1
Full i le: Moni o ing o uel oil p ocess o Ma ine Diesel Engine
Au ho s: Da id Boullosa-Falces a,*, Juan Luis La abe Ba ena a, Albe o Lopez-
A aiza a, Jaime Menendezb, Miguel Angel Gomez Solae xe a.
a Depa men o Nau ical Sciences and Ma ine Sys ems Enginee ing, Uni e si y o he
Basque Coun y UPV/EHU, Ma ia Diaz de Ha o 68, 48920 Po ugale e, Spain.
b Ibaizabal Tanke s Shipping Company.
*Co esponding au ho . Tel.: +34 946014850. E-mail add ess: da [email protected]
(Da id Boullosa-Falces).
Highligh s
- Selec ion uel oil p ocess a iables o a Ma ine Diesel engine.
-P ocess moni o ing h ough Small Sudden De ia ion Me hod (SSDM).
- De ec ion o small and sudden de ia ion o he p ocess.
- Iden i ica ion o he a iables ha caused de ia ions in he p ocess.
Abs ac
Ho elling´s T2 con ol cha is e y e icien o de ec ing sudden changes in a p ocess;
howe e , i loses sensi i i y o de ec small and p og essi e changes and i s pe o mance
dec eases when he numbe o a iables moni o ed a he same ime is high. Because o
his, con en ional me hods o a iable educ ion such as PCA we e used, bu hey ha e
di icul ies in de ec ing he a iabili y o he p ocess when he co ela ion be ween
a iables is poo .
We p opose a Me hod o de ec ion o Small and Sudden De ia ions in he p ocess
(SSDM), applicable when he co ela ion be ween a iables is low; which is ypical in
ma ine p opulsion p ocesses.
Fi s , uel oil p ocess a iables o a ma ine diesel engine unning, poo ly co ela ed
be ween hem, we e educed h ough he analysis o co ela ions.
A e wa ds, he selec ed a iables we e moni o ed h ough Ho elling´s T2 con ol cha s
and sudden, ou -o - ange changes we e de ec ed. The a iable ha gene a ed he
de ia ion in he p ocess was iden i ied and he p edic i e a iables we e moni o ed
h ough Cusum cha s; he o igin o small and p og essi e changes in he p ocess below
he ala m h eshold se by he manu ac u e was iden i ied.
This is he accep ed manusc ip o he a icle ha appea ed in inal o m in Applied The mal Enginee ing 127 : 517-526 (2017), which
has been published in inal o m a h ps://doi.o g/10.1016/j.appl he maleng.2017.08.036. © 2017 Else ie unde CC BY-NC-ND license
(h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/)
2
The p oposed me hod (SSDM), based on he combina ion o (Ho elling´s T2 + Cusum),
can be implemen ed in any ype o p ocess in ma ine p opulsion in a sa is ac o y and
economical way, helping in he iden i ica ion o he o igin o any ype o de ia ion
(small and sudden) in he p ocess ea ly enough o implemen he igh p edic i e
ac ions.
1. In oduc ion
Nowadays, in he indus y, due o echnological ad ances and complexi y o he
p ocesses, he e a e many si ua ions in which he moni o ing o wo o mo e a iables is
equi ed [1].
Moni o ing o edundan a iables unnecessa ily inc eases he cos s o measu emen [2]
and hinde s he in e p e a ion by he use when he e is a high numbe o signals o be
moni o ed.
Wi hin he moni o ing echniques, we can dis inguish be ween compu a ional and
s a is ical. Compu a ional echniques such as A i icial Neu al Ne wo ks (ANN) ha e
been used in he moni o ing o di e en indus ial p ocesses [3]; i.e., ene gy e iciency
wi h imp o ed uel consump ion educ ion on a ma ine diesel engine.
In he s a is ical p ocess con ol (SPC), he con ibu ions ha e been made h ough he
con ol cha o Shewa [4]. This me hods moni o s a iables h ough independen
con ol cha s, igno ing he possible co ela ion o in e ac ion be ween hem, so when
he e is a a ia ion in he p ocess, se e al o hese cha s de ec i a he same ime,
being complex o de ec he exac cause o he ailu e and some imes gi ing ise o alse
ala ms.
In an a icle published in Janua y 2017 [5], in a 2-s oke ma ine diesel engine, o some
speci ic wo king condi ions, we e moni o ed a small g oup o se en a iables,
co esponding o he cylinde lub ica ion p ocess h ough Ho elling´s T2 con ol cha s
in combina ion wi h he echnique Mason, Young and T acy (MYT).
