scieee Science in your language
[en] (orig)

Novel speed and current sensor FDI schemes with an improved AFTC for induction motor drives

Abstract

This paper focuses on speed and current sensor Faults Detection and Isolation (FDI) in an Induction Motor (IM) drive. The effect of sensors faults on the IM vector control is presented, then, new detection and isolation approaches are suggested. Speed sensor faults are detected when an error between only two points from speed data exceeds a certain threshold. An algorithm based on RMS currents is developed to detect and isolate any faulty current sensor. This requires three current sensors, each per phase. Besides, open circuit faults of inverter power switches are taken into account too. To ensure continuous functionality of the drive, we conceived an Active Fault Tolerant Controller (AFTC) with smoother reconfiguration feature. Simulations in Matlab/Simulink are carried out to show the efficiency of the suggested schemes.

Read accessible full text

Novel speed and current sensor FDI schemes with an improved AFTC for induction motor drives

Author: Bouakoura, Mohamed
Publisher: Vysoká škola báňská - Technická univerzita Ostrava
Year: 2018
DOI: 10.15598/aeee.v16i1.2573
Source: https://dspace.vsb.cz/bitstreams/bde955f7-1840-4495-b7ff-6de95303b07c/download
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
No el Speed and Cu en Senso FDI Schemes wi h
an Imp o ed AFTC o Induc ion Mo o D i es
Mohamed BOUAKOURA, Nas eddine NAIT-SAID,
Mohamed-Said NAIT-SAID, Adel BELBACH
LSP-IE’2000 Labo a o y, Elec ical Enginee ing Depa men , Facul y o Technology, Uni e si y o Ba na 2,
Rue Chahid Mohamed El-Hadi Boukhlou , 05000 Ba na, Alge ia
bouak[email p o ec ed], n_nai [email protected], medsnai said@yahoo. , adel.belbac[email p o ec ed]
DOI: 10.15598/aeee. 16i1.2573
Abs ac . This pape ocuses on speed and cu en sen-
so Faul s De ec ion and Isola ion (FDI) in an Induc-
ion Mo o (IM) d i e. The e ec o senso s aul s
on he IM ec o con ol is p esen ed, hen, new de ec-
ion and isola ion app oaches a e sugges ed. Speed sen-
so aul s a e de ec ed when an e o be ween only wo
poin s om speed da a exceeds a ce ain h eshold. An
algo i hm based on RMS cu en s is de eloped o de ec
and isola e any aul y cu en senso . This equi es
h ee cu en senso s, each pe phase. Besides, open
ci cui aul s o in e e powe swi ches a e aken in o
accoun oo. To ensu e con inuous unc ionali y o he
d i e, we concei ed an Ac i e Faul Tole an Con olle
(AFTC) wi h smoo he econ igu a ion ea u e. Simu-
la ions in Ma lab/Simulink a e ca ied ou o show he
e iciency o he sugges ed schemes.
Keywo ds
Ac i e aul ole an con ol, cu en senso , in-
e e powe swi ch, MRAS, RMS alue, speed
senso .
1. In oduc ion
Faul Tole an Con ol (FTC) is a echnique imple-
men ed in many c i ical and high a ailabili y sys ems.
I s main pu pose is o mi iga e aul s and ensu e a
con inuous unc ionali y o a sys em wi h aul y ele-
men s a he han o al ailu e. The i s FTC was
implemen ed in ai c a s [1]. A e ha , i has been
b oadened o many o he ields, such as powe plan s
[2], anspo a ion [3], [4] and [5], and wind ene gy con-
e sion sys ems [6], [7] and [8]. To con ol any p ocess,
accu a e eedback in o ma ion is equi ed, and his one
is p o ided by senso s. Thus, in his pape , we in es-
iga e pa icula ly speed and cu en senso aul s in
an induc ion mo o d i e. The choice was aken since
induc ion mo o d i es a e in ol ed in mos p opulsion
and ac ion applica ions [9].
Faul ole an echniques a e di ided in o wo ypes:
passi e FTCs and ac i e FTCs. The i s ones in ol e
obus con olle s such as H_∞[10] and [11] and slid-
ing mode con olle s [12], [13] and [14], i.e. he aul y
elemen emains in eg a ed in he sys em whe e he
con olle abso bs i s e ec . Ne e heless, his ech-
nique has a limi ed e ec i eness because i ole a es
only low se e i y aul s [15] and [16]. Ac i e FTCs
(AFTC), o econ igu able FTCs, a e mo e sui able o
se e e aul s since he aul y componen is eplaced au-
oma ically by a heal hy one o i s signal is gene a ed
by a ma hema ical model based on o he a ailable sen-
so s. AFTC necessi a es a Faul De ec ion and Isola-
ion (FDI) mechanism.
