based control and engineering systems by: natalia danilova advisor: prof. tatiana · pdf...
TRANSCRIPT
By:By:
Natalia Danilova
Natalia Danilova
Advisor:
Advisor:
Prof. Tatiana A. Gavrilova
Prof. Tatiana A. Gavrilova
SPbSPU
SPbSPUJASS 2008
JASS 2008
KNOWLEDGE
KNOWLEDGE-- BASED CONTROL
BASED CONTROL
AND ENGINEERING SYSTEMS
AND ENGINEERING SYSTEMS
Intr
oduc
tion
Con
cept
sA
ppro
ache
sC
ase-
stud
ies
Per
spec
tives
Con
clus
ion
N. Danilova, Knowledge-Based Control and Engineering Systems
2
Co
nte
nts
Kno
wle
dge-
base
dsy
stem
sar
eth
em
ostm
atur
ean
dw
idel
y-us
edco
mm
erci
alfie
ldof
artif
icia
lint
ellig
ence
. The
sp
ectr
um o
f app
licat
ions
of t
hese
sys
tem
s to
indu
stria
l and
co
mm
erci
al p
robl
ems
is s
o w
ide
as to
def
y ea
sy
char
acte
rizat
ion.
The
app
licat
ions
find
thei
r w
ay in
to m
ost
area
s of
kno
wle
dge
wor
k.K
now
ledg
e-ba
sed
syst
ems
achi
eve
high
leve
lsof
perf
orm
ance
inta
skar
eas
that
, for
hum
anbe
ings
, req
uire
year
sof
spec
iale
duca
tion
and
trai
ning
.
N. Danilova, Knowledge-Based Control and Engineering Systems
3
Intr
od
uct
ion
Diagnostics
Design
Decision support
N. Danilova, Knowledge-Based Control and Engineering Systems
4
Kn
ow
led
ge-
Bas
ed
Sys
tem
s (M
ind
Map
)
Diagnostics
Design
Decision support
Kn
ow
led
ge
and
Dat
a
N. Danilova, Knowledge-Based Control and Engineering Systems
5
DATA
DATA
DATA
DATA
Inst
ance
s an
d fa
cts
char
acte
rizin
g ob
ject
’s, p
roce
sses
and
thei
r
prop
ertie
s
KNOWLEDGE
Rul
es,
conc
aten
atin
g th
e da
ta
and
obta
ined
from
exp
erie
nce
.
Def
init
ion
A k
now
ledg
ekn
owle
dge
base
dba
sed
syst
emsy
stem
, als
okn
own
asan
exp
ert
expe
rtsy
stem
syst
em, i
sa
com
pute
r pr
ogra
m th
atco
ntai
nsth
ekn
owle
dge
and
anal
ytic
alsk
ills
ofon
eor
mor
ehu
man
expe
rts,
rel
ated
toa
spec
ific
subj
ect.
A s
oftw
are
that
per
form
s a
task
that
wou
ld o
ther
wis
e be
per
form
ed b
y a
hum
an e
xper
t
N. Danilova, Knowledge-Based Control and Engineering Systems
6
His
tory
The
Log
ical
P
erio
d
1960
-197
6
Axi
omat
ic a
ppro
ach
of
solv
ing
prob
lem
s
Heu
ristic
app
roac
h
1965
: LIS
P
(firs
t sym
bol
proc
essi
ng la
ngua
ge)
1973
: PR
OLO
G
(axi
om-b
ased
la
ngua
ge)
Kno
wle
dge
Bas
ed
Per
iod
1976
-199
0
Con
clus
ion
is m
ade
from
vas
t am
ount
of
prev
ious
kno
wle
dge
colle
cted
DE
ND
RA
L
MY
CIN
Indu
stria
l A
I Sys
tem
s
1990
-now
aday
s
Exp
erts
yste
ms
are
appl
ied
toa
wid
eva
riety
ofar
eas
N. Danilova, Knowledge-Based Control and Engineering Systems
6
Th
e M
ain
Dif
fere
nce
Fro
m O
ther
Pro
gra
ms
•E
S m
odel
nott
heph
ysic
sof
the
dom
ain
butt
hew
ayof
prob
lem
solv
ing
byhu
man
expe
rtor
the
way
thin
king
and
know
ledg
epr
oces
sing
inth
isdo
mai
n.