In ha wo k, he Ho elling´s T2 con ol cha , moni o s in a mul i a ia e way and
e ec i ely de ia ions a e de ec ed ega ding he op imal wo king condi ion o he
p ocess; ne e heless, small and p og essi e change wasn´ de ec ed due o he
di icul ies ha hese ypes o con ol cha ha e o de ec hese ypes o beha iou s.
Fu he mo e, he a iables ha gene a ed he de ia ion in he p ocess we e iden i ied
h ough MYT decomposi ion; i acili a ed he diagnosis o he change in he p ocess.
The p oblem comes when he size o he moni o ed a iables begins o be mode a ely
la ge, hus complica ing he in e p e a ion o he a iable which caused he de ia ion in
he p ocess.
3
The MYT decomposi ion o he T2 s a is ic has been shown o be a g ea aid in he
in e p e a ion o signaling T2 alues, bu when he numbe o a iables is g ea e han
10, and he cause o signaling is no clea om he unique e ms o decomposi ion, he
possible combina ions among hem a e inc eased exponen ially and hide which a iable
is esponsible o i . [6].
Al hough his p oblem has been no ed by o he au ho s e.g. [7], i o una ely has led o
he de elopmen o compu e p og ams ha can apidly p oduce he signi ican
componen s o he decomposi ion o mode a ely la ge se s o a iables.
Howe e , he ques ion on how hese compu a ional me hods will wo k when he e a e
hund eds o housands o a iables has ye o be answe ed.
The e o e, he e icien selec ion and educ ion o he a iables o moni o is a way o
op imize he p ocess, maximizing he e iciency and educing he cos s o measu emen
[8].
The e a e many ac o s in luencing he capabili y o his p ocedu e, and hese include
compu e capaci y, compu e speed, he size o he da a se , and he p og amming o an
algo i hm.
The e a e di e en s a is ics echniques educing he a iables o moni o such as he
P incipal Componen s Analysis (PCA). This echnique is capable o educing he
a iables space, gene a ing unco ela ed p incipal componen s (PCs) [9]; howe e ,
moni o ing h ough PCA, [10], has di icul ies in de ec ing he a iabili y o he p ocess
when he co ela ion be ween a iables is low, like in main p opulsion engines ela ed
p ocesses.
Yi ei Wang, Xiandong Ma e al. [11], p oposed an op imal senso selec ion me hod
based on p incipal componen s analysis (PCA) o condi ion moni o ing o a dis ibu ed
gene a ion (DG) sys em o ien ed o wind u bines. The aim was o iden i y a se o
a iables om a huge amoun o measu emen da a which could educe he numbe o
physical senso s ins alled o condi ion moni o ing, while main aining su icien
in o ma ion o assess he sys em´s condi ions.
The esul s showed ha unde a aul y condi ion, he algo i hm o selec ion educed he
da ase dimension and kep he i al unc ions associa ed wi h he aul in he e ained
da ase wi h a high accu acy.
In he ma ine indus y [12], he condi ion o he ship h ough sa elli e using PCA was
moni o ed. The so wa e de eloped o ansmission using PCA educes he amoun o
da a sen ia sa elli e, educing ime and cos o communica ions in case o ansmission
o all signals oge he .
Vinicius Ba oso Soa es e al. [13], implemen ed a sys em o ala m managemen
h ough he use o di e en co ela ion me hods (P incipal Componen Analysis,
Co ela ion Analysis and Clus e Analysis), in h ee na u al gas p ocessing plan s,
4
ge ing o eplace g oups o ala ms co ela ed by a mo e meaning ul one, p o ided ha
he p ocesses we e linea .
O he echniques o educ ion o a iables such as Pa ial Leas Squa es (PLS) we e
p esen ed; José Ca los Vega-Vilca e al [14], compa ed he echnical P incipal
Componen s Analysis (PCA) and Pa ial Leas Squa es (PLS) on a da abase o 252
cases, 17 p edic o a iables and 1 dependen a iable, wi h he aim o educing he
dimensionali y.
To selec he bes eg ession model, hey used he p edic i e esidual sum o squa es
(PRESS), de e mining ha he bes model o PCA was wi h 6 componen s and he
eg ession PLS was wi h 7 componen s; Fo easons o compa ison, bo h models we e
es ima ed wi h 6 componen s, being he alues PRESS o each o hem 88.31 and
266.54 espec i ely. These esul s showed ha he PLS eg ession exceeded hose o he
PCA.
These echniques o a iables educ ion can be combined wi h Ho elling´s T2 con ol
cha s o educe he limi a ion ha hey ha e when he numbe o a iables is high; S.