In some pape s, ha dwa e edundancy is conside ed
o aul de ec ion [17] and [18], whe eas in o he s an-
aly ical edundancy is p e e ed. This las ely on es i-
ma o s and obse e s, o example; in [19] an ex ended
Kalman il e is conside ed as a speed i ual senso .
Alkaya and Eke applied a Luenbe ge obse e wi h a
DC mo o o de ec speed senso aul s [20]. MRAS is
also a widely used me hod o speed es ima ion. I is
p esen ed by Wang e al. as a subs i u e o he aul y
speed senso [21]. Usually, when i is ha d o model
a p ocess, signal p ocessing and machine lea ning ap-
p oaches a e e ec i e. In [22], wa ele analysis has a
undamen al ole in FTC scheme. In [23] he au ho s
conside ed he s a o cu en signa u e as a eliable
ool o de ec eccen ici y aul s o induc ion mo o s.
Fuzzy logic was an e icien ool used by Kamal e al.
o es ima e senso aul s in a wind-diesel hyb id sys em
[24] and also in [25] o de elop a mo e e icien con ol
o an induc ion mo o .
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 1
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
Many esea che s deal wi h senso s aul s. In [7],
[26] and [27] senso s mal unc ion causes signi ican loss
o he con olle pe o mance. Hence, his issue was a
mo i a ion o FTC design. This pape co e s se e al
con ibu ions o speed and cu en senso aul s diag-
nosis and ole ance in induc ion mo o d i es. Fi s ,
he e ec o senso s aul s on he IM ec o con ol
(IFOC) is p esen ed. Then, we p opose new de ec-
ion and isola ion s a egies based on signal p ocessing.
Las men ioned a e e ec i e and easy o implemen . To
handle hese aul s, an imp o ed AFTC scheme is de-
eloped wi h smoo he econ igu a ion ea u e a sen-
so aul momen .
This pape is o ganized as ollows: Sec ion 2. is
dedica ed o he de ec ion and isola ion o speed and
cu en senso aul s. Sec ion 3. p esen s an ex-
ension o cu en senso FDI algo i hm o de ec and
isola e in e e leg open swi ches aul s. The main im-
p o emen s on he AFTC a e explained in Sec. 4.
Sec ion 5. concludes he pape .
Rema k 1. No e ha in Fig. 2, Fig. 5, Fig. 6, Fig. 8,
Fig. 11 and Fig. 12 he ec angle wi h ounded angles
ep esen s a condi ional es . I s ou pu is bina y, so
i equals “1” i he condi ion is e i ied, o he wise i
equals “0”.
2. Speed and Cu en Senso
Faul Diagnosis
Be o e we p esen he FDI algo i hms, we show he
in luence o speed and cu en senso aul s on he pe -
o mance o he IM d i e wi h indi ec ield-o ien ed
con ol. We chose i since i is one o he mos pe o -
man and widesp ead echnique.
The in e mi en aul in speed senso s o DC gen-
e a o ype o al e na o s is usually caused by o o
eccen ici y [28] and [29] o he a i ion o b ushes o
bea ings. Whe eas o se aul s may be caused by he
a ia ion o elec ical pa ame e s o he senso in some
ope a ing condi ions. In o a y encode s, an insu i-
cien ligh sou ce (LED) o a mal unc ion o he pho-
o ansis o p oduces unce ain measu emen [8] and
[29]. The mechanical sliding in bo h ypes o speed
senso s (encode , gene a o ) causes ab up changes in
measu emen .
Since IFOC con ols mo o cu en s, he las men-
ioned a e usually measu ed by a leas wo cu en
senso s, bu some imes hey a e es ima ed om speed
and DC bus ol age. To inc ease he eliabili y o he
d i e we p e e o measu e phase cu en s ins ead o
es ima ing hem elying on o he senso s.
Cu en senso aul s a e less se e e han hose o
speed senso , ye , hey al e he con olle pe o -
mance. The causes o cu en senso mal unc ion a e
ela ed o i s physical s uc u e. In some unc ioning
condi ions, he change in ma e ial p ope ies and also
he deg ada ion a e a long ime o use p oduces sen-
so aul s. Cu en senso s based on Hall e ec a e no
linea wi h espec o magne ic lux densi y, so hey
may be sa u a ed i he measu ed cu en exceeds he
nominal suppo ed alue, which engende s a bias in
measu emen [30] and [31]. A disconnec ion o he elec-
ical link o b eakdown o he senso is an o igin o he
o al loss o eedback in o ma ion.
Rema k 2. In all simula ions in his sec ion, a ed
load oque Tl = 6.1[Nm] is applied a = 1 [s] and
each aul is ac i a ed a = 1.5[s].
