Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Vib a ion Analysis and Op imal Design o
Pneuma ic Ci cui s Using A i icial Neu al
Ne wo ks
Şahin YILDIRIM
Mecha onic Enginee ing Depa men
E ciyes Uni e si y, Facul y o Enginee ing
Kayse i, Tu key
[email p o ec ed]
Aslı DURMUŞOĞLU
Mechanical Enginee ing Depa men
E ciyes Uni e si y, Facul y o Enginee ing
Kayse i, Tu key
[email p o ec ed]u.
Abs ac —Pneuma ic sys ems a e commonly used in indus y
p ocesses and applica ions o au omo ion. These sys ems a e
make a ac i e powe ansmission wi h he comp essed ai ,
because o hey a e economic, clea , sa e and simple
s uc u ed.So in his sys ems, as noise and ib a ion e ec s, i ’ s
undesi able si ua ions bo h human heal h and sys em
pe o mances yield and wo king li e. In his s udy, ib a ion and
noise da as a e ob ained om wo ype o pneuma ic sys ems
p o o ype which a e classi y me al and non-me al ma e ials and
pe o med an analysis wi h help o his da as and used neu al
ne wo k which has adap i e and quick con uc ion.
Keywo ds—pneuma ic ci cui s; ib a ion analysis; neu al
ne wo ks
I. INTRODUCTION
Capable o con olling he p essu e o he si ua ion ha can
be changed, wi h sys ems ha a e unning ai and gases,
pneuma ic sys ems a e called. Pneuma ic sys ems; clean,
secu e and wi h comp essed ai powe ansmission due o i s
simple s uc u e makes i a ac i e. Thanks o his sys em,
au oma ion p oduc ion is smoo h, as and con ollable
sys ems has gained impo ance in ad ancing he echnology
o use as. And conside ing he a ea hey a e used in
au oma ion echnology, his impo ance is e en mo e e iden .
Pneuma ic cylinde s, ep oducibili y, low cos and has a
wide use in indus ial applica ions in e ms o supe io
pe o mance. In many cases, o achie e good con ol
pe o mance, nonlinea con ol echniques ha e been used.
Simaui and colleagues [1] one o hese echniques by using he
cascade me hod he back elec o-pneuma ic posi ion con ol
sys em ha e wo ked on.
Pneuma ic sys ems, indus ial p ocesses, and a e used
ex ensi ely in au oma ion applica ions. Howe e , his sys em
mus be made o he analysis o ins an aneous dynamic
changes.
Akkaya and his colleques [2] ha e been ca ied ou he
simula ion o he dynamic p ope ies o a linea pneuma ic
ac ua o sys em, and hus s udied he e ec o pa ame e
a ia ions o he sys em despi e. Simula ion s udy in he
Ma lab-Simulink model was ca ied ou using a compu e
p og am c ea ed in.
Pneuma ic ib a ion isola ion sys em (PVIS), e ec i e o
ib a ion isola ion due o he pe o mance, a e widely used.
Moon and Lee [3] we e in ended o do p is sys ems modelling
and sensi i i y analysis. Pu and colleagues [4] in hei s udy o
he police sys em o adjus he damping cha ac e is ics o
double-chambe pneuma ic sp ing is used. The pu pose o his
adjus able p is o imp o e he pe o mance cha ac e is ics o
he sys em.
In his s udy, wo di e en sepa a o pneuma ic es and
sepa a ion s a ion o ANNS (a i icial neu al ne wo ks) using
ib a ion analysis and modeling a e e iewed. Minimum and
maximum alues o ma e ials o di e en weigh and
p essu es aken a he p essu e di e ence cylinde pneuma ic
ib a ion he e ec s o ib a ion on he sys em we e
in es iga ed.
II. SEPARATOR PNEUMATIC TEST AND SEPARATION
STATIONS
A. The Physical S uc u e O The Sys em
Sepa a o and sepa a ion PLC con olled pneuma ic es
s a ions, colo , heigh , and ese oi samples a e au oma ically
seg ega ed in o di e en me al wi h di e en cha ac e is ics as
o whe he o no . The sepa a ion s a ion also is con olled
wi h wo pneuma ic cylinde s and a PLC con olle wi h DC
mo o s and he mo emen and he sepa a ion p ocess a e
p o ided. The samples o he de e mina ion o he p ope ies
o a ious senso s dis ance senso s induc i e senso s a e used.