Joe Qin [15], analyzed he use o Ho elling´s T2 con ol cha s oge he wi h PCA and
o he me hods o de ec ion, iden i ica ion and diagnosis o ailu es; Joyce M. F. Fonseca
e al. [9], p oposed a me hodology based on he combina ion o PCA and Ho elling´s T2
con ol cha s, capable o dealing p ocesses wi h mul iple se poin s and non-s a iona y.
The p oposed me hodology was implemen ed in a he moelec ic powe plan ,
moni o ing in eal ime o de ec any changes in he ope a ion condi ions o c i ical uni s
o he powe plan , boile and u bine-gene a o uni .
Finally, he Ho elling´s T2 con ol cha s and P incipal Componen s Analysis (PCA), o
moni o ing and con ol o a mul i a ia e no mal p ocess we e p oposed in me al
indus y [16].
The Ho elling´s T2 con ol cha s de ec ed when he p ocess had de ia ions ega ding
he no mal ope a ion condi ions, bu didn´ iden i y he a iables which we e ou o
ange o possible ends ha migh be in he p ocess a iables; bu he con ol cha o
PCA de ec ed when he p ocess was ou o ange and also showed he end ha made
he p ocess o be in ha si ua ion.
As men ioned abo e, ano he ea u e o Ho elling´s T2 con ol cha s, simila o
Shewha cha s o uni a ia e p ocess con ol, is ha hey lose sensi i i y o small
changes, below 1.5 and p og essi e in he ec o o a e ages o he p ocess [17].
Thus, Apa isi F and Ga cia JC [18] es ablished a zone o a en ion as a way o inc ease
he powe o he cha o his ype o beha iou ; in low-speed machines, ailu es may
de elop slowly and hey s ay la en ill some c i ical poin o hei de elopmen in e al
when i is oo la e o ac p e en i ely [19].
In hese cases, al e na i e p ocedu es such as Cumula i e Sum cha s (Cusum) a e
widely ecommended. This ype o cha s ep esen he cumula i e sum o de ia ions,
5
which con ains in o ma ion o all he p e ious samples [20]; in his issue, du ing he
p ocess o elabo a ion o a piece o he au omo i e indus y, Shewha and Cusum
con ol cha s we e compa ed, o a same magni ude in he p ocess changes. While he
Cusum cha s de ec ed changes success ully, Shewha cha s we e no able o de ec
hem, indica ing ha he p ocess was in con ol.
These Cusum cha s [21], also ha e been used o de ec possible de ec s in he
downwind main bea ing; he me hod was as and eliable, and o e ed an es ima e on
he de elopmen o he wea as a unc ion o ime.
The e o e, Ho elling´s T2 con ol cha s lose sensi i i y o de ec small and p og essi e
changes in he p ocess and hey ha e di icul ies in iden i ying he a iable esponsible
o he change when he numbe o moni o ed a iables is g ea e han 10.
On he o he hand, e ised educ ion o a iables me hods such as PCA ha e di icul ies
o pe o m his ask, when he co ela ion be ween a iables is low.
Thus, we p opose a new Me hod o de ec ion o Small and Sudden De ia ions in he
p ocess (SSDM), applicable when he co ela ion be ween a iables is low; which is
ypical in ma ine p opulsion p ocesses.
Fi s h ough he analysis o co ela ions be ween a iables, we implemen ed a
me hodology o educe he numbe o moni o ed a iables, poo ly co ela ed be ween
hem, o uel p ocess o a ypical low-speed diesel engine unning ins alled on a anke
ship and hus o imp o e he limi a ion ha has MYT decomposi ion when he size o
he moni o ed a iables is la ge.
A e wa ds, he selec ed a iables we e moni o ed h ough he Ho elling´s T2 con ol
cha , o some speci ic wo king condi ions and h ough MYT decomposi ion; he
a iable ha caused he ou o ange s a e wi h espec o he no mal mode ope a ion o
he ship was iden i ied.
In addi ion, he echnique Ho elling´s T2 was combined wi h uni a ia e Cusum cha s,
o de ec hose a iables which can gene a e small and p og essi e de ia ions in he
p ocess, ypical in p ocess whe e he e a e he mal exchanges, and canno be de ec ed
h ough (Ho elling + MYT) con ol cha s.
The main di e ence wi h cu en li e a u e lies in he use o a new me hod called SSDM
based on he combina ion o echniques (Ho elling´s T2 + Cusum) in he main engine o
a ship in seagoing condi ions, o e ing eliable esul s and a he same ime an economic
and easy implemen a ion.