2.1. Speed Senso Faul s E ec on
IFOC
In his pape , he in es iga ed aul s a e: in e mi en
aul , o se aul , and o al loss aul . Each one is
pe o med in Ma lab as ollows:





In e mi en aul : Ω = Ω + δ,
O se aul : Ω = Ω + γ,
To al loss: Ω = Ω ×0,
(1)
whe e γis he o se alue, γ= 10 ad·s−1, and δis
a andom numbe wi h a mean alue equal o 0 and a
a iance equal o 10 ad·s−1.
Figu e 1 shows phase cu en , o o luxes on “dq”
e e ence ame, and speed o each speed senso aul .
Speed senso aul s ha e a clea impac on he ec-
o con ol. The in e mi en aul causes signi i-
can o que ipples which cause he changes in speed
(Fig. 1(a)). Rega ding o se aul (Fig. 1(b)), i is
ea ed almos as a load o que by he ec o con ol.
So he phase cu en s ise ins an ly. Howe e , he e -
ec o he aul on he ac ual speed is no elimina ed
since he o se alue added o he ac ual speed makes
he speed alue p o ided by he aul y senso equal o
he e e ence. To al loss aul is he mos isky be-
cause he speed becomes no longe con olled. Thus,
we limi ed he s a o elec ic speed “ωs” and he cu -
en “Isq” o p e en speed di e gence. S a o cu en s
ise o wo imes he a ed cu en a he aul momen
and o o luxes do no ollow hei e e ences a e he
aul (Fig. 1(c)). The To que ipple a e inc eases o
Temax −min ≈5.8Nm.
2.2. Speed Senso Faul s De ec ion
Mos o en, he speed a ia ion due o a senso aul is
as e han i s a ia ion due o a o que load, change
in speed e e ence o aul s in o he componen s o he
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 2
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
-5
0
5
Ia [A]
0
0.5
1
Ro o Flux
[Wb]
5
6
7
To que
[Nm]
1.4 1.42 1.44 1.46 1.48 1.5 1.52 1.54 1.56 1.58 1.6
Time [s]
99.5
100
100.5
Speed
[Rad·s-1]
Fi dFi q
(a) In e mi en aul .
-10
0
10
Ia [A]
-2
0
2
Ro o Flux
[Wb]
-5
0
5
10
To que
[Nm]
1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Time [s]
80
100
120
Speed
[Rad·s-1]
Fi d
Fi q
(b) O se aul .
-10
0
10
Ia [A]
-2
0
2
Ro o Flux
[Wb]
-5
0
5
10
To que
[Nm]
1.3 1.4 1.5 1.6 1.7 1.8 1.9
2
Time [s]
80
100
120
Speed
[Rad·s-1]
Fi d
Fi q
(c) To al loss aul .
Fig. 1: Phase cu en , o o luxes on “dq” e e ence ame, and
ac ual speed in p esence o speed senso aul : (a) in e -
mi en aul , (b) o se aul , (c) o al loss aul . Wi h
Ω∗= 100 ad·s−1,Tl = 6.1Nm and φ∗
= 1 Wb. Each
aul is applied a = 1.5s.
d i e [32]. F om his s andpoin , he de ec ion could be
achie ed by compa ing only wo poin s om speed da a
be ween which he dis ance is p opo ional o he sam-
pling ime. In ou case, since T= 5 ·10−6s , i e s eps
dis ance is su icien ; τd= 5 ×T. The de ec ion sig-
nal is compu ed as illus a ed by he scheme in Fig. 2.
This p oposed speed senso aul s de ec ion scheme is
less un ime consuming compa ed o obse e based
app oaches o some simple signal p ocessing echniques
such as a e age s anda d de ia ion in [32].
Ω(𝑡 −𝜏𝑑)
Ω(𝑡)
− +
+ −
Th eshold
𝑖𝑓 𝜀1>0 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑖𝑚𝑝𝑢𝑙𝑠𝑒
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝜀1
𝜀2
∫
dsw
dw
|𝑥(𝑡)|
𝑖𝑓 𝜀2>0 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 = 0
Fig2
Fig. 2: De ec ion scheme o speed senso aul s. Wi h τd=
5×T.
Figu e 3 shows he simula ion esul s o he sug-
ges ed de ec ion me hod wi h h ee di e en aul s. I
is clea ha any ab up change in speed da a gene a es
impulses in “dω” cu e.
Hence, any impulses due o measu emen noise, load
o que o ansien s in speed a e kep unde a p ese
h eshold. A aul occu ence momen , “dω” exceeds
he h eshold gene a ing a de ec ion signal “dsω”. The
di e ence be ween “dω” and he h eshold is ep e-
sen ed by “1”. When i s alue is g ea e han ze o,
we ge an impulse which is hen in eg a ed o ge a
cons an signal “2”. Since his alue is e y small, i is
ans o med ia a elay o p oduce a meaning ul bina y
signal.
