[PDF] Concept and practice of expert systems in civil engineering E Bernard

PDF document for free

1. PDF document for free

50598_3AIENG93029FU.pdf

Concept and practice of expert systems in

civi l engineerin g E . Bernard ! & P.R.C . Sardinh a Civi l Constructio n Divisio n Compute r Section , Sa o Paul o Stat e Institut e o f Technologica l

Research

, Brazi l

ABSTRAC

T

Firstly

, thi s pape r give s a n overvie w o f genera l concept s o f exper t systems : thei r parts , whe n exper t syste m buildin g i s justifie d an d appropriate , an d phase s an d tool s necessar y t o develo p them .

Secondly

, exper t system s ar e associate d wit h som e basi c characteristic s o f civi l engineerin g problems , suc h a s incompletenes s an d inaccurac y o f data , multidisciplinar y approache s an d th e nee d o f complementar y qualitativ e analysis . Example s o f exper t system s i n man y civi l engineerin g branche s suc h a s structural , geotechnica l an d environmenta l engineerin g ar e cited .

Finally

, a practica l experienc e o f buildin g a n exper t syste m demonstratio n prototyp e i n flexibl e pavemen t maintenanc e domai n i s reported .

INTRODUCTIO

N

Althoug

h i n th e beginnin g o f th e 60
* s som e researcher s wer e workin g t o reproduc e huma n reasonin g t o solv e generi c problems , onl y i n th e beginnin g o f th e 80
* s exper t systems , tha t ha d worldwid e repercussion s du e t o thei r abilit y t o solv e problem s i n determine d knowledg e areas , wer e developed . Som e o f thos e exper t system s wer e th e MYCI N fo r diagnosi s o f infectiou s diseases , th e PROSPECTO R fo r geologica l dat a interpretatio n an d th e MOLGE N fo r plannin g experiment s i n molecula r genetics

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

418 Artificial Intelligence in Engineering

Wit h suc h positiv e repercussion , exper t system s wer e considere d a grea t solutio n an d a lo t o f mone y wa s investe d i n them . But , i n a shor t perio d o f tim e peopl e notice d that , althoug h exper t system s ar e capabl e o f solvin g problem s tha t on e couldn' t solv e wit h traditiona l programmin g techniques , the y ha d som e limitations . Thei r developmen t wa s tim e consumin g an d the y didn' t solv e al l type s o f problems . Afte r tha t initia l euphoria , som e partiall y unsuccessfu l experience s le d t o a certai n discredit . But , i n th e followin g years , som e criteri a an d shell s fo r helpin g exper t system s buildin g wer e developed , an d a t th e en d o f th e 80'
s a mor e realisti c vie w o f the m emerged . Nowadays , ther e ar e new , wel l succeeded , experience s i n man y knowledg e domain s an d exper t system s begi n t o occup y thei r stric t role .

CONCEP

T I n thi s topi c basi c concept s o f a n exper t syste m an d it s developmen t ar e treated . Wha t exper t system s ar e

Artificia

l Intelligenc e i s th e par t o f compute r scienc e tha t studie s theme s relate d t o huma n intelligenc e simulatio n i n a computer . I t ha s branche s suc h a s natura l language , robotics , perceptio n (visio n an d speech) , neura l network s an d exper t systems .

Resultin

g fro m researche s abou t simulatio n an d reproductio n o f huma n reasonin g t o solv e problems , exper t system s ar e compute r program s that , quickl y an d efficiently , ca n solv e problem s a s a huma n exper t does . The y ar e interactiv e program s an d ca n proces s experience , judgemen t an d intuition .