•E
S m
ake
allc
oncl
usio
nsus
ing
Kno
wle
dge
that
ispu
tby
expe
rtan
dkn
owle
dge
engi
neer
.
•M
athe
mat
ics
isno
tess
entia
lin
ES
, heu
ristic
san
dfu
zzy
met
hods
are
mor
esu
bsta
ntia
l.
N. Danilova, Knowledge-Based Control and Engineering Systems
8
Exp
ert
Sys
tem
S
tru
ctu
re
N. Danilova, Knowledge-Based Control and Engineering Systems
9
Kn
ow
led
ge
eng
inee
rEn
d u
ser
INFERENCE
ENGINE
USER
INTERFACE
KNOWLEDGE BASE
E EEE X XXX P PPP E EEE R RRR T TTT S SSS Y YYY S SSS T TTT E EEE M MMM
Sol
utio
ns
Sol
utio
ns
Rul
es
Fac
ts a
bout
prob
lem
Fac
ts a
bout
prob
lem
Pra
ctic
al d
ata
(app
lied
know
ledg
e)H
ard
data
(sci
entif
ic k
now
ledg
e)
Pub
lishe
d da
taF
ield
stu
dies
Exp
ert
Mai
n B
lock
s
KNOWLEDGE BASE
cons
ists
ofse
nten
ces
whi
chde
fine
know
ledg
ew
ithth
eus
eof
supe
r-hi
gh-le
vell
angu
ages
whi
char
eca
lled
know
ledg
ere
pres
enta
tion
lang
uage
s. It
isth
eke
rnel
ofth
eex
pert
syst
em.
INFERENCE ENGINE
isa
prog
ram
whi
chsi
mul
ates
the
proc
ess
ofex
pertre
ason
ing
orde
cisi
onm
akin
g.
The
func
tion
ofth
eUSER INTERFACE
isto
pres
entq
uest
ions
and
info
rmat
ion
toth
eus
eran
dsu
pply
the
user
'sre
spon
ses
to
the
infe
renc
een
gine
. N. Danilova, Knowledge-Based Control and Engineering Systems
10
N. Danilova, Knowledge-Based Control and Engineering Systems
11
Cla
ssif
icat
ion
By relation
By relation
By relation
By relation
to real time
to real time
to real time
to real time
Sta
ticS
tatic
Qua
siQ
uasi
-- dyn
amic
dyna
mic
Dyn
amic
Dyn
amic
By problem
By problem
By problem
By problem
Dat
a D
ata
Inte
rpre
tatio
nIn
terp
reta
tion
DIAGNOSTICS
DIAGNOSTICS
DIAGNOSTICS
DIAGNOSTICS
CONTROL
CONTROL
CONTROL
CONTROL
For
ecas
tF
orec
ast
Pla
nnin
gP
lann
ing
Inst
ruct
ion
Inst
ruct
ion
DESIGN
DESIGN
DESIGN
DESIGN
DECISION SUPPORT
DECISION SUPPORT
DECISION SUPPORT
DECISION SUPPORT
By computer
By computer
By computer
By computer
type
type
type
type
On
supe
rcom
pute
rO
n su
perc
ompu
ter
On
mai
nfra
mes
On
mai
nfra
mes
On
sym
bolic
O
n sy
mbo
lic
proc
esso
rspr
oces
sors
On
wor
ksta
tions
On
wor
ksta
tions
On
pers
onal
O
n pe
rson
al
com
pute
rco
mpu
ter
By
By
By
By
integration
integration
integration
integration
degree
degree
degree
degree
Sta
ndS
tand
-- alo
neal
one
Hyb
ridH
ybrid
(( int
egra
ted
inte
grat