2.3. Cu en Senso Faul s E ec on
IFOC/Senso less IFOC
Th ee di e en aul s a e conside ed; o se aul , gain
aul , and o al loss o eedback in o ma ion. They a e
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 3
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
0
10
20
30
40
dw
0 0.5 1 1.5 2 2.5 3
Time [s]
-0.5
0
0.5
1
1.5
dsw
dw
Th eshold
Speed senso aul
(a) In e mi en aul .
0
5
10
15
20
dw
0 0.5 1 1.5 2 2.5 3
Time [s]
-0.5
0
0.5
1
1.5
dsw
dw
Th eshold
Speed senso aul
(b) O se aul .
0
50
100
150
dw
0 0.5 1 1.5 2 2.5 3
Time [s]
-0.5
0
0.5
1
1.5
dsw
Th eshold
Speed senso aul
dw
(c) To al loss aul .
Fig. 3: De ec ion esul s o speed senso aul s: (a) in e mi en
aul , (b) o se aul , (c) o al loss aul .
simula ed in Ma lab as ollows:





O se aul : Ic =Ic+ρ,
Gain aul : Ic =Ic×$,
To al loss: Ic =Ic×0,
(2)
whe e ρis he o se alue (ρ= 2A) and $is he gain
coe icien ($= 0.5).
The o se alue and he gain a e chosen ela i ely
small, so hey will no ha e a conside able impac on
elec ical and mechanical a iables, ye , hey should
be de ec ed by an algo i hm (so wa e) using he mea-
su emen s o o he non- aul y senso s. In he case o
IFOC wi h speed encode , cu en senso aul s cause
ei he luc ua ion o o o luxes (o se aul ) o de-
ia e hem om hei e e ences (gain and o al loss
aul ). Consequen ly, conside able o que oscilla ions
a e no iced wi h Temax −min ≈5Nm, which may lead
o long- e m o a mechanical de e io a ion o he sha .
In senso less ope a ion, MRAS speed es ima o loses
i s e iciency when cu en senso s p o ide inaccu a e
alues. Hence, his leads o a o al con olle ailu e.
Figu e 4 illus a es: “Ib” phase cu en , o o luxes
on “dq” e e ence ame, o que, and speed. We chose
o simula e he aul o only one cu en senso since
he occu ence p obabili y o wo o h ee aul s in a
sho ime pe iod is e y low. Bo h ope a ions a e
conside ed: wi h and wi hou a speed senso .
2.4. Faul y Cu en Senso
De ec ion and Isola ion
The simples way o de ec a cu en senso aul in
a balanced h ee-phase sys em is he sum o he h ee
cu en s. This sum is p ac ically null in no mal ope a-
ion o he d i e, ye i changes due o a cu en senso
aul . We adop ed he absolu e mean alue o h ee
cu en s sum “I ” as a aul indica o “dsi” when i ex-
ceeds ce ain h eshold “ζ” (Fig. 5). As o localiza ion,
a new algo i hm is p oposed based on RMS alues o
phase cu en s. The use o RMS alues pe mi s he
localiza ion o a cu en senso unde gain aul , un-
like a e age alues which a e null when cu en s s ill
al e na ing a e he aul . Hence, he e iciency o he
echnique p oposed in [7] is no e i ied wi h gain aul .
Mo eo e , he de eloped me hod in his pape is less
compu a ionally demanding han he one in [7].
The key idea o localiza ion is o look o he mini-
mum alue be ween wo RMS alues o phase cu en s.
Because his alue co esponds o he di e ence be-
ween he RMS cu en s measu ed by heal hy senso s,
hen he emaining phase cu en is measu ed by a
aul y one. F om Fig. 6, i he senso o phase “b” is
aul y, bwill be equal o “0” since i is he di e ence
be ween he alue chosen by he unc ion minimum and
he di e ence |Ia ms −Ib ms |. When b= 0, hen he
condi ional es is e i ied and he middle ou pu will
be equal o “1”. This la e is mul iplied by 2 which
is he index o phase “b”. Since aul y senso index is
no cons an be o e he aul occu ence we mul iply
i by “dsi” o a oid any alse localiza ion signal. The
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 4
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
-5
0
5
Ib [A]
-0.5
0
0.5
1
1.5
Ro o Flux
[Wb]
5
10
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
95
100
105
Speed
[ ad·s-1]
Fi d
Fi q
(a1)
To que
[Nm]
-20
0
20
Ib [A]
-0.5
0
0.5
1
1.5
Ro o
Flux [Wb]
-50
0
50
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
100
200
Speed
[ ad·s-1]
Fi dFi q
(a2)
To que
[Nm]
(a) O se aul .