2. Ma e ial and Me hods
2.1 Machine s udy
6
The machine o s udy was he p opulsion engine o a ypical low-speed diesel engine,
which is equen ly ins alled in anke ships and bulka ie s as he main engine. The
basic echnical de ails o he engine a e lis ed in Table 1.
Manu ac u e
MAN B&W
Type
6S70ME-C8
Cycle
Low-speed
2 s oke
Nominal speed
91 .p.m.
Ra e powe
19620 kW
Numbe o cylinde s
6
S oke
2800 mm
Bo e
700 mm
Table 1. Technical de ails o engine s udied.
The engine was ins alled on a Suez Max C ude Ca ie wi h he cha ac e is ics lis ed in
Table 2. This ship no mally ca ies ou egula oyages, be ween Wes e n A ica, and
No he n Eu ope whe e she is discha ged.
Name o Ship
Con iden ial
Shipya d Build
Con iden ial
Yea Buil
2012
Type o Ship
C ude Oil Tanke
Class
Suez Max
Leng h O e all
274.20 m.
Ex eme B ead h
48.04 m.
D augh
17 m.
G oss Tonnage
81.187
Ne Tonnage
51.148
Table 2. Ship´s speci ica ions.
In he uel oil sys em [22], he uel om he se ice ank is led o an elec ically d i en
supply pump by means o which a p essu e o app oxima ely 4 ba can be main ained in
he low p essu e pa o he uel ci cula ing sys em.
F om he e he uel oil is led o an elec ically-d i en ci cula ing pump, which pumps i
h ough a hea e and a ull low il e si ua ed immedia ely be o e he inle o he engine.
This sys em is shown in Figu e 1.
7
Figu e 1 – Fuel Oil Sys em
The uel injec ion is pe o med by he elec onically-con olled p essu e boos e loca ed
on he Hyd aulic Cylinde Uni (HCU).
The Cylinde Con ol Uni (CCU) o he Engine Con ol Sys em calcula ed he iming
o he uel injec ion and he exhaus al e ac i a ion, in acco dance wi h he commands
ecei ed om he Engine Con ol Uni (ECU).
To ensu e ample illing o he HCU, he capaci y o he elec ically-d i en ci cula ing
pump is highe han he amoun o uel consumed by he diesel engine. Su plus uel oil
is eci cula ed om he engine h ough he en ing box.
2.2 Applica ion o me hod
2.2.1. S ep 1 – Da a acquisi ion
The main engine has wo moni o ing sys ems and da a acquisi ion: on one hand he
CoCos EDS, a su eillance and diagnosis con ol sys em c ea ed by he engine
manu ac u e M.A.N.; and on he o he hand he In eg a ed Au oma ion Sys em (IAS),
whe e he he modynamic p ocess da a a e collec ed.
Ou s udy only ocused on he laden condi ion due o he high a iabili y in he ballas
condi ion, moni o ing he beha iou o uel oil p ocess in he main engine du ing i s
oyage om A ica o Eu ope.
The uel oil p ocess o he main engine, was de ined by p=11 a iables: Engine Load,
Fuel Index, Tu bocha ge speed Rpm ( hey we e measu ed in he local con ol), Fuel
Plunge S oke (i is he a e age o he alue o all he injec o s), Sca enge ai coole ai
inle empe a u e (i was measu ed om inle o in e coole ), Exhaus gas empe a u e a
u bine inle (i was measu ed om inle o u bocha ge ), P (sca ) (ai p essu e inle
combus ion chambe ), Es ima e E ec i e Powe (measu ed a he sha ), Comp ession
P essu e (Pcom) and Maximum P essu e (Pmax) ( hey we e he a e age o he alue o
8
all he cylinde s, measu ed in he combus ion chambe ) and SFOC ( uel oil consumed
by he engine) measu ed in a b ake. Fo he selec ion o hese, we ha e had he
collabo a ion o he ship’s enginee s, and he manu ac u e ’s da a.
Da a acquisi ion was pe o med unde he ollowing condi ions: Speed o e g ound
(SOG) be ween 12 and 14 kno s wi h less han 18% slip, an a e age empe a u e o sea
wa e o 20 ° C, a e age ambien empe a u e o 30 ° C and a e age empe a u e o he
engine oom o 37 ° C.
Fou samples we e aken daily, om all he selec ed a iables, du ing 1 oyage which
ob ained a o al o n=47 alid samples ollowing he c i e ia p e iously men ioned.
The minimum, maximum, mean and s anda d de ia ions alues o each a e lis ed in
Table 3. Each a iable was iden i ied wi h a co ela i e numbe ing.