ed))
Exp
ert
Sys
tem
Lif
eC
ycle
N. Danilova, Knowledge-Based Control and Engineering Systems
12
Kno
wle
dge
Acq
uisi
tion
Kno
wle
dge
Rep
rese
ntat
ion
Pro
toty
pe
syst
em
Com
mer
cial
syst
em
Pro
blem
D
efin
ition
Ad
van
tag
es a
nd
Dis
adva
nta
ges
•P
rovi
des
cons
iste
ntan
swer
sfo
rre
petit
ive
deci
sion
s,
proc
esse
san
dta
sks
•H
olds
and
mai
ntai
nssi
gnifi
cant
leve
lsof
info
rmat
ion
•E
ncou
rage
sor
gani
zatio
nsto
clar
ifyth
elo
gic
ofth
eir
deci
sion
-mak
ing
•N
ever
"for
gets
" to
ask
a qu
estio
n, a
sa
hum
anm
ight
•La
cks
com
mon
sens
ene
eded
inso
me
deci
sion
mak
ing
•C
anno
tmak
ecr
eativ
ere
spon
ses
ashu
man
expe
rtw
ould
inun
usua
lci
rcum
stan
ces
•E
rror
sm
ayoc
cur
inth
ekn
owle
dge
base
, and
lead
tow
rong
deci
sion
s•
Can
nota
dapt
toch
angi
ngen
viro
nmen
ts, u
nles
skn
owle
dge
base
isch
ange
d
N. Danilova, Knowledge-Based Control and Engineering Systems
13
Fo
rwar
d c
hai
nin
g(D
ata
Dri
ven
rea
son
ing
)
The
syst
emke
eps
trac
kof
the
curr
ents
tate
ofpr
oble
mso
lutio
nan
dlo
oks
for
rule
sw
hich
will
mov
eth
atst
ate
clos
erto
a fin
also
lutio
n.
The
syst
emm
ustb
ein
itial
lypo
pula
ted
with
data
,in
cont
rast
toth
ego
aldr
iven
syst
emw
hich
gath
ers
data
asit
need
sit.
DA
TA
Rul
e 1
Rul
e 2
Rul
e N
Rul
e 3
Pos
sibi
lity
1
Pos
sibi
lity
2
Pos
sibi
lity
K
A=
1If
A=
1 &
B=
2 th
en C
=3;
If
C=
3 th
en D
=4
D=
4B
=2
N. Danilova, Knowledge-Based Control and Engineering Systems
14
KNOWLEDGE
KNOWLEDGE
KNOWLEDGE
KNOWLEDGE
PROBLEM
PROBLEM
PROBLEM
PROBLEM
Bac
kwar
d C
hai
nin
g(G
oal
-Dri
ven
Rea
son
ing
)
An
effic
ient
way
toso
lve
"str
uctu
red
sele
ctio
n" p
robl
ems.
Rul
e 1
Rul
e 2
Rul
e N
Rul
e 3
Sub
-pro
blem
3
Sub
-pro
blem
4
Sub
-pro
blem
M
Sub
-pro
blem
2
Sub
-pro
blem
1P
ossi
bilit
y 1
Pos
sibi
lity
2
Pos
sibi
lity
K
A=
1If
A=
1 &
B=
2 th
en C
=3;
If
C=
3 th
en D
=4
D=
4B
=2
N. Danilova, Knowledge-Based Control and Engineering Systems
15
Cas
e-S
tud
ies
Cas
e 1:
Bla
st F
urna
ce C
ontr
ol
•T
he c
ompa
ny: S
teel
Com
pany
's F
ukuy
ama
Wor
ks, J
apan
•
The
pro
blem
: Bec
ause
the
blas
t fur
nace
feed
s al
l oth
er
proc
esse
s in
the
stee
l mill
, any
inst
abili
ty in
th
e op
erat
ion
of th
e fu
rnac
e is
com
poun
ded
by th
e im
pact
on
othe
r pr
oces
ses
furt
her
dow
n th
e pr
oduc
tion
line.