-5
0
5
Ib [A]
-0.5
0
0.5
1
1.5
Ro o Flux
[Wb]
5
10
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
95
100
105
Speed
[ ad·s-1]
Fi d
Fi q
(a1)
To que
[Nm]
-20
0
20
Ib [A]
-0.5
0
0.5
1
1.5
Ro o
Flux [Wb]
-50
0
50
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
100
200
Speed
[ ad·s-1]
Fi dFi q
(a2)
To que
[Nm]
(b) O se aul wi h MRAS.
-5
0
5
Ib [A]
-0.5
0
0.5
1
1.5
Ro o Flux
[Wb]
5
10
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
95
100
105
Speed
[ ad·s-1]
Fi qFi d
(b1)
To que
[Nm]
-10
0
10
Ib [A]
-1
0
1
2
Ro o
Flux [Wb]
-50
0
50
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
50
100
150
Speed
[ ad·s-1]
Fi q
Fi d
(b2)
To que
[Nm]
(c) Gain aul .
-5
0
5
Ib [A]
-0.5
0
0.5
1
1.5
Ro o Flux
[Wb]
5
10
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
95
100
105
Speed
[ ad·s-1]
Fi qFi d
(b1)
To que
[Nm]
-10
0
10
Ib [A]
-1
0
1
2
Ro o
Flux [Wb]
-50
0
50
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
50
100
150
Speed
[ ad·s-1]
Fi q
Fi d
(b2)
To que
[Nm]
(d) Gain aul wi h MRAS.
-10
0
10
Ib [A]
-1
0
1
2
3
Ro o Flux
[Wb]
-20
0
20
To que
[Nm]
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
90
100
110
Speed
[ ad·s-1]
Fi d
Fi q
(c1)
-10
0
10
Ib [A]
-1
0
1
2
3
-20
0
20
To que
[Nm]
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
50
100
150
Speed
[ ad·s-1]
Fi d
Fi q
(c2)
Ro o Flux
[Wb]
(e) To al loss aul .
-10
0
10
Ib [A]
-1
0
1
2
3
Ro o Flux
[Wb]
-20
0
20
To que
[Nm]
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
90
100
110
Speed
[ ad·s-1]
Fi d
Fi q
(c1)
-10
0
10
Ib [A]
-1
0
1
2
3
-20
0
20
To que
[Nm]
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
50
100
150
Speed
[ ad·s-1]
Fi d
Fi q
(c2)
Ro o Flux
[Wb]
( ) To al loss aul wi h MRAS.
Fig. 4: Phase cu en , o o luxes on “dq” e e ence ame, o que and ac ual speed in p esence o cu en senso aul o phase
“b” wi h speed senso and wi h MRAS. Ω∗= 100 ad·s−1,Tl = 6.1Nm and φ∗
= 1 Wb. Each aul is applied a = 1.5s.
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 5

POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
+ −
+ −
+ −
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
Minimum
+ −
+ −
− +
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
𝑟𝑐
𝑟𝑏
𝑟𝑎
×3
×2
×1
+ + +
×
dsi
Faul y senso index
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
Minimum
+ −
− +
− +
𝑧𝑎
𝑧𝑏
𝑧𝑐
×3
×2
×1
+ + +
×
NAND
NOT dsi
𝑑Ω∗
𝑑𝑡
×𝜆𝒕𝒔
+ +
𝜆𝒔𝒔
− +
𝑦
×
Faul y in e e leg index
LPFLPFLPF
LPF LPF LPF
LPF LPF LPF
+ + +
Faul y
in e e leg
Localiza ion
RMS
dsi
NOT
𝐼𝑓
Ω∗
Faul y
cu en
senso
localiza ion
|𝑥(𝑡)|
mean
alue
𝐼𝑎𝑏𝑐
dsps
lsps
𝑑𝑠𝑖
𝑦0
Faul y senso
numbe
Faul y
in e e leg
numbe
𝑖𝑓 𝐼𝑓≥ 𝜁 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 = 1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡=0
𝑖𝑓 𝑟𝑗= 0 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝜌𝑚𝑖𝑛 <𝑧𝑗< 𝜌𝑚𝑎𝑥 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝑑Ω∗
𝑑𝑡 ≥1⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑖𝑓 𝑦>0⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
Fig5
Fig6
Fig8
Fig. 5: Global block scheme o cu en senso /in e e leg aul
de ec ion and isola ion.
p oposed algo i hms a e e i ied by simula ion and he
esul s a e shown in Fig. 7. Fo all h ee conside ed
aul s, he gap be ween he RMS alues co esponding
o heal hy senso s is he smalles . Faul y senso lo-
caliza ion block is in en ionally ac i a ed a e 0.035 s
om de ec ion momen , which is he ime equi ed o
ge a cons an localiza ion signal.