The i s es s a ion and componen s o he sys em and
sepa a ion Figu e 1 seen. Sampling a his s a ion, which is
mo ed by a DC mo o o ha piece o p og ess is p o ided by
he ma ching band. As can be seen om he igu e, he i s
pa o he ape consis s o a ious senso s o making
measu emen s and es . The senso s henacco ding o he
esul s o he es p ocess, o pe o m he sepa a ion p ocess,
pneuma ic cylinde s, and ma e ial o he channels is loca ed.
This es samples a he s a ion and sepa a ion o whi e and
black me al depending on whe he hey a e pneuma ic
sepa a o s is pa sed h ough o di e en si es.
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Fig 1.Tes and sepa a ion S a ion and he sys em componen s(Mechanism 1)
Some ins umen s o he Figu e 1, can be desc ibes in he
ollowing;
1.Elec o-pneuma ic solenoid al e g oup
2. and 3. Op ic senso s
4. The dis ance (heigh ness) senso
5. Induk i senso
6. Con as senso
7. Single ac ing pneuma ic cylinde s
8. Bel
9. Re lec o s
10. Ma e ial
11. P og amming cables
12. PLC
13. Ope a o panel
The second es seen in Figu e 2, and he sepa a ion S a ion
and he sys em componen s. A his s a ion he p og ess o a
DC Mo o he mo ion o pneuma ic cylinde s and sample
piece p o ided by he sys em. Samples placed in he s a ing
posi ion o he i s senso and he measu emen s o be made
a e being mo ed o wo posi ions. The las poin o he
measu emen esul om he ma e ial o he channel abo e o
below, again a e pushed by a pneuma ic cylinde .
Fig 2. Tes and sepa a ion S a ion sys em componen s(Mechanism 2)
Sepa a o and sepa a ion PLC con olled pneuma ic es
s a ions, colo , heigh , and as o whe he o no o me al
Figu e 2 common sys em componen s;
1. PLC (S7-300)
2. Pneuma ic Cylinde s
3. DC Mo o
4. Elec o-Pneuma ic Val e G oup
5. Pneuma ic Vacuum G ippe
6. Senso s
The schema ic o he sys em s uc u e, he s uc u e is qui e
simila o he p e ious schema ic o he s a ion. (Fig 1.) The
di e ence he e, he sys em consis s o wo wo ks a ions. The
mo emen o he sample o he anspo o he i s s a ion,
he second S a ion and he measu emen s a e used o he
pa sing p ocess. These wo s a ions, and is con olled by wo
di e en PLC s a ions wo ks in e ac i ely wi h each o he .
Tes specimens om his es and sepa a ion s a ion he Black,
Whi e and whe he hey a e p ope ies such as me al a e
s udied. Also he e is a hole on one side o he samples. In his
way he specimen is placed ha will be pe o med o check
whe he o no la o in e ed.
B. Ma hema ical S uc u e O The Sys em
In es iga ed pneuma ic sys ems, single ac ing cylinde ,
double ac ing cylinde and elec o-pneuma ic di ec ional
con ol al es a e discussed as including h ee sepa a e
elemen s. The s anda d o i ice using he heo y o dynamic
p ope ies o elemen s and each elemen is gi en ma hema ical
model o a pneuma ic.
When you c ea ed he model, he ollowing assump ions
we e made
 Ai is an ideal gas.
 Feeding p essu e (Ps) is cons an
 Exhaus p essu e (PR) is equal o a mosphe ic
p essu e (Pa).
 The p ocess is isen opic.
 Feed cylinde is cons an and equal o he
empe a u e he empe a u es in he compa men s
in he ank.
 The dynamics o he pis on al e and he hose has
been neglec ed.
1. Cylinde Model
Push and pull o ces single ac ing pneuma ic cylinde s can
be calcula ed as ollows. In he di ec ion o he h us o ce
han he o ce in he di ec ion o pull. The eason o his is
due o ensile comp essed ai in he di ec ion o he applied
ield will be smalle han he sha s.
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Fig 3. Single ac ing pneuma ic cylinde [8]
Theo e ical o ce calcula ion;
(1)
(2)
Figu e 4, 4/2 di ec ional con ol al e double ac ing cylinde
con olled by a also has been gi en. He e 󰇗 󰇘, espec i ely,
posi ion, speed, accele a ion [m, m/s ,m/s2 ]; o cylinde 1.
and 2. he p essu e in he compa men s [Pa]; supply and
exhaus p essu e [Pa]; cylinde 1. and 2. he a ea on he
side o he pis on [m2]; 󰇗 cylinde 1. and 2. on he side o
he mass ai low [kg/s]; con ol sys em and al e pis on
weigh M [kg] ep esen s.
