2000r
849r
i
processfaultdiag
osisJAutomatica1993294843r
监控的整体框架这一方法不需建立系统的数学模r
a提出了一种基于ICASVM对系统性能进行r
pe
de
tcompo
e
ta
alysisJIEEETra
sactio
so
Neuralr
450Networks1999362663410r
中南工业大学学报然科学版自r
第34卷r
4LIRFWa
gXZDime
sio
reductio
ofprocessdy
amica
dChemicalE
gi
eeri
g200274732646tre
dsusi
gi
depe
de
tcompo
e
ta
alysisJComputersr
atestatisticalprocessmo
itori
gmethodswithapplicatio
stoE
gi
eeri
g200216117426r
theEastma
challe
geproblemJComputersa
dChemicalr
5Dow
sJJVogelEFApla
twidei
dustrialprocessco
trol17525524r
7Ka
oMHasebeSHashimotoIA
ewmultivariatestatisticalprocessmo
itori
gmethodusi
gpri
cipalcompo
e
ta
alysisJComputersa
dChemicalE
gi
eeri
g200125r
problemJComputersa
dChemicalE
gi
eeri
g1993r
6Ka
oMNagaoKHasebeSetalCompariso
ofmultivarir
11031113r
Themethodofi
depe
de
tcompo
e
ta
alysisa
dsupportvectormachi
ebasedfaultdiag
osisr
GUOMi
gXIELeiHENi
gWANGShuqi
gr
tio
alKeyLaboratoryofI
dustrialCo
trolTech
ologyNar
I
stituteofAdva
cedProcessCo
trolZhejia
gU
iversityHa
gzhou310027hi
aCr
AbstractAchemicalprocesshasalarge
umberofmeasuredvariablesbutitisusuallydrive
byfeweresse
tialvariableswhichmay
otbemeasuredExtracti
gtheseesse
tialvariablesa
dmo
itori
gthemwillimprovetheprocessmo
itori
gperforma
ceI
thispapera
i
tegratedframeworkforprocessmor
itori
ga
dfaultdiag
osisisprese
tedwhichcombi
esi
depe
de
tcompo
e
ta
alysisAforfeaICtodetermi
etheprojectio
coefficie
tmatrixwhichreprese
tsthefeaturescharacterizi
gthecurre
toperr
tureextractio
a
dsupportvectormachi
eMforide
tificatio
ofdiffere
tfaultsourcesICAisusedSVr
atio
co
ditio
Multiplesupportvectormachi
esaretrai
eda
dusethecoefficie
tmatrixastheiri
putsEastma
processr
toide
tifythefaultsThemethodisprovedtobeeffectivebytheapplicatio
tomo
itori
gTe
esseeKeywordssupportvectormachi
eprocessmo
itori
gfaultdiag
osisr
r
1r
白日放歌须纵酒，青春作伴好还乡。江山代有才人出，各领风r