•
The
pur
pose
: the
pre
dict
ion
of a
bnor
mal
con
ditio
ns w
ithin
the
blas
t fur
nace
(to
min
imiz
e th
e un
cert
aint
yin
the
oper
atin
g te
mpe
ratu
re)
N. Danilova, Knowledge-Based Control and Engineering Systems
16
Cas
e 1:
B
last
Fu
rnac
e C
on
tro
l•
Sub
-pro
blem
s:–
char
acte
rizin
g th
e cu
rren
t sta
te o
f the
furn
ace
and
proj
ectin
g th
e co
nditi
ons
occu
r se
vera
l hou
rs;
–tr
aini
ng a
ski
lled
blas
t fur
nace
ope
rato
r ta
kes
man
y ye
ars;
–th
e co
mpl
exity
of m
odel
ing
a bl
ast f
urna
ce;
–th
ere
are
no s
ymm
etrie
s to
sim
plify
the
geom
etric
m
odel
ing;
–th
e th
erm
al s
tate
of t
he fu
rnac
e ca
nnot
be
mea
sure
d di
rect
ly, b
ut m
ust b
e in
ferr
ed fr
om v
ario
us s
enso
r m
easu
rem
ents
;
N. Danilova, Knowledge-Based Control and Engineering Systems
17
The
sm
alle
r th
e un
cert
aint
y, th
e lo
wer
the
over
all t
empe
ratu
re
need
ed to
pro
duce
the
pig
iron,
res
ultin
g in
ver
y la
rge
fuel
sav
ings
.
Fue
lCos
tSav
ings
:
N. Danilova, Knowledge-Based Control and Engineering Systems
18
Cas
e 1:
B
last
Fu
rnac
e C
on
tro
l
The
feat
ures
of t
he e
xper
t sys
tem
:
•m
odel
s th
e cu
rren
t sta
te;
•pr
edic
ts fu
ture
tren
ds w
ith s
uffic
ient
accu
racy
to m
ake
cont
rol d
ecis
ions
;•
mak
es th
e co
ntro
l dec
isio
ns;
•de
cisi
ons
can
be im
plem
ente
d au
tom
atic
ally
or
man
ually
Cas
e 1:
Bla
st F
urn
ace
Co
ntr
ol
N. Danilova, Knowledge-Based Control and Engineering Systems
19
BF
sen
sors
N. Danilova, Knowledge-Based Control and Engineering Systems
20
Bla
stF
urna
ceE
xper
tSys
tem
:
Hot
sto
ve
Wat
er in
ject
ion
Ste
am Air
Dis
trib
uted
dig
ital
cont
rolle
rHot
sto
veco
ntro
ller
Gas
cle
anin
gco
ntro
ller
Bla
se
tem
pera
ture
Bla
se m
oist
ure
(Pro
cess
com
pute
r)(A
I pro
cess
or) In
fere
nce
engi
ne Kno
wle
dge
base
Con
clus
ion
Req
uest
Dat
a pr
oces
sing
for
reas
onin
g
Pro
cess
data
bas
e
Fur
nace
co
ntro
l
Sen
sor
data
Gat
herin
g
CR
TC
RT
IF…
TH
EN
…
Use
r t/
F
Mol
ten
iron,
sl
ag
Rac
eway
Coh
esiv
e zo
ne
Sys
tem
Com
pone
nts:
•a
proc
ess
com
pute
rga
ther
s in
put d
ata
from
var
ious
se
nsor
s in
the
furn
ace,
mai
ntai
ns a
pro
cess
dat
abas
e an
d ge
nera
tes
furn
ace
cont
rol i
nfor
mat
ion;
•th
e A
I pro
cess
orpr
ovid
es th
e kn
owle
dge
and
reas
onin
g fo
r as
sess
ing
and
inte
rpre
ting
the
sens
or d
ata,
hy
poth
esiz
ing
the
inte
rnal
sta
te o
f the
furn
ace,
and
de
term
inin
g ap
prop
riate
con
trol
act
ions
;
•a
dist
ribut
ed d
igita
l con
trol
ler
uses
the
furn
ace
cont
rol
data
from
the
proc
ess
com
pute
r to
con
trol
the
actu
al
blas
t fur
nace
.