Differentl

y fro m ordinar y programs , exper t system s represen t knowledg e symbolicall y usin g character s involvin g complexity , uncertaint y an d ambiguousnes s (an d no t onl y certaint y an d determinis m s o commo n whe n workin g wit h algorithms) . T o hol d thi s knowledge , exper t system s appl y man y strategie s an d us e heuristi c intensivel y (tha t is , th e us e o f procedure s that , althoug h the y can' t b e proved , lea d t o th e solutio n o f th e proble m quickly) . Part s o f a n exper t syste m A n exper t syste m ha s fiv e mai n parts : th e knowledg e acquisitio n module , knowledg e base , inferenc e engine , explanatio n mechanis m an d use r interface

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Artificial Intelligence in Engineering 419

Th e knowledg e acquisitio n modul e i s th e syste m par t tha t interface s wit h th e knowledg e enginee r o r wit h th e huma n expert , offerin g resource s t o ad d an d modif y th e knowledg e represente d i n th e knowledg e base . I n th e knowledg e bas e th e procedures , strategie s an d reasonin g tha t a huma n exper t use s t o solv e a certai n proble m ar e represented . Thi s knowledg e i s symbolicall y represente d an d organize d usin g fact s (data ) an d rule s (o r othe r representation ) tha t ca n b e associate d wit h certaint y factors . Th e inferenc e engin e i s a structur e tha t ha s a n algorith m capabl e o f usin g th e knowledg e bas e effectively . I t ha s a n interprete r tha t decide s ho w t o appl y th e rule s t o infe r ne w knowledge , an d a schedule r tha t decide s th e orde r i n whic h th e rule s shoul d b e applied . Ther e ar e tw o way s o f chainin g th e rule s tha t ar e bein g analyzed : th e forward-chainin g an d th e backward - chaining . Th e explanatio n mechanis m store s informatio n abou t ho w exper t system s reac h thei r conclusions , tha t is , th e decision s tha t ar e bein g mad e b y th e syste m durin g th e solutio n proces s o f a certai n problem . Th e use r interfac e i s th e componen t throug h whic h th e use r ca n communicat e wit h th e system . I t supplie s dat a an d informatio n o r obtain s results , conclusion s an d explanation s abou t ho w an d wh y th e syste m reache d a certai n solution . Whe n t o buil d a n exper t syste m I t i s ver y difficul t t o describ e i n genera l term s th e characteristic s tha t mak e a proble m appropriat e fo r a n exper t syste m development . Waterma n [1 1 talk s abou t som e aspect s tha t ca n hel p on e t o decide : - Exper t syste m developmen t i s possibl e whe n i t ca n b e validated , thi s meanin g tha t ther e ar e huma n expert s studyin g th e problem , an d the y agre e abou t thei r solvin g methods , th e choic e an d precisenes s o f th e solution . • " A n exper t syste m i s justifie d whe n huma n expert s ar e expensiv e o r scarc e o r ther e i s a grea t deman d fo r them . O r whe n th e huma n exper t decisio n makin g mus t tak e plac e i n a hostil e environment . O r th e knowledg e i s gettin g lost , fo r example , becaus e o f hig h rotativit y o f personne l i n a company . -I t i s appropriat e whe n i t involve s heuristi c knowledg e an d th e proble m ca n b e solve d naturall y b y manipulatin g symbol s an d thei r structures . Whe n th e proble m ca n b e solve d wit h mathematica l model s on

e Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

420 Artificial Intelligence in Engineering

mus t us e algorithmi c methods . Moreover , th e proble m mus t b e comple x o r difficul t enoug h t o justif y th e cos t an d effor t o f a n exper t syste m development . Phase s o f exper t syste m developmen t Th e developmen t o f a n exper t syste m involve s th e followin g phases : knowledg e acquisition , knowledg e formalization , implementatio n an d tests .

Knowledg

e acquisitio n involve s direc t an d interactiv e contac t betwee n th e huma n exper t an d th e knowledg e engineer . Th e knowledg e enginee r take s th e exper t knowledge , conceptualize s an d formalize s it . I n thi s phas e on e ca n b e face d wit h th e mai n difficult y i n exper t system s building : exper t knowledg e acquisitio n an d formalization . Face d wit h realisti c problem s t o b e solved , th e huma n exper t ha s a tendenc y t o sho w hi s reasonin g an d conclusion s i n a generi c way , ver y fa r fro m tha t necessar y fo r a compute r analysis . H e combine s piece s o f hi s basi c knowledg e s o quickly , tha t i t i s difficul t fo r hi m t o describ e thi s proces s i n detail .