No.
Va iables
Uni
Min.
Value
Max.
Value
Means
(µ)
S anda d
De ia ions
(σ)
1
Engine Load
%
54
61
56.85
1.546
2
Fuel Index
%
62.6
70.6
65.136
1.7644
3
Fuel plunge s oke
mA
2.58
2.77
2.662
0.0358
4
Sca enge ai coole ai inle
empe a u e
°C
142
160
149.21
4.287
5
Exhaus gas empe a u e a u bine
inle
°C
354
407
372.63
15.306
6
Tu bocha ge speed
.p.m.
10366
11202
10818.84
185.504
7
P (sca )
Ba
1.57
1.96
1.834
0.1093
8
Es ima e E ec i e Powe
kW
10318
10909
10540.11
130.761
9
Comp ession P essu e, Pcom
Ba
110.18
129.7
123.832
5.6346
10
Maximum P essu e, Pmax
Ba
138.43
142.47
140.739
1.0883
11
SFOC
g/kWh
154.28
164.01
158,511
2,042
Table 3. Means, s anda d de ia ions, maximum and minimum alues.
The e was a p oblem o los da a. The a ailable da a sampling pe iod on boa d was oo
slow, he e o e, i was necessa y he use o in e pola ion echnique o ge he samples
needed o implemen he me hod; i hese pe iods had been sho e , i.e, one sample,
e e y hal hou , he adjus ed R2 coe icien s would had been highe , he eby inc easing
he eliabili y o he me hod.
Fu he mo e, o c ea e he p elimina y da abase, n=599 samples we e gene a ed o each
a iable h ough cubic spline in e pola ion [23].
Wi h his, he numbe o samples needed o alida e he s udy was achie ed.The
minimum sample size ollows he equa ion (2) acco ding o he numbe o a iables, p
[24]:
(2)
9
2.2.2. S ep 2 Va iable selec ion
Wi h he samples gene a ed in he p elimina y da abase, a Pea son co ela ion analysis
was pe o med [25], among he 11 a iables in which uel p ocess was de ined, lis ed in
able 4.
Va iable iden i ica ion numbe
1
2
3
4
5
6
7
8
9
10
11
Va iable iden i ica ion numbe
1
1.000
0.880
0.750
0.245
0.114
0.171
0.096
0.415
0.012
0.047
0.277
2
0.880
1.000
0.769
0.373
0.125
0.282
0.156
0.358
0.046
-0.340
0.262
3
0.750
0.769
1.000
0.181
0.101
0.119
0.067
0.494
-0.025
0.142
0.384
4
0.245
0.373
0.181
1.000
0.160
0.790
0.414
0.409
0.235
-0.313
0.340
5
0.114
0.125
0.101
0.160
1.000
-0.392
-0.792
-0.061
-0.900
-0.693
-0.377
6
0.171
0.282
0.119
0.790
-0.392
1.000
0.857
0.494
0.739
0.115
0.623
7
0.096
0.156
0.067
0.414
-0.792
0.857
1.000
0.404
0.975
0.457
0.648
8
0.415
0.358
0.494
0.409
-0.061
0.494
0.404
1.000
0.286
0.318
0.843
9
0.012
0.046
-0.025
0.235
-0.900
0.739
0.975
0.286
1.000
0.563
0.570
10
0.047
-0.034
0.142
-0.313
-0.693
0.115
0.457
0.318
0.563
1.000
0.439
11
0.277
0.262
0.384
0.340
-0.377
0.623
0.648
0.843
0.570
0.439
1.000
Table 4. Co ela ion be ween a iables.
The p ocess was moni o ed by he minimum numbe o a iables, (p=3 a iables: Fuel
Index, Exhaus gas empe a u e a u bine inle and Tu bocha ge speed), ollowing he
c i e ia: he selec ed a iables ha e one co ela ion be ween hem less han 0.49 and he
selec ed a iables had a co ela ion wi h a leas one o he unselec ed a iables equal o
o highe han 0.49.
Finally, h ough SPSS so wa e, i was ound he adjus men o models among he h ee
selec ed a iables and hei p edic i e a iables using a mul i a ible eg ession analysis,
ob aining he ollowing coe icien s o de e mina ion R2 adjus ed, 0.8, 0.95 and 0.96 o
each model espec i ely.
Con en ional me hods o a iable educ ion such as PCA we e no e icien ; wi h wo
p incipal componen s ep esen ed only he 81% o he p ocess. Fi e p incipal
componen s we e equi ed o ep esen 96% o he p ocess.