3. Faul y Powe Swi ch
De ec ion and Localiza ion
As an ex ension o senso aul diagnosis, we added a
block o iden i y he in e e leg wi h a aul y powe
swi ch. Sho ci cui aul s o powe swi ches canno
be localized as enough since he cu en o he DC
ol age sou ce inc eases in milliseconds o a high alue
which igge s he p o ec ion componen s ( use o e-
lay) o shu down he d i e. Only open ci cui aul s o
con olled powe swi ches a e conside ed because hey
a e mo e p one o aul s han an ipa allel diodes. One-
hal cycle o phase cu en passes due o he loss o an
in e e powe swi ch. Consequen ly, he RMS cu -
en in aul y in e e leg dec eases. Figu e 8 shows
he algo i hm o de ec and localize he aul y in e e
swi ch. Fo example, i op swi ch o he second leg is
open-ci cui ed, he RMS cu en o he second phase
"Ib ms " will ha e he lowes alue, and he wo o he
RMS cu en s "Ia ms " and "Ic ms " will ise o compen-
sa e he cu en d op. As illus a ed in Fig. 8, a e he
aul , zbs ays in he de ined in e al: ρmin < zb< ρmax
howe e : za,zc a y such as: ρmax < za,ρmax < zc.
The di e ence be ween he ou pu o he unc ion min-
imum and Ij ms is no ed as zj. Wi h jis he phase
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
+ −
+ −
+ −
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
Minimum
+ −
+ −
− +
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
𝑟𝑐
𝑟𝑏
𝑟𝑎
×3
×2
×1
+ + +
×
dsi
Faul y senso index
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
Minimum
+ −
− +
− +
𝑧𝑎
𝑧𝑏
𝑧𝑐
×3
×2
×1
+ + +
×
NAND
NOT dsi
𝑑Ω∗
𝑑𝑡
×𝜆𝒕𝒔
+ +
𝜆𝒔𝒔
− +
𝑦
×
Faul y in e e leg index
LPFLPFLPF
LPF LPF LPF
LPF LPF LPF
+ + +
Faul y
in e e leg
Localiza ion
RMS
dsi
NOT
𝐼𝑓
Ω∗
Faul y
cu en
senso
localiza ion
|𝑥(𝑡)|
mean
alue
𝐼𝑎𝑏𝑐
dsps
lsps
𝑑𝑠𝑖
𝑦0
Faul y senso
numbe
Faul y
in e e leg
numbe
𝑖𝑓 𝐼𝑓≥𝜁 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡=0
𝑖𝑓 𝑟𝑗= 0 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 = 1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝜌𝑚𝑖𝑛 <𝑧𝑗< 𝜌𝑚𝑎𝑥 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝑑Ω∗
𝑑𝑡 ≥1⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑖𝑓 𝑦>0⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
Fig5
Fig6
Fig8
Fig. 6: Isola ion scheme o aul y cu en senso .
index and j=ao bo c. The in e al limi s: ρmin,
ρmax a e chosen close o ze o, whe e ρmin =−0.04 and
ρmax = 0.04. The condi ional es ou pu equals one
i zjis included in he in e al [ρmin,ρmax] o he wise
i equals ze o. In he conside ed case, he ou pu o
he condi ional es block is [1 1 1] be o e he aul
and [0 1 0] a e he aul occu ence. When hese
bina y alues pass by he NAND unc ion, hey gene -
a e a de ec ion signal “dsps” which is null be o e he
aul -in s eady s a e- and equal o 1 a e i . Since
he an ipa allel diode allows he con inui y o cu en ,
an open swi ch aul a ec s sligh ly he sum o h ee
cu en s. This is used o di e en ia e be ween senso
and powe swi ches aul s. We mul iply by he in e se
o he senso aul de ec ion signal “(dsi)” o u n o
he localiza ion block o powe swi ches aul s when a
cu en senso aul occu s. Also, o elimina e any alse
ala m due o di e en changes in speed e e ence, we
se an adap i e h eshold in unc ion o speed e e ence,
whe e he λss and λ s a e he s eady and ansien s a e
h esholds espec i ely. λ s is ac i a ed only i he e -
e ence speed changes and i s de i a i e is supe io o
one. The localiza ion o he uppe swi ch aul in he
second in e e leg is simula ed in Ma lab and he e-
sul is illus a ed in Fig. 9. No ice ha in a ansien
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 6
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
2
3
4
IabcRMS
[A]
IaRMS IbRMS IcRMS
-10
0
10
Cu en s
Sum
0
1
2
absolu e
mean o
he sum
0
0.5
1
De ec ion
Signal
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
0
1
2
Faul y
Senso
Numbe
I
ζ
(a) O se aul .
1
2
3
4
IabcRMS
[A]
IaRMS IbRMS IcRMS
-10
0
10
Cu en s
Sum
0.5
1
1.5
Absolu e
mean o
he sum
0
0.5
1
De ec ion
Signal
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
0
1
2
Faul y
Senso
Numbe
I
ζ
(b) Gain aul .