Knowledg

e formalizatio n involve s expressin g concept s an d relation s i n a forma l way . Ther e ar e man y technique s tha t us e mechanism s associate d t o th e characteristic s o f huma n intelligenc e t o represen t th e knowledge . The y cop y th e wa y human s represen t thei r knowledge . Thes e technique s ar e calle d forma l wa y o r knowledg e representatio n method s an d th e mos t use d are : rules , semanti c net s an d frames . Rule s ar e base d o n structure s lik e I F (premise ) THE N (conclusion ) o r I F (condition ) THE N (action ) an d the y ar e a natura l wa y t o describ e dynami c processes ; the y wor k ver y wel l fo r problem s drive n b y th e data , wher e deviation s ar e common . Semanti c net s an d frame s provid e a natura l wa y t o structur e a taxonomy , tha t is , problem s tha t involv e relations/hierarchy . Th e implementatio n phas e turn s th e formalize d knowledg e int o a compute r program . I n thi s phas e on e mus t us e a n exper t syste m too l fo r development , an d it s choic e wil l depen d o n th e typ e o f problem , knowledg e formalizatio n an d too l features .

Finally

, th e tes t phas e involve s th e evaluation , revisio n an d validatio n o f th e exper t syste m facin g rea l problems . Exper t syste m tool s fo r developmen t Tool s fo r buildin g exper t system s ar e basicall y softwar e resource s tha t ca n b e divide d int o thre e categories : programmin g languages , system-buildin g aid s an d knowledg e engineerin g languages

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Artificial Intelligence in Engineering 421

Programmin

g language s use d t o develo p exper t system s ar e generall y th e language s tha t hav e flexibilit y fo r th e knowledg e bas e implementation , inferenc e engin e construction , an d whic h allo w fo r th e developmen t o f man y interfac e resources . Som e example s ar e problem-oriente d language s suc h a s

FORTRA

N an d PASCAL , symbol-manipulatio n language s suc h a s LIS P an d

PROLOG

, an d object-oriente d language s a s SMALLTALK .

System-buildin

g aid s ar e projecte d mainl y t o hel p knowledg e acquisitio n an d design . Throug h th e definitio n o f th e problem , whic h involve s al l th e possibl e decisions , it s attribute s an d values , th e syste m wil l as k th e use r fo r example s describin g th e condition s tha t le d hi m t o thi s decision . Som e o f thes e system s ar e th e EXPERT-EASE , RULE-MASTE R an d SEE K (e.g . Waterma n [11) .

Knowledg

e engineerin g language s ar e a complet e se t o f resource s use d t o buil d exper t systems , combinin g languag e powe r wit h a sophisticate d interfac e an d suppor t environment . The y ar e commonl y calle d shells . Mos t o f the m ar e base d o n rule s an d hav e thei r ow n inferenc e engine . The y ar e usefu l fo r inexperience d developmen t groups , becaus e the y offe r facilitie s fo r editing , debugging , executin g an d interfacin g (suc h a s explanatio n facilities) . Som e o f thes e shell s ar e general-purpos e suc h a s OPS5 , EMYCIN , M.I , KEE , ROSIE , VP-EXPERT , FUZZY-EXPERT , GURU , NEXPERT , TEARS , G 2 an d PATE R (th e las t on e wa s develope d i n Brazil ) (e.g . IP T [21) . Som e other s ar e mor e oriente d fo r diagnosti c problem s suc h a s

PERSONA

L CONSULTAN T an d SeRIS .

Suppor

t service s fo r usin g thes e shell s are , i n general , no t availabl e i n

Brazil

, an d thi s fac t ha s brough t man y difficultie s durin g exper t syste m development s an d sometime s ha s postpone d them .