N. Danilova, Knowledge-Based Control and Engineering Systems
21
Cas
e 1:
Bla
st F
urn
ace
Co
ntr
ol
Res
ult:
•C
ompa
ny a
nnua
l sav
ings
of $
6 m
illio
n•
Red
uctio
n in
sta
ff of
4 p
eopl
e
•Im
prov
emen
t in
the
qual
ity o
f the
furn
ace
outp
ut
Det
ails
:
•T
he s
yste
m is
impl
emen
ted
in L
ISP
with
FO
RT
RA
N u
sed
for
data
pr
epro
cess
ing
•T
he k
now
ledg
e in
the
AI p
roce
ssor
is c
onta
ined
in 4
00 r
ules
, 35
0 fr
ames
, and
200
LIS
P p
roce
dure
s
•F
uzzy
theo
ry is
em
ploy
ed in
its
infe
renc
e en
gine
•T
he s
yste
m h
as a
cyc
le ti
me
of 2
0 m
inut
es, c
ompa
red
to th
e fu
rnac
e tim
e co
nsta
nt o
f six
to e
ight
hou
rs.
Cas
e 1:
Bla
st F
urn
ace
Co
ntr
ol
N. Danilova, Knowledge-Based Control and Engineering Systems
22
Cas
e 2:
Mo
ther
bo
ard
Exp
ert
Sys
tem
•T
he p
robl
em: n
eed
to c
all t
o th
e se
rvic
e ce
nter
to a
sk fo
r th
e so
lutio
n fr
om th
e te
chni
cian
and
be
char
ged
for
that
; har
d un
ders
tand
ing
of th
ete
rms
writ
ten
in m
anua
l
•T
he p
urpo
se: t
o re
cogn
ize,
man
age
and
fix th
e m
othe
rboa
rd’s
pro
blem
, and
pro
vide
use
rsw
ith a
ppro
pria
te s
olut
ion
base
on
the
accu
rate
dia
gnos
is
N. Danilova, Knowledge-Based Control and Engineering Systems
23
The
feat
ures
of t
he e
xper
t sys
tem
:
•fig
ures
out
the
mai
n pr
oble
ms
of th
e M
SI’s
mot
herb
oard
;
•gi
ves
solu
tion
rega
rdin
g to
the
accu
rate
di
agno
sisCas
e 2:
Mo
ther
bo
ard
Exp
ert
Sys
tem
N. Danilova, Knowledge-Based Control and Engineering Systems
24
Met
hodo
logy
:•
Pha
se 1
: Pro
blem
Ass
essm
ent
–ga
ther
ing
all t
he in
form
atio
n–
iden
tifyi
ng th
e go
als
and
requ
irem
ents
•P
hase
2: K
now
ledg
e A
cqui
sitio
n an
d A
naly
sis
–co
llect
ion
of k
now
ledg
e (in
terv
iew
ing
the
MS
I mot
herb
oard
str
uctu
re e
xper
ts, c
olle
ctio
n da
ta fr
om th
e us
erm
anua
l boo
ks a
nd th
e da
ta fr
om th
e M
SI’s
web
site
).
•P
hase
3: D
esig
n an
d Im
plem
enta
tion
–ru
les,
sys
tem
pro
gram
min
g pa
rt a
nd s
yste
m in
terf
ace
Cas
e 2:
Mo
ther
bo
ard
Exp
ert
Sys
tem
N. Danilova, Knowledge-Based Control and Engineering Systems
25
•P
hase
4: T
estin
g–
all t
he k
now
ledg
e an
d ru
les
in th
e sy
stem
are
tota
lly a
ccur
ate
with
the
know
ledg
e th
at c
olle
cted
from
the
expe
rtis
e–
the
abili
ty o
f Mo
bE
Son
sol
ving
the
user
’s p
robl
ems
–us
ers
are
com
fort
able
with
the
desi
gn–
the
syst
em r
each
es th
e go
als
(pro
vide
s so
lutio
n an
d ad
vice
to
the
user
s so
that
they
can
han
dle
thei
r ow
n co
mpu
ter
prob
lem
s)
Mo
bE
Sac
tual
ly is
stil
l und
er d
evel
opm
ent
stat
us, a
nd it
stil
l nee
d tim
e an
d su
ppor
t fr
om th
e ex
pert
to r
each
it g
oals
.