Fig 4. Double ac ing pneuma ic cylinde [8]
The eed p essu e o he al e posi ion o which
de e mines how he cylinde is opened o a mosphe ic p essu e
o o. The load o wa d (+) o mo e he cylinde 1.
compa men opens in o he eed ank. Simul aneously, he
cylinde 2. he compa men is opened o he a mosphe e. This
condi ion gi es ise o he p essu e di e ence be ween he
compa men s o he cylinde . The di e ence in p essu e
inc eases enough o de ea ic ion and ex e nal o ces when
he pis on and hus he load s a s o mo e. The equa ion o
mo ion o he cylinde is as ollows [5];
(3)
󰇘 (4)
󰇘 (5)
( 󰇗 󰇘 ) (6)
Equa ion (4), he ne o ce on he sys em due o he
di e ence in p essu e, i he o ce o ic ion is .
Vol s mass ai low comp essed om he cylinde elemen
o he inpu a iables, he posi ion and eloci y o he pis on.
A cylinde 1. ou pu a iables and 2. he p essu e in he
compa men s. The ela ionships be ween p essu e and low
a e gi en below [6];
󰇗
[ 󰇗
󰇗] (7)
󰇗
[ 󰇗
󰇗] (8)
2. Val e Model
Val e elemen om he p e ious elemen o
p essu ized ai and an elec ical signal inpu o he a iable
ha is se by he loca ion o he al e. As ou pu a iables
depending on he posi ion o he al e cylinde 1. pane o 2.
en e ed in o he pane, o he comp essed mass ai low.
The po a ea al e , and al e discha ge coe icien
cons an s, such as inpu sys em can be conside ed as
cons an . The ela ionships be ween al e inpu and ou pu
a iables a e gi en in he ollowing equa ions [5].
The pis on cylinde o wa d (+) mo emen in he case;
󰇗
√ ( ) (9)
󰇗
√ ( ) (10)
whe e;
{
(
)
√ (
)
(11)
The pis on cylinde backwa d (-) in he case o p og ess;
󰇗
√ ( ) (12)
󰇗
√ ( ) (13)
whe e;
{
(
) √ (
)( ) (14)
√
(
)
√
( ) ; (
)
(15)
whe e is he speci ic hea a io, he gas cons an c i ical
p essu e and he mass low pa ame e .
III. EXPERIMENTAL WORK
Two di e en alues a e measu ed a di e en p essu es
o di e en es and sepa a ion s a ion he ib a ional
modeling a i icial neu al ne wo ks a e e iewed. S a ions o
es ing and so ing he ma e ial in 3 di e en weigh s wi h he
ib a ion alues a e measu ed a minimum and maximum
p essu es in pneuma ic cylinde and has pe o med analysis
and modeling. While modeling algo i hms ha belong o 5
di e en a i icial neu al ne wo ks algo i hms ha gi e he
bes and wo s esul s a e p esen ed in g aphs and ied.
Tes and sepa a ion s a ion he i s pneuma ic sepa a o
o black, whi e and me al (25 g, 15 g, 10 g) single ac ing
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
cylinde used o sepa a ing h ee ma e ials, (1., 2. and 3.
measu emen poin s) 2 ba wo king p essu e and 6 ba
ib a ion alues we e aken.
Fig 4. Sepa a o pneuma ic es and sepa a ion s a ion and an expe imen al
se up o measu ing he accele a ion (1).
Second sepa a o pneuma ic es and sepa a ion s a ion 3
di e en weigh s (50 g, 100 g, 240 g) ma e ial o a wo king
p essu e o 2 ba and 6 ba , wo double ac ing cylinde s (1.
and 2. measu ing poin ) and DC mo o (3. measu ing poin
h ough he ib a ion alues we e aken.
Fig 5. Sepa a o pneuma ic es s a ion and an expe imen al se up o
measu ing he accele a ion and sepa a ion (2).
IV. ARTIFICIAL NEURAL NETWORKS (ANN )
Sepa a o pneuma ic es s a ions, expe imen al se ups and
sepa a ion o he B uel - Kjae po able and mul i-channel
pulse da a has been ecei ed and wi h he help o
accele ome e senso s using a i icial neu al ne wo ks by using
modeling and analysis conduc ed. In Figu e 6 The Block
diag am o his si ua ion a e gi en.