ENGINEERIN

G APPLICATION S

Althoug

h engineerin g i s a n exac t science , i t i s als o a knowledg e are a tha t involve s a lo t o f intuitio n an d experience . Thi s i s naturall y clea r whe n w e thin k abou t th e numbe r o f expert s an d consultant s necessar y t o mak e diagnose s an d reac h solution s fo r th e majorit y o f problem s w e encounter . O n th e othe r hand , engineerin g usuall y manipulate s a grea t volum e o f incomplet e and/o r inaccurat e dat a t o se e th e feasibl e alternative s fo r th e problem . I n thes e case s qualitativ e analysi s i s frequentl y mor e importan t tha n th e quantitativ e ones .

Additionally

, engineerin g nowaday s use s mor e an d mor e multidisciplinar y knowledg e t o trea t problems , suc h a s makin g a

n Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

422 Artificial Intelligence in Engineering

environmen t impac t analyse s o f a grea t civi l work , o r plannin g th e traffi c contro l o f a bi g city , o r designin g an d analyzin g a medica l prosthesi s performanc e (e.g . Bernard i f31) . Thes e characteristic s ar e extremel y favorabl e fo r usin g exper t systems . Ver y successfu l i n medicin e an d financia l areas , exper t system s too k a lon g tim e t o b e disseminate d i n engineering , bu t toda y ther e ar e man y prototype s tha t ar e leavin g universitie s an d researc h center s t o becom e product s use d b y companies . Civi l engineerin g application s Al l characteristic s describe d abov e ca n b e foun d i n civi l engineering . I n environmenta l civi l engineering , on e o f th e mos t importan t characteristic s i s th e involvemen t o f multidisciplinar y area s (chemistry , biology , flui d mechanics , mathematics , statistics , economics , law , etc.) . Exper t system s i n thi s are a usuall y dea l wit h hazardou s wast e management , wate r qualit y protectio n an d wate r resources . Som e example s (e.g . Mahe r [4] ) ar e th e FRE S fo r decidin g emergenc y response s i n chemica l spills ; th e RP I an d GEOTO X fo r aidin g th e evaluatio n o f hazardou s wast e sites ; th e QUAL2 E Adviso r fo r suggestin g appropriat e inpu t value s fo r QUAL2 E tha t predict s wate r qualit y impact s o f pollutan t discharge s o n river s an d streams ; an d th e Floo d Adviso r fo r modelin g estimat e desig n flood s fo r civi l engineerin g projects . I n structura l engineerin g exper t system s hav e aide d design , analysis , inspectio n an d maintenance , an d cod e checking . Ther e i s als o a grea t potentia l applicatio n fo r diagnosi s an d interpretatio n problems . Mos t o f th e exper t system s develope d i n thi s are a involv e interactio n wit h conventiona l calculu s an d analysi s programs , an d th e exper t syste m i s a n ai d t o conceiv e model s an d t o understan d structura l behavior . Som e example s (e.g . Mahe r f4] ) are : HI-RIS E fo r preliminar y desig n o f high-ris e buildings ; BDE S fo r helpin g th e analysis , modelin g an d decisio n abou t th e bette r bridg e superstructur e design ; an d VibDia g fo r vibratio n diagnosis . Othe r example s ar e INDE X fo r industria l buildin g desig n (e.g . Adel i [51) ; RAISE- 1 fo r diagnosi s o f safet y an d integrit y o f structura l system s (e.g . Adel i [5]) ; an d a n Exper t Syste m fo r th e diagnosi s o f buildin g defect s (develope d i n Singapore , e.g . Mathu r [6]) .

Geotechnica

l engineerin g i s a n are a marke d b y th e manipulatio n o f incomplet e and/o r inaccurat e data , tha t depen d o n th e sensitivit y an d skil l o f a n exper t wh o i s abl e t o extrapolat e an d infe r goo d results . Tha t i s wh y thi s are a i s ver y goo d fo r buildin g an d usin g exper t systems . Ther e ar e als o man y application s involvin g th e interpretatio n o f parameter s o f a heuristi c natur

e Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Artificial Intelligence in Engineering 423

i n general , an d other s tha t interac t wit h conventiona l programs . Som e example s ar e CESSO L fo r analysi s o f geotechnica l investigatio n problem s (e.g . Bigo t [7] ) an d EXSE L fo r diagnosi s an d treatmen t o f da m seepag e problem s (e.g . Adel i [51) . Othe r example s (e.g . Mahe r [4] ) ar e CON E fo r soi l classificatio n an d inferenc e o f shea r stresse s usin g con e penetratio n tes t dat a an d RETWAL L fo r th e choic e o f alternativ e design s fo r retainin g walls .