16
caused he de ia ion om i s no mal ope a ing mode was he Tu bocha ge speed
a iable.
Obse a ions
Va iables
1
Exhaus gas empe a u e
4
Tu bocha ge speed
5
Tu bocha ge speed
6
Exhaus gas empe a u e and Tu bocha ge
speed.
7
Fuel Index
8
Tu bocha ge speed
10
Fuel Index
11
Tu bocha ge speed
12
Tu bocha ge speed
13
Tu bocha ge speed
Table 7. Decomposi ion MYT
3.3 Applica ion o Cumula i e sum
In his s age, i was moni o ed he p edic i e a iables o he Tu bocha ge speed
a iable, using he Cusum cha s, o de ec i any o hem was esponsible o he ou o
ange s a e o he p ocess.
The mean, s anda d de ia ion alues o each p edic i e a iable in con ol a e lis ed in
Table 8.
Va iables
Uni
Means(µ)
S anda d
De ia ions
(σ)
Sca enge ai coole ai inle
empe a u e
°C
147.35
2.24
P (sca )
Ba
1.84
0.1
Es ima e E ec i e Powe
kW
10542.36
103.63
Comp ession P essu e, Pcom
Ba
124.8
5.38
SFOC
g/kWh
158.58
1.7
Table 8. Mean and s anda d de ia ion o p edic i e a iables.
61 obse a ions o each o he a iables we e moni o ed; he i s 48 obse a ions
co esponded o he ARL0 and he ollowing 13 we e new inpu da a.
Figu es 3a, 3b, 3c, 3d, 3e, show Cusum cha s o each o he a iables. I was no ed
ha he only a iable ha exceeded i s decision in e al was he SFOC a iable, whe e
17
a sample 50 is C50+ = 10.8. Since his is he i s pe iod a which Ci+ > H=8.5, we would
conclude ha he a iable was ou o ange in his poin .
Howe e , he abula Cusum also indica es when he shi p obably occu ed. The i s
consecu i e sample in which Ci+ > 0 i s exceed he alue o H, was he pe iod 49, C49+
= 5.36, hus indica ing ha he misma ch in he a iable could ha e s a ed in he sample
49.
Figu e 3a - Sca enge ai coole ai inle empe a u e (ARL1=1.23).
18
Figu e 3b - P (sca ) (ARL1=4.29)
Figu e 3c - Es ima e E ec i e Powe (ARL1=0.79)
19
Figu e 3d - Pcom (ARL1=3.37)
Figu e 3e - SFOC (ARL1=0.87)
20
4-Discussion
The uel oil p ocess o a 2-s oke ma ine diesel engine was moni o ed by only h ee
a iables wi h low co ela ion be ween hem, h ough a combina ion o uni a ia e and
mul i a ia e echniques (Ho elling´s T2 + Cusum).
Ho elling´s T2 con ol cha s pe o mance dec eased as i inc eased he numbe o
a iables o be moni o ed. I was chosen he minimum numbe o a iables o be
moni o ed, p=3, om among he 11 a iables ep esen ing he en i e p ocess h ough a
mul i a ia e eg ession analysis, ensu ing i ing models be ween a iables and hei
p edic i e a iables, wi h coe icien s o de e mina ion R2 adjus ed highe han 0.8.
Mul i a ia e cha s de ec ed obse a ions ou o ange wi h espec o he op imal
condi ions o he p ocess; in he able 9, he e is he ch onology o he ou o ange
obse a ions, wi h i s espec i e T2 alues.
Da e
Obse a ions
T2
25/08/2016
1
8.042
28/08/2016
4
30.395
29/08/2016
5
9.294
30/08/2016
6
21.884
31/08/2016
7
16.745
01/09/2016
8
8.489
03/09/2016
10
31.785
04/09/2016
11
37.641
05/09/2016
12
21.212
06/09/2016
13
221.034
Table 9. Ch onology o Obse a ions.
Obse a ions ha we e abo e he limi o con ol we e decomposed, iden i ying he
a iable u bocha ge speed as he main a iable ha o igina ed he ou o ange s a e o
he mul i a ia e p ocess.
Ho elling´s T2 Technique has he ad an age ha e ec i ely de ec s high and sudden
changes in he p ocess bu can’ de ec small and p og essi e changes.
Fo his eason, he p edic i e a iables in he a iable u bocha ge speed we e
moni o ed h ough Cusum cha s, o y o de ec small and p og essi e changes in he
p ocess ha had no been de ec ed by means o mul i a ia e cha s.