Cas
e 2:
Mo
ther
bo
ard
Exp
ert
Sys
tem
N. Danilova, Knowledge-Based Control and Engineering Systems
26
•T
echn
olog
y w
ill c
lear
ly b
ecom
e m
ore
help
ful i
n de
alin
gw
ith in
form
atio
n ov
erlo
ad.
•T
he c
urre
nt c
apab
ility
of m
achi
ne in
telli
genc
e is
suc
h th
athu
man
kno
wle
dge
will
con
tinue
to b
e a
valu
able
res
ourc
e fo
rth
e fo
rese
eabl
e fu
ture
, and
tech
nolo
gy to
hel
p to
leve
rage
it w
ill b
e in
crea
sing
ly v
alua
ble
and
capa
ble.
•H
owev
er, i
n m
any
case
s ex
pert
s ar
e be
ing
aske
d to
surr
ende
r th
eir
know
ledg
e an
d ex
perie
nce—
the
very
trai
ts
that
mak
e th
em v
alua
ble
as in
divi
dual
s.
N. Danilova, Knowledge-Based Control and Engineering Systems
27
Per
spec
tive
s
Co
ncl
usi
on
Tha
t was
the
brie
f ove
rvie
w o
fkno
wle
dge-
base
d sy
stem
s fo
r co
ntro
l and
en
gine
erin
g. T
wo
diffe
rent
rea
l exp
ert
syst
ems
wer
e co
nsid
ered
–B
last
Fur
nace
C
ontr
ol s
yste
m a
nd M
othe
rboa
rd E
xper
t sy
stem
. Per
spec
tives
of k
now
ledg
e-ba
sed
syst
ems
deve
lopm
ent w
ere
conc
lude
d.
N. Danilova, Knowledge-Based Control and Engineering Systems
28
Ref
eren
ces
•A
mzi
inc.
Bui
ldin
gE
xper
tSys
tem
sin
Pro
log
: w
ww
.am
zi.c
om/E
xper
tSys
tem
sInP
rolo
g•
Cla
rke
R. K
now
ledg
e-B
ased
Exp
ertS
yste
ms
: w
ww
.anu
.edu
.au/
peop
le/R
oger
.Cla
rke/
SO
S/K
BT
.htm
l•
Day
J. K
now
ledg
e-ba
sed
Eng
inee
ring
–A
utom
atin
gfo
rP
rofit
abili
tyin
Pro
duct
Des
ign
: ww
w.d
esig
nnew
s.co
m•
Fei
genb
aum
E.,
Frie
dlan
dP
. Kno
wle
dge-
Bas
ed S
yste
ms
in J
apan
-w
ww
.wte
c.or
g/lo
yola
/kb/
c2_s
3.ht
m•
Gav
rilov
a T
. Cou
rse
of L
ectu
res
abou
t Kno
wle
dge
Eng
inee
ring
, -S
aint
-P
eter
sbur
g S
tate
Pol
ytec
hnic
alU
nive
rsity
, 200
7•
Nor
man
A. C
ours
e of
Lec
ture
s ab
out E
xper
t Sys
tem
s, -
Uni
vers
ity o
f Tex
as
at A
ustin
:ww
w.e
co.u
texa
s.ed
u/fa
culty
/Nor
man
/long
.ext
ra/In
fo.S
98/E
xp•
Sie
wW
ai.M
obE
SE
xper
tSys
tem
: w
ww
.gen
erat
ion5
.org
/con
tent
/200
5/M
obE
S.a
sp•
Tay
lor
J., S
trin
ger
P. A
nau
tosy
nthe
sizi
ngno
n-lin
ear
cont
rols
yste
mus
ing
a ru
le-b
ased
expe
rtsy
stem
:Int
erna
tiona
l Jou
rnal
of A
dapt
ive
Con
trol
and
Sig
nal
Pro
cess
ing,
5 M
ar20
07,V
olum
e 5,
Issu
e 1
, Pag
es21
–39
•W
ikip
edia
: ww
w.w
ikip
edia
.org
N. Danilova, Knowledge-Based Control and Engineering Systems
29
N. Danilova, Knowledge-Based Control and Engineering Systems
30
THANK YOU!
THANK YOU!
THANK YOU!
THANK YOU!