Transportatio

n engineerin g involve s planning , design , contro l an d operation , managemen t an d maintenanc e activities ; an d i n al l o f the m computer s ar e essentia l tools . Though , man y problem s can' t b e solve d b y th e usag e o f numerica l algorithm s becaus e the y involv e skil l an d judgemen t o f huma n expert s an d the y encompas s social , politica l an d huma n behavio r elements . Exper t system s hav e bee n a ver y usefu l alternativ e i n thes e cases . A n exampl e o f thi s i s plannin g an d desig n o f traffi c networks . Som e example s o f exper t system s (e.g . Mahe r [41 ) ar e PARADIG M tha t integrate s som e exper t system s t o presen t strategie s fo r regiona l pavemen t maintenance ; CHIN A fo r th e desig n o f highwa y nois e barriers ; an d HERCULE S t o generat e traffi c contro l plan s t o avoi d congestin g situations . Othe r example s ar e Pavemen t Exper t fo r evaluatin g concret e pavemen t performanc e (e.g .

Toppin

g [81 ) an d Managemen t o f Lo w Volum e Flexibl e Pavemen t tha t recommend s pavemen t rehabilitatio n an d maintenanc e strategie s (e.g . Aouga b [91) .

Application

s fo r plannin g an d managemen t o f civi l constructio n work s ca n als o b e foun d i n variou s reference s (e.g . Mahe r [41) .

PRACTIC

E A practica l experienc e i n buildin g a demonstratio n prototyp e o f a n exper t syste m i n engineering , performe d a t th e Civi l Constructio n Divisio n o f IPT , wil l b e reported . Th e choic e o f a domai n Ther e ar e man y applicatio n area s (domains ) fo r exper t system s developmen t i n engineering . Fo r thi s firs t experienc e pavemen t maintenanc e wa s chosen . Thi s domai n ha s tw o basi c characteristics : i t ha s a grea t shor t ter m potentia l interes t fo r th e communit y an d th e Civi l Constructio n Divisio n o f IP T ha s a lo t o f accumulate d experienc e i n thi s domai n (e.g . IP T [21) . Th e dimensio n o f th e roa d networ k o f th e bigges t state s i n Brazil , it s importanc e a s a mean s o f communicatio n an d transportatio n o f good s amon g regions , an d pavemen t age , sinc e man y road s wer e buil t man y year s ago , sho w th e increasin g nee d o f a constant , quic k an d efficien t maintenance

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

424 Artificial Intelligence in Engineering

I n man y countrie s suc h a s Chil e (e.g . C . [101) , th e U.S.A . (e.g . Aouga b [9]) , Canad a an d Englan d (e.g . Toppin g [81 ) exper t syste m developmen t i s emergin g t o hel p engineer s wh o wor k i n publi c office s responsibl e fo r roa d an d stree t maintenance . Exper t systems , gatherin g experienc e fro m pavemen t expert s wh o hav e develope d successfu l maintenanc e strategies , ca n hel p les s experience d engineer s t o adop t adequat e maintenanc e an d recuperatio n system s eve n wit h limite d o r incomplet e data . I t wa s mad e a n evaluatio n tha t a n exper t syste m fo r pavemen t maintenanc e shoul d b e appropriat e an d usefu l i n Brazil . I t shoul d als o b e o f extrem e interes t an d usefulnes s i n man y cit y office s an d publi c organization s directl y connecte d wit h th e proble m a s th e Roa d Engineerin g Departmen t o f th e Sa o Paul o State . Therefore , th e Compute r Sectio n an d Pavemen t Grou p o f Civi l Constructio n Divisio n o f IP T gathere d thei r effort s t o buil d a demonstratio n prototyp e o f a n exper t syste m fo r urba n flexibl e pavemen t maintenance .