0
2
4
6
8
IabcRMS
[A]
IaRMS IbRMS IcRMS
-20
0
20
Cu en s
Sum
0
5
Absolu e
mean o
he sum
0
0.5
1
De ec ion
Signal
1.2 1.3 1.4 1.5 1.6 1.7 1.8
Time [s]
0
1
2
Faul y
Senso
Numbe
I
ζ
(c) To al loss aul .
Fig. 7: De ec ion and isola ion esul s o cu en senso aul s:
(a) o se aul , (b) gain aul , (c) o al loss aul .
s a e (0< < 0.25) “dsps” is no null, which could
p oduce a alse ala m i we did no use an adap i e
h eshold. The p e-localiza ion signal “lsps” is ac i-
a ed by “dsps” because i has no meaning un il he
aul occu ence, whe e i s alue indica es he aul y
leg.
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
+ −
+ −
+ −
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
Minimum
+ −
+ −
− +
|𝑥(𝑡)|
|𝑥(𝑡)|
|𝑥(𝑡)|
𝑟𝑐
𝑟𝑏
𝑟𝑎
×3
×2
×1
+ + +
×
dsi
Faul y senso index
𝐼𝑏𝑟𝑚𝑠
𝐼𝑎𝑟𝑚𝑠
𝐼𝑐𝑟𝑚𝑠
Minimum
+ −
− +
− +
𝑧𝑎
𝑧𝑏
𝑧𝑐
×3
×2
×1
+ + +
×
NAND
NOT dsi
𝑑Ω∗
𝑑𝑡
×𝜆𝒕𝒔
+ +
𝜆𝒔𝒔
− +
𝑦
×
Faul y in e e leg index
LPFLPFLPF
LPF LPF LPF
LPF LPF LPF
+ + +
Faul y
in e e leg
Localiza ion
RMS
dsi
NOT
𝐼𝑓
Ω∗
Faul y
cu en
senso
localiza ion
|𝑥(𝑡)|
mean
alue
𝐼𝑎𝑏𝑐
dsps
lsps
𝑑𝑠𝑖
𝑦0
Faul y senso
numbe
Faul y
in e e leg
numbe
𝑖𝑓 𝐼𝑓≥ 𝜁 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 = 1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡=0
𝑖𝑓 𝑟𝑗= 0 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 = 1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝜌𝑚𝑖𝑛 <𝑧𝑗< 𝜌𝑚𝑎𝑥 ⇒𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑤𝑖𝑡ℎ 𝑗=𝑎 𝑜𝑟 𝑏 𝑜𝑟 𝑐
𝑖𝑓 𝑑Ω∗
𝑑𝑡 ≥1⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
𝑖𝑓 𝑦>0⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =1
𝑒𝑙𝑠𝑒 ⇒ 𝑜𝑢𝑡𝑝𝑢𝑡 =0
Fig5
Fig6
Fig8
Fig. 8: Faul y leg isola ion algo i hm.
1.4 1.45 1.5 1.55 1.6 1.65 1.7 1.75 1.8
2
4
6
IabcRMS
[A]
IaRMS IbRMS IcRMS
1.45 1.5 1.55 1.6 1.65 1.7
0
0.2
0.4
z1,z2,z3
& h esholds
z1 z2 z3 ho1 ho2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
5
10
lsps,
dsps,
NOT(dsi)
lsps dsps NOT(dsi)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
0
2
4
6
y0
y0 Adap ed h eshold
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time [s]
0
1
2
Faul y leg
Index
Fig. 9: Faul y leg isola ion esul s.
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 7
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
Since no cu en senso aul occu ed, “(dsi)” is con-
s an ly equal o “1”. The de ec ion and localiza ion de-
lay is due o he low pass il e s used o educe signals
luc ua ions.
4. Imp o ed AFTC
4.1. O e iew o he AFTC Scheme
Figu e 10 shows he o e all scheme o he IM d i e wi h
an AFTC and a de ec ion mechanism. A speed encode
and h ee cu en senso s a e used o measu emen .
All senso s da a pass by a de ec ion block o de ec any
senso mal unc ion. Re e ence ol ages a e gene a ed
by he ac i e aul ole an con olle .
The AFTC block inco po a es ou di e en con ol
echniques, each one o hem equi es a ce ain mini-
mum o senso s o unc ion p ope ly. The selec ion o
he con ol s a egy is achie ed au oma ically depend-
ing on he ou pu s o he senso s aul s de ec o . In
Tab. 1, we summa ize he possible de ec o ou pu s,
emaining heal hy senso s, and he chosen con olle .
