    l i n e s ,

    i s

    communi cated

    t o

    S

    as

    t he

    answer

    t h a t

    t he

    const rai nt

    i s

    not

    s a t i s f i e d

    S

    now

    i ssues a command t o

    s a t i s f y

    t h i s constrai nt whi l e

    keepi ng

    Beam3 f i x ed, whi ch i s

    t ransl ated

    i n t o make the angl e

    between

    Beam3. p i c

    and

    Col umn3

    . p i c at J oi nt3 . p i c be 9

    degrees

    wi t hout

    modi f yi ngBeam3

    p i c

    f o r

    Carr yi ng

    out the

    operati on wi l l r es ul t i n t he

    shape

    shown i n

    Fi gure 4 c)

    Communi cat i on

    wi l l cont i nue i n

    t h i s

    manner

    unt i l

    al l t he

    const rai nt s

    a r e s a t i s f i e d

    Fi gure

    5

    shows REDRAWs

    symbol i c

    r easoni ng a c t i v i t y f or the

    same

    exampl e

    a

    I

    b

    d )

    e )

    Fi gure

    4

    RE R W

    ol uti on t o f rame

    s t r u c t u r e probl emsketched

    i n

    Fi gure 1

    24 5

  • 7/25/2019 Tessler 1993 Qualitative Structural Analysis Diagrammic Reasoning

    7/10

    Def l ect

    BEA.0

    i n t he

    same

    di rect i on as

    l oad

    i o t r r r 3 i s

    of support

    type f i xed at

    90

    .

    Const r ai nt

    A Angl e

    at

    coL3 must be

    90

    t o BEAMS

    Get angl e

    bet ween

    COL

    and

    BEAMSat J o t r r r 3

    Sat i s fy

    Constr ai nt

    A

    Make

    angl ebetweencoL3

    andBEAMS

    be 90 whi l e keepi ngBEAMS f i x ed

    ( S im l a r l y ,

    f or

    angl e

    between coL4 and

    BEAMS )

    suPPoRT3 i s

    of support

    type

    f i x ed at

    90

    .

    Get angl e

    of

    COL3 at

    suppoRT3

    S a t i s f y

    Constr ai nt B Make

    angl e

    o f c oL3

    t o

    suPPoRT3 be

    90

    ( S im l a r l y , f o r angl e of coL4to suppoRT4.

    Constr ai nt C

    Moment

    ar ound

    i our r 3must

    be

    zero

    Get

    moment

    ar ound i o m r 3

    Sat i s fy

    Constr ai nt

    C

    Establ i sh

    moment

    equi l i br i um

    ar ound i o i N - r 3

    S i m l a r l y

    f o r momnt ar ound j o i a

    C 1

    3 . 2

    Di scussi on

    Constr ai nt B Angl e at

    suppoRT3

    must be 90

    t o

    COL3

    6

    BendBEAMPic i n t he negati ve

    d i r e c t i o n

    of t he

    y-coordi nate .

    Get angl e

    bet ween

    coL3 . P i c andBEAM3 PIcat

    i o L NT 3

    . P I G

    Angl e

    bet ween

    coL3

    . P i c andBEAM3 Pl c i s