Knowledg

e acquisitio n Th e knowledg e acquisitio n phas e wa s don e throug h bibliographi c technica l materia l (suc h a s technica l bulletins , reports , photos ) an d periodi c meeting s wit h tw o pavemen t experts . Bot h activitie s wer e conducte d b y tw o engineerin g researcher s tha t ac t a s knowledg e engineers . The y ar e graduate d i n civi l engineerin g an d computers .

Initia

l wor k wa s don e t o defin e th e limit s o f th e chose n domai n mor e precisely . Initiall y a broa d attac k o f th e proble m wa s discussed . Throug h th e us e o f diagram s an d flowchart s mor e stric t limit s wer e redefined . Th e initia l approac h ha d involve d proble m identification , analysis , test s an d suggestion s o f preventiv e an d correctiv e attitudes , an d late r th e domai n wa s restricte d t o th e identificatio n o f th e proble m an d it s causes . Th e firs t phas e o f knowledg e acquisitio n wa s characterize d b y terminological , conceptua l an d mainl y commo n sens e problems . Thes e problem s wer e surpasse d b y th e adoptio n o f concept s an d commo n definitions , resulte d fro m th e discussio n betwee n th e expert s an d thei r revie w o f th e informatio n obtaine d i n th e bibliographi c material . Th e conclusio n tha t som e defect s calle d "concentrate d settlement " an d "depression " coul d b e treate d a s a uniqu e defec t i s a n example . Anothe r exampl e i s th e concept s o f norma l an d excessiv e traffic , th e firs t on e define

d Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Artificial Intelligence in Engineering 425

a s traffi c fo r whic h th e roa d wa s projecte d an d th e las t a s traffi c abov e th e projected .

Acquisition

, tha t wa s don e initiall y withou t a methodolog y (onl y wit h a gatherin g o f knowledg e i n th e area) , wa s don e i n a mor e organize d for m afte r th e star t o f th e knowledg e formalization . Usin g a pavemen t defec t enumeration , knowledg e engineer s an d expert s simulat e man y situation s an d phenomen a relate d t o th e defect , obtainin g fo r eac h simulatio n th e probabl e caus e o f th e defect . Fo r example , th e knowledg e enginee r aske d th e exper t t o imagin e a pavemen t wit h "bloc k cracking" . Then , th e enginee r adde d mor e information , suc h a s norma l traffic , pavemen t i s thre e year s ol d an d s o on , unti l th e exper t reache d th e defec t cause s o r conclude d tha t th e informatio n ha d n o consistency . Th e proces s wen t o n unti l th e complet e acquisitio n o f th e proble m knowledg e wa s reached .

Knowledg

e formalizatio n Th e knowledg e formalizatio n phas e starte d immediatel y afte r th e limi t definitio n o f th e domai n i n th e knowledg e acquisitio n phase . Sinc e th e beginnin g i t wa s natura l an d immediat e tha t th e representatio n woul d b e don e usin g rule s (a s define d i n precedin g topics) .

Actually

, thes e tw o phase s wer e develope d together , an d th e methodolog y use d fo r knowledg e formalizatio n generate d a paralle l methodolog y fo r knowledg e acquisition . Initiall y a defec t wa s chose n an d a grou p o f situation s (attributes ) mor e commo n an d immediat e wer e added . Usin g th e assemblag e defec t plu s situation s th e caus e (conclusion ) wa s reached . Then , othe r value s fo r th e situation s (maintainin g th e sam e defect ) wer e adde d an d ne w cause s fo r thes e ne w combination s wer e obtained . A n exampl e o f formalizatio n is : BLOC K CRACKIN G + NORMA L TRAFFI C +

PAVEMEN

T AG E LES S THA N 5 YEAR S + N O DEFORMATIO N = > BA D QUALIT Y O F WEARIN G COURS E (VER Y STIFF ) Too l choic e

Considerin

g th e inexperienc e o f th e grou p i n exper t syste m developmen t an d th e lac k o f financia l resources , i t wa s decide d t o acquir e a nationa l an d simpl e too l calle d PATE R 2. 0 (e.g . TECSI S [111)

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

426 Artificial Intelligence in Engineering

PATE R 2. 0 i s a shel l fo r exper t syste m developmen t base d o n rule s typ e "IF.. . THE N .. . ELSE..." , easil y use d du e t o it s friendl y use r interface , i n spit e o f it s limite d resources . I t allow s fo r creation , executio n an d debuggin g exper t system s i n a completel y integrate d environment . I t work s wit h concept s suc h a s attributes , value s an d variable s an d ha s a n inferenc e engin e tha t use s backward-chaining . I t wa s a goo d too l fo r buildin g a demonstratio n prototyp e wit h th e characteristic s describe d before .

Knowledg

e bas e implementatio n Thi s phas e starte d wit h th e definitio n o f th e attributes , thei r value s an d variables . The n th e rule s wer e constructed . Th e situations/condition s a s bein g attribute s an d variable s wer e define d wit h th e knowledg e formalization . Th e attribute s ar e situations/condition s tha t hav e thei r value s well-define d b y intervals . Th e variable s ar e als o situations/condition s bu t thei r value s mus t b e treate d i n mor e flexibl e interval s i n th e rules . A n exampl e o f attribut e i s "TRAFFIC " tha t ca n assum e th e value s "NORMAL " an d "EXCESSIVE" . A n exampl e o f variabl e i s "PAVEMEN T AGE " that , dependin g o n th e rule , ca n hav e differen t interval s o f value , tha t is , "PAVEMEN T AG E < 5 YEARS " an d " 2 YEAR S < PAVEMEN T AG E < 1 0 YEARS" . Du e t o th e proble m domai n an d th e methodolog y adopte d i n knowledg e acquisitio n an d formalization , th e firs t ste p i n knowledg e bas e constructio n wa s th e creatio n o f th e attribut e "DEFECT " wit h it s value s (name s o f th e defects ) wit h it s descriptions . Then , ne w attributes , value s an d variable s wer e incorporate d i n th e bas e fo r th e constructio n o f th e rules . A n exampl e o f a n implemente d rul e is : I F Defec t = Block_Crackin g an d

Traffi

c = Norma l an d

Pavement_Ag

e < 5 an d

Deformatio

n = N o THE N

Proble

m = Ba d qualit y o f wearin g cours e (ver y stiff

) Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Artificial Intelligence in Engineering 427

Statu s repor t Th e knowledg e acquisitio n an d formalizatio n phase s ar e concluded . Wit h th e conclusio n o f th e implementatio n phas e i t wa s possibl e t o consul t th e base , creatin g a feed-bac k proces s tha t le d t o a refinemen t o f th e acquisitio n an d formalizatio n alread y done . I t i s interestin g t o observ e tha t wit h thi s feed - bac k proces s al l phase s wer e execute d a t th e sam e time . Now , wit h th e en d o f knowledg e bas e constructio n som e test s wit h othe r expert s an d user s shoul d b e carrie d ou t t o validat e and , eventually , refin e th e knowledg e base . Afte r thi s a n extensio n o f th e domai n o f thi s demonstratio n prototyp e wil l b e discussed . I n thi s extensio n th e goa l syste m wil l b e abl e t o giv e suggestion s abou t preventiv e an d correctiv e attitude s t o b e adopte d fo r th e analyze d pavement .

CONCLUSION

S

Nowaday

s exper t system s ar e a consolidate d techniqu e an d man y engineerin g problem s hav e characteristic s tha t ca n b e treate d b y thi s technique . Th e tendenc y i s t o increas e th e numbe r o f product s availabl e i n th e market . I n th e cas e o f th e demonstratio n prototyp e describe d i n thi s paper , experienc e ha s show n on e mor e tim e tha t th e "bottleneck " i n a n exper t syste m buildin g reside s i n th e knowledg e acquisition . But , th e result s wer e hopeful . I t i s intende d t o buil d a fina l produc t probabl y usin g mor e advance d tool s fo r th e developmen t an d includin g som e graphi c resource s fo r use r interface s i n th e nea r future .