1
Ω∗
DC powe supply
Ac i e Faul
Tole an
Con olle
𝑉
𝑎
∗
𝑉
𝑏
∗
𝑉
𝑐
∗
dsw
dsi
Speed and Cu en
Senso Faul s
De ec o
𝑖𝑎
𝑖𝑏
𝑖𝑐
Ω
Faul
Faul
Fig. 10: O e all scheme o he IM d i e wi h AFTC.
Tab. 1: De ec ion signals and he selec ed con ol echnique.
dsi dsw
Remaining
heal hy
senso s
Con olle numbe 1:
IFOC 0 0
Speed senso ,
h ee cu en
senso s
Con olle numbe 2:
IFOC senso less 0 1
Th ee cu en
senso s, DC
ol age
measu emen
Con olle numbe 3:
V/ CL 1 0 Speed senso
Con olle numbe 4:
V/ OL 1 1 No senso s
a ailable
4.2. Smoo hening he T ansi ion o
V/ CL Con ol
In AFTC, a e de ec ing a senso aul , a econ igu a-
ion o he con ol scheme is necessa y, bu his s ep is
no always s aigh o wa d. As shown in Fig. 11, he e
is a phase shi be ween he e e ence ol age o IFOC
and V/ CL. Thus, he ansi ion be ween hese wo
con ol echniques p oduces signi ican o que dis o -
ion and conside able speed luc ua ion due o a decel-
e a ion o he o a ing magne ic ield.
Fig. 11. Re e ence ol age o IFOC and V/
CL on “𝛼𝛽” e e ence ame
𝑉
𝑠𝛽
𝑉
𝑠
∗
ሬ
ሬ
ሬ
ሬ
Ԧ
𝐼𝐹𝑂𝐶
𝑉
𝑠
∗
ሬ
ሬ
ሬ
ሬ
Ԧ
𝑉
𝑓 𝐶𝐿
Di ec ion o
Ro a ion
𝜑
𝑉
𝑠𝛼
𝜃
𝑉
𝑓𝐶𝐿
𝜃
𝐼𝐹𝑂𝐶
Fig. 11: Re e ence ol age o IFOC and V/ CL on “αβ” e e -
ence ame.
Se e al esea che s sugges ed some ideas o educe
he e ec o he ansi ion, o example, Diallo e
al. ecommend ha he swi ching mus be pe o med
when he phase shi be ween he e e ence ol ages o
ec o con ol and scala con ol is almos ze o [33].
Howe e , in hei pape , he selec ion o he sui able
swi ching momen is no done au oma ically by he
con olle bu p og ammed by he au ho s. In [34],
he au ho s educed he phase shi be ween he con-
olle s by eadjus ing he PI pa ame e s o V/ CL.
In hese pape s, he p ocess om de ec ion o ansi-
ion is no well cla i ied. In his sec ion, we p esen a
new app oach o smoo hen he ansi ion in he AFTC
by linking e e ence ol ages o wo con olle s whe e
one is ac i e and he o he is on s andby be o e he
ansi ion momen . Then, he con olle in s andby is
libe a ed g adually o ake o e when i is selec ed. Be-
sides, an adap a ion o he e e ence speed is necessa y
o achie e a be e pe o mance.
The new scheme o V/ CL con ol is shown in
Fig. 12. Basically, wo main modi ica ions on V/ CL
con ol a e done o smoo hen he ansi ion o i om
IFOC due o a cu en senso aul :
•The i s modi ica ion consis s o p epa ing he
con olle V/ CL o ake o e by ixing i s e e -
ence ol age on IFOC’s, i.e. he elec ic equen-
cies o he wo con olle s a e equalized. This is
pe o med by he ollowing equa ion:
ω0
sV
CL =ωsV
CL −ωsV
CL −ωsIF OC ·RcCL,
(3)
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 8
POWER ENGINEERING AND ELECTRICAL ENGINEERING VOLUME: 16 |NUMBER: 1 |2018 |MARCH
PI
Vboos
+ -
+ +
V/
𝑽𝒂𝒃𝒄
∗
𝑽𝒎
𝜽𝒔
PWM
In e e
+ +
𝑽𝒂𝒃𝒄
𝜔𝑠𝑉𝑓
+ -
𝜔𝑠𝐼𝐹𝑂𝐶
×
Con olle
Numbe
1s
O de
LPF
2nd
O de
LPF
+ -
∫
𝒑
𝛀∗
+ -
×
𝛀
+ + +
Vboos
+ +
V/
𝑽𝒂𝒃𝒄
∗
𝑽𝒎
𝜽𝒔
PWM
In e e
𝑽𝒂𝒃𝒄
𝜔𝑠𝑉𝑓
𝑂𝐿
+ -
×
1s
O de
LPF
+ +
∫
𝑘𝑑2(𝑡−𝜏𝑑2)
𝛀∗
+ +
𝜔𝑠𝑉𝑓