I was es ablished a decision in e al, less han he one ma ked by he manu ac u e ,
15% o e he a e age alue o each a iable in op imal condi ion ope a ion. The SFOC
a iable exceeded he h eshold and was de ec ed when i began o de ia e om i s
no mal condi ion be o e he es ablished h eshold.
21
The cleaning o he in e coole , o se ice easons, only was made wi h chemical
p oduc s, he las cleaning in dep h had been 6 mon hs ago; his si ua ion gene a ed a
p og essi e ouling in he in e coole . In o de o main ain he speed o he essel a
small de ia ion in he SFOC a iable was caused.
5- Conclusions
The p oposed me hodology o educ ion o a iables, h ough he analysis o
co ela ions be ween a iables, was capable o educe he numbe o a iables, poo ly
co ela ed be ween hem, o uel p ocess o a unning ma ine diesel engine;
con en ional me hods o a iable educ ion such as PCA was shown ha we e no
e icien when he co ela ion be ween a iables was poo .
Th ough p oposed me hodology o moni o ing o a iables SSDM based on he
combina ion o (Ho elling T2 + Cusum) cha s, high and sudden and also small and
p og essi e de ia ions in he p ocess we e de ec ed.
The alue o he di e en ial p essu e in he in e coole was no enough o o e come he
h eshold se by he manu ac u e ; a small de ia ion in he SFOC a iable was
gene a ed. Wi hou his iden i ica ion, hey would ha e had o wai o he alue o he
di e en ial p essu e was abo e he h eshold se by he manu ac u e , esul ing in a
highe ouling o he in e coole and an inc ease o he SFOC a iable.
Many p ocesses in ol ed in he ope a ion o a ma ine diesel engine ha e decay small
and p og essi ely in addi ion o suddenly, o his eason, his me hod has he ad an age
ha is can be cus omized o any ype o engines because i is capable o de ec ing any
ype o de ia ion (small and sudden) in he p ocess; his can be pe o med in a simple
and economical way, a he eques o he shipowne , depending on he ope a ional
condi ions o he ship.
In u u e wo k, he e ec i eness o he me hod using Mul i a ia e Cusum Cha s
(MCusum) in his ype o p ocess could be s udied.
6- Re e ences
[1] Be simis S, Psa akis S, Pana e os J. Mul i a ia e s a is ical p ocess con ol cha s: an
o e iew. Qual Reliab Eng In 2007;23(5):517-543.
[2] González I, Sánchez I. Va iable selec ion o mul i a ia e s a is ical p ocess con ol.
Jou nal o Quali y Technology 2010;42(3):242.
[3] Basu ko OC, U iondo Z. Condi ion-Based Main enance o medium speed diesel
engines used in essels in ope a ion. Appl The m Eng 2015;80:404-412.
22
[4] Ga cia-Al a ez D, Fuen e M. Es udio compa a i o de écnicas de de ección de allos
basadas en el Análisis de Componen es P incipales (PCA). Re is a Ibe oame icana de
Au omá ica e In o má ica Indus ial RIAI 2011;8(3):182-195.
[5] Boullosa D, La abe JL, Lopez A, Gomez MA. Moni o ing h ough T 2 Ho elling o
cylinde lub ica ion p ocess o ma ine diesel engine. Appl The m Eng 2017;110:32-38.
[6] Mason RL, Young JC. Mul i a ia e s a is ical p ocess con ol wi h indus ial
applica ions. : Siam; 2002.
[7] Kou i T, MacG ego JF. Mul i a ia e SPC me hods o p ocess and p oduc
moni o ing. Jou nal o quali y echnology 1996;28(4):409-428.
[8] Con ol Es adís ico Mega Va ian e pa a los p ocesos del siglo XXI. Dp o de
Es adís ica e In es igación Ope a i a. UPV. Cong eso Nacional de Es adís ica e
In es igación Ope a i a. Valencia, España; 2003.
[9] Moni o ing o a he moelec ic powe plan based on mul i a ia e s a is ical p ocess
con ol. E ol ing and Adap i e In elligen Sys ems (EAIS), 2016 IEEE Con e ence on:
IEEE; 2016.
[10] Peña D. Análisis de da os mul i a ian es. : McG aw-Hill Mad id; 2002.
[11] Wang Y, Ma X, Joyce MJ. Reducing senso complexi y o moni o ing wind
u bine pe o mance using p incipal componen analysis. Renewable Ene gy
2016;97:444-456.
[12] La abe J, Gómez M, So és I, Al a ez F, Rey M, Mielgo V, e al. Vessel condi ion
moni o iza ion h ough sa elli e using p incipal componen analysis. Jou nal o
Ma i ime Resea ch 2014;7(3):3-14.