Fig 6. Block diag am o expe imen al sys em
The mos ypical o m o he NN s uc u e, he cell
gene a ed wi h he model o he laye s o be assembled can be
ins alled wi h he sequen ial o ma . A neu al ne wo k model
used is shown in Figu e 7.
Fig 7. ANN’ s schema ic ep esen a ion
He e he applied laye o he inpu laye o he inpu signal,
he ou pu signal o he eceip o he esponse laye is called
he ou pu laye . In he p esen s udy, a linea cell in he inpu
laye , hidden and ou pu laye s, espec i ely, en and h ee
nonlinea cell was used. Gi en he ne wo k ob ained o he
aining, 70%, o he es phase, 30% we e used. ANN’ s da a
o be used o he pu pose o es ing o he s able s uc u e o
he ne wo k du ing he aining phase we e no used.
A. Back P opaga ion Algo i hm
The mos common applica ions used in he lea ning
algo i hm. Mos p e e ed because i can be easily e i ied
ma hema ically easy o unde s and and he lea ning algo i hm.
Re e se his algo i hm o educe e o s back p opaga ion go
i s name because o wo king om he ou pu o inpu [7].
Back p opaga ion algo i hm, he slope is dec easing and
MLP’ s he mos basic lea ning algo i hm used in aining. In
his algo i hm he weigh s be ween he elemen s and wi h
p ocessing imes ( ), he change in is calcula ed. This
exp ession,
( ) ( ) (16)
as i is p o ided. The equali y (16) he lea ning
coe icien , momen um coe icien and o belonging o any
neu on in he ou pu laye Dec o is a ac o . Fo he ou pu
laye , his ac o is p o ided in he ollowing manne ;
( ) (17)
whe e, ∑ (18)
and he a ge ou pu o he elemen p ocesso . Dec
ac o o he neu ons in he laye s;
(
)∑ (19)
B. QuickP opaga ion Algo i hm (QP)
Quick p opaga ion algo i hm, de eloped by Fahlman and
new on me hod based on a heu is ic lea ning algo i hm used
o MLP aining. P opaga ion es s a e usually as when
compa ed wi h o he echniques he pe o mance o he
algo i hm is p e y good. This algo i hm gi es be e esul s in
less da a p oblems wi h he noise le el especially [7].
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
This algo i hm is neglec ed and he weigh in accele a ion
and clipped he om up o in weigh change,
( ) ( ) ( ) (20)
wi h he o mula is ob ained. He e ε is he lea ning
coe icien , and momen um coe icien α is. He e o upda e he
weigh unc ion, he del a unc ion and pa icipa es in he
p ocessing o he accele a ion due o g a i y, he coe icien o
weigh . ( ) ( ) ( ) ( ) (21)
He e is he coe icien accele a ion. Finally, i he weigh
is oo small, 0 is clipped by aking. | ( )| i ( ) is
aken. He e, he weigh ac o c op [6].
C. Del a- Ba - Del a Algo i hm
Del a-Ba -Del a mul i-laye ed pe sep on con e gence o
he connec ion weigh s in a heu is ic app oach ha is used o
inc ease he speed. In expe imen al s udies, he weigh o each
dimension in e ms o space ha e shown ha all he e o
su ace can be comple ely di e en . To explain he changes in
he e o su ace, especially he coe icien o each link o he
ne wo k should ha e own lea ning. The app op ia e s ep size
o he size weigh only he weigh o his hough may no be
sui able o all sizes [7]. The s anda d back p opaga ion
algo i hm, he slope componen is p o ided in he ollowing
way, ( ) ( )
( ) (22)
He e ( )a he ime o he e o alue , ( )and he
weigh o he connec ion ( )illus a es he slope o he
change in weigh componen . The s anda d back p opaga ion
algo i hm o he Connec ion Weigh ;
( ) ( ) ( ) (23)
will be upda ed as. He e is a lea ning cons an
coe icien . DBD lea ning ule, he lea ning a e o each
connec ion, he a iable ( )is assigned o, and upda e o he
Connec ion Weigh ;
( ) ( ) ( ) ( ) (24)
V. SIMULATION RESULTS
The scope o wo k es ing and so ing s a ions in i e
expe imen al se ups o ib a ion analysis neu al ne wo k
algo i hm was used. Expe imen ally ob ained ib a ion da a is
aken as he desi ed alue. A S a ion 107 he i s es o
ib a ion and sepa a ion o he ne wo k he aining da a, he
ib a ion da a we e used in he es ing phase 50. 175 he
second es and sepa a ion s a ion o he ne wo k du ing
aining, ib a ion da a, ib a ion da a we e used o he es
phase 90. ANN’ s e alua ion o he accu acy o modeling wi h
he da a o be used du ing aining o he da a o be used in
he es phase is di e en .