ACKNOWLEDGEMENT

S Th e author s than k IPT , especiall y th e colleague s o f th e Pavemen t Group , an d FAPES P fo r th e opportunit y o f divulgin g thi s paper .

REFERENCE

S 1 . Waterman , D.A . A Guid e t o Exper t Systems . Addison-Wesley , U.S.A. , 1987
. 2 . IP T - Institut e d e Pesquisa s TecnoWgica s d o Estad o d e Sa o Paulo .

Sistema

s Especialista s e m Engenhari a Civil . Relatdri o n ° 28.932
. Sa o Paulo , 1991

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

428 Artificial Intelligence in Engineering

3 . Bernard! , E. , Sardinha , P.R.C . 'Sistema s Especialistas : Conceit o e

Pnitica'

, i n XII I CILAMCE , Vol . 2 , pp . 421-431
, Proceeding s o f th e XII I

Congress

o Iber o Latin o American o sobr e M6todo s Computadonai s par a

Engenharia

, Port o Alegre , Brazil , 1992
. 4 . Maher , M.L . (Ed.) . EXPER T SYSTEM S fo r Civi l Engineers : Technolog y an d Application . ASCE , Ne w York , 1987
. 5 . Adeli , H . (Ed.) . Microcompute r Knowledge-Base d Exper t System s i n Civi l

Engineering

. ASCE , Ne w York , 1988
. 6 . Mathur , K.S. , Leng , A.A . 'Efficac y o f exper t syste m technolog y fo r th e diagnosi s o f buildin g defect s - a cas e study' . Buildin g Researc h an d

Information

, Vol . 20 , n . 5 , pp . 307-312
, London , 1992
. 7 . Bigot , G. , Marchal , J. , Moreau , M.C . 'CESSO L - System e exper t pou r le s reconnaissance s d e so l dan s l e domain e d u batiment' . Bulleti n d e Liaiso n de s Laboratoire s de s Font s e t Chauss&s , n . 154
, pp . 110-112
, Paris , 1988
. 8 . Topping , B.H.V . (Ed.) . TH E APPLICATIO N O F ARTIFICIA L

INTELLIGENC

E TECHNIQUE S T O CIVI L AN D STRUCTURA L

ENGINEERING

, Pro . A I CIVIL-COMP , Civil-Comp , Edinburgh , 1987
. 9 . Aougab , H. , Schwartz , C.W. , Wentworth , J . 'Exper t Syste m fo r

Managemen

t o f Lo w Volum e Flexibl e Pavements' , pp . 759-768
, Proceeding s o f th e Fift h Conferenc e o n Computin g I n Civi l Engineering : Microcomputer s t o Supercomputer s (Ed . Will , K.M.) , Alexandria , Va. , 1988
. ASCE , Ne w York , 1988
. 10 . C. , Carlo s Videla , G. , G . Echeverria , L. , Marcel o Pui g e t al . 'GIMP : U N SISTEM A D E INFORMACIO N COMPUTADORIZAD O PAR A L A

GESTIO

N INTEGRA L D E PAVIMENTO S ASFALTICOS' , i n X I

CILAMCE

, pp . 623-642
, Proceeding s o f th e X T Congress o Iber o Latin o

American

o sobr e Mgtodo s Computacionai s par a Engenharia , Ri o d e Janeiro ,

Brazil

, 1990
. 11 . TECSIS . Manua l d o UsuAri o d o PATER . Ri o d e Janeiro , Brazil , 1990

. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517

Systems Engineering Documents PDF, PPT , Doc

[PDF] 2019 systems engineering exam

1. Engineering Technology

2. Industrial Engineering

3. Systems Engineering

[PDF] adamson systems engineering canada

[PDF] adamson systems engineering careers

[PDF] aero systems engineering zoominfo

[PDF] afp systems engineering office

[PDF] agile systems engineering pdf

[PDF] best systems engineering schools

[PDF] biological systems engineering careers

[PDF] civil engineering systems examples

[PDF] cognitive systems engineering what is

12345 Next 200000 acticles