[13] Soa es VB, Pin o JC, de Souza MB. Ala m managemen p ac ices in na u al gas
p ocessing plan s. Con ol Eng P ac 2016;55:185-196.
[14] Vega-Vilca JC, Guzmán J. Reg esión PLS y PCA como solución al p oblema de
mul icolinealidad en eg esión múl iple. Re is a de Ma emá ica Teo ía y Aplicaciones
2011;18(1):09-20.
[15] Joe Qin S. S a is ical p ocess moni o ing: basics and beyond. J Chemome ics
2003;17(8-9):480-502.
[16] Zama ón AMC, P ado EM, Luis FZ. Moni o eo y con ol de un p oceso no mal
mul i a iado. Conciencia Tecnológica 2012(43):29-35.
[17] Mason RL, Young JC. Imp o ing he sensi i i y o he T2 s a is ic in mul i a ia e
p ocess con ol. Jou nal o Quali y Technology 1999;31(2):155.
[18] Apa isi F, GARCIA JC. Aumen o de la po encia del g á ico de con ol
mul i a ian e T 2 de Ho elling u ilizando señales adicionales de al a de con ol.
Es adís ica española 2001;43(148):171-188.
23
[19] S amboliska Z, Rusiński E, Moczko P. P oac i e condi ion moni o ing o low-
speed machines. P oac i e Condi ion Moni o ing o Low-Speed Machines: Sp inge ;
2015. p. 53-68.
[20] Rod íguez LI, Rod íguez MI, Rod íguez MA. Análisis Compa a i o de los
G á icos de Con ol Shewha , CUSUM y EWMA, pa a P ocesos en Condiciones de
Con ol y Fue a de Con ol.
[21] S a is ical aul diagnosis o wind u bine d i e ain applied o a 5MW loa ing
wind u bine. Jou nal o Physics: Con e ence Se ies: IOP Publishing; 2016.
[22] MAN B&W S70ME-C8.5-TII – P ojec Guide. Elec onically Con olled wo-
s oke engines. Copenhagen: MAN B&W Diesel, 2016. A ailable a :
h p://ma ine.man.eu/applica ions/p ojec guides/2s oke/con en /p in ed/S70ME-
C8_5.pd . Accessed 12/08, 2016.
[23] McKinley S, Le ine M. Cubic spline in e pola ion. College o he Redwoods
1998;45(1):1049-1060.
[24] Williams B. A sample on sampling. : JSTOR; 1978.
[25] Pea son K. Ma hema ical con ibu ions o he heo y o e olu ion. III. Reg ession,
he edi y, and panmixia. Philosophical T ansac ions o he Royal Socie y o
London.Se ies A, con aining pape s o a ma hema ical o physical cha ac e
1896;187:253-318.
[26] Ho elling H. Mul i a ia e quali y con ol. In C. Eisenha , M. W. Has ay and W.A.
Wallis, eds.”, Techniques o s a is ical analysis 1947. New Yo k: McG aw Hill, pp 111-
184,1947
[27] Mason RL, Chou Y, Sulli an JH, S oumbos ZG, Young JC. Sys ema ic pa e ns in
T (2) cha s. Jou nal o Quali y Technology 2003;35(1):47.
[28] Ma a elakis P. An in es iga ion o some cha ac e is ics o uni a ia e and
mul i a ia e con ol cha s. Unpublished doc o al disse a ion, Depa men o S a is ics,
A henas Uni e si y o Economics and Business.(A ailable om in e ne :
h p://www.s a -a hens.aueb.g /~jpan/dia i es/Ma a elakis/Index.h ml) 2003.
[29] Mason RL, T acy ND, Young JC. A p ac ical app oach o in e p e ing
mul i a ia e T2 con ol cha signals. Jou nal o Quali y Technology 1997;29(4):396-
406.
[30] Agog NS, Dikko HG, Asi ibo OE. Decomposi ion ho elling's T2 s a is ic using
ou a iables. Jou nal o Inno a e Resea ch in Science, Enginee ing and Technology
2014;3 (4).
[31] Mason RL, T acy ND, Young JC. Decomposi ion o T2 o mul i a ia e con ol
cha in e p e a ion. Jou nal o quali y echnology 1995;27(2):99-1108.
24
[32]Mon gome y DC. In oduc ion o s a is ical quali y con ol. John Wiley & Sons;
2009. 6 h. ed.
[33] Siegmund DO. Sequen ial Analysis: Tes s and Confidence In e als Sp inge -
Ve lag. New Yo k 1985.