Vib a ion analysis o he s udy o back p opaga ion
algo i hm, ba ch algo i hm, backp op, del a-ba -del a
algo i hm, quickp opaga ion ( as p opaga ion) algo i hm and
p op ( lexible p opaga ion) algo i hm was used. The i s es
and sepa a ion s a ion he bes esul s o as p opaga ion
algo i hm, while he wo s esul s del a-ba -del a algo i hm
we e ob ained.
The second es and sepa a ion s a ion o bes esul s he
wo s esul s ha e been ob ained in he back p opaga ion
algo i hm and is s ill he del a-ba -del a algo i hm we e
ob ained.
TABLE I. CHARACTERSTİCS OF DİFFERENT LEARNİNG ALGORİTHMS İN
FEEDFORWARD ANNS
A. Max. and min. p essu e in he simula ion esul s
(Mechanism 1)
Max. and min. he apid p essu e p opaga ion algo i hm
and Del a - ba - Del a he esul s o he simula ion igu e 8,9,
10 and 11 a e gi en.
Measu emen poin -1
Measu emen poin -2

Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Measu emen poin -3
Fig 8. Di e en measu emen poin s max.expe imen al and neu al ne wo k
esul s o p essu e( as -p opaga ion algo i hm)
In he abo e cha max. apid p essu e p opaga ion
algo i hm esul s a e gi en. Figu e. 8. as seen, he highes
ib a ion 2. ANN and occu ed la gely a he poin o
measu emen o he desi ed answe unco e ed. 1. and 3. he
ib a ion alues a he measu emen poin s smalle neu al
ne wo k has been able o ack he desi ed alues.
Measu emen poin -1
Measu emen poin -2
Measu emen poin -3
Fig 9. Di e en measu emen poin s max.p essu e-expe imen al and neu al
ne wo k esul s(Del a-ba -del a algo i hm)
Del a-ba -del a neu al ne wo k algo i hm wi h h ee
measu emen poin s in he desi ed esponse could no ca ch
up. This algo i hm ANN’ s in he es phase he co ec
pe cen age is 15%.
Measu emen poin -1
Measu emen poin -2
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Measu emen poin -3
Fig 10. Di e en measu emen poin s o min.expe imen al and neu al
ne wo k esul s o p essu e( as -p opaga ion algo i hm)
Figu e 10, he maximum ib a ion alue, as shown in 2.
we e aken om he measu emen poin . Fas -p opaga ion
algo i hm neu al ne wo k ga e he desi ed answe . 1. despi e
he lowe alue o he ib a ion measu emen a he poin o
sudden NN could no ca ch up wi h he changes a he poin s
whe e he desi ed alues. 3. he measu emen poin o he NN,
he desi ed answe has ollowed.
Measu emen poin -1
Measu emen poin -2
Measu emen poin -3
Fig 11. Di e en measu emen poin s o min.p essu e-expe imen al and
neu al ne wo k esul s(Del a-ba -del a algo i hm)
Del a-ba -del a algo i hm wi h neu al ne wo k has been
able o ack he desi ed alues. 1. and 3. despi e he low
equency alue a he poin o measu emen , neu al ne wo k
did no ge he desi ed esponse. ANN wi h his algo i hm’ s
accu acy pe cen age o 2% due o emained.
B. Max. and min. p essu e in he simula ion esul s
(Mechanism 2)
Max. and min. p essu e p opaga ion algo i hm back
p opaga ion and Del a - ba - Del a he esul s o he
simula ion igu e 12,13, 14 and 15 a e gi en.
Measu emen poin -1
Measu emen poin -2
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Measu emen poin -3
Fig 12. Di e en measu emen poin s max .expe imen al esul s o p essu e
and a i icial neu al ne wo k(back p opaga ion algo i hm)
In he abo e cha max. back p opaga ion algo i hm esul s
in p essu e a e gi en. 1. wi h he inc ease in p essu e be ween
measu emen poin s change in ib a ion da ahas been
obse ed. ANN, his has la gely ollowed he desi ed da a a
he poin o measu emen . 2. and 3. e y la ge di e ences
we e obse ed in he measu emen poin s. ANN, his
algo i hm is 80% and achie ed an accu acy a e o .
Measu emen poin -1
Measu emen poin -2
Measu emen poin -3
Fig 13. Di e en measu emen poin s max.p essu e-expe imen al and neu al
ne wo k esul s(Del a-ba -del a algo i hm)
Del a-ba -del a algo i hm, neu al ne wo k he ou pu da a
o he desi ed alues a h ee measu emen poin s has been
able o ollow he e ec o ib a ion.
Measu emen poin -1
Measu emen poin -2
0
0,5
1
1,5
2
2,5
110 19 28 37 46 55 64 73 82
İ me [m/s2]
zaman[s]
ysa deneysel
Recen Inno a ions in Mecha onics (RIiM) Vol. 3. (2016). No. 1-2.
DOI: 10.17667/ iim.2016.1-2/14.
Measu emen poin -3
Fig 14. Di e en measu emen poin s min..p essu e-expe imen al and neu al
ne wo k esul s (Backp opaga ion algo i hm )
In he abo e g aph, p essu e, esul s o back p opaga ion
algo i hm a e gi en. ANN, a he poin s whe e he ib a ion is
able o ack sudden changes o he desi ed alues. Apa om
ha ANN’ s accu acy he pe cen age o he algo i hm o 50%.
Whe e 2 alues a e lowe ib a ion. ela i ely good esul s
we e ob ained a he poin o measu emen .
Measu emen poin -1
Measu emen poin -2
Measu emen poin -3
Fig 15. Di e en measu emen poin s o min.p essu e-expe imen al and
neu al ne wo k esul s(Del a-ba -del a algo i hm)
As seen in Figu e 15, he la ges alue o 1 ib a ion we e
ob ained a he measu emen poin . ANN, del a-ba -del a
algo i hm did no gi e he desi ed esul . ANN’ s displayed a
s able image ou pu da a. ANN’ s he accu acy a e is almos
negligible.
VI. CONCLUSIONS
In his s udy, wo di e en es s a ions, PLC con olled
pneuma ic sepa a o and he sepa a ion o es se s om
di e en poin s o accele a ion and ib a ion analysis using
neu al ne wo k modeling wi h noise e ec s was ca ied ou .
As a esul , he ib a ion occu ing a di e en wo king
p essu es and p essu e changes in pneuma ic sys ems and
e iewed he e ec o ib a ion was in es iga ed.
Expe imen al and i e di e en algo i hm by using
simula ion esul s ob ained a e e alua ed, he as p opaga ion
algo i hm and in he back p opaga ion algo i hm simula ion
esul s ob ained o he same beha io expe imen al esul s
ibu e capabili y, he algo i hms he ib a ion pa ame e s can
be used in eal- ime.
    
REFERENCES
[1] Smaoui, M., B un, X., Thomasse , D., 2006. A s udy on acking
posi ion con ol o an elec opneuma ics sys em using backs epping
design. Con ol Enginee ing P ac ice, 14: 923-933.
[2] Akkaya, A. V., Se ilgen, S. H., E dem, H. H., Çe in, B., 2005. Simulink
kullana ak bi pnöma ik sis emin simülasyonu. Doğuş Üni e si esi
De gisi, 6 (2): 155-162.
[3] Moon, J., Lee, B., 2010. Modelling and sensi i i y anaysis o a
pneuma ic ib a ion isola ion sys em wi h wo ai chambe s. Mechanism
and Machine Theo y, 45: 1828-1850.
[4] Pu, H., Luo, X., Chen, X., 2011. Modelling and analysis o dual-
chambe pneuma ic sp ing wi h adjus ible damping o p ecision
ib a ion isola ion. Jou nal o Sound and Vib a ion, 330: 3578-3590.
[5] Ka penko, M., Sepeh i, N., 2002. Neu al ne wo k classi ie s applied o
condi ion moni o ing o a pneuma ic p ocess al e ac ua o . Enginee ing
Applica ions o a i icial In elligence, 15: 273-283.
[6] So li, M., Gas aldi, L., Codina, E., He as, S., 1999. Dynamic analysis o
pneuma ic ac ua o s. Simula ion P ac ice and Theo y, 7: 589-602.
[7] Sağı oğlu, Ş., Beşdok, E., E le , M., 2003. Mühendislik e Yapay Zeka
Uygulamala ı-1. U uk Yayıncılık, Kayse i, 426 s.
[8] MEGEP, 2007. Temel Pnöma ik. Milli Eği im Bakanlığı, Anka a, 75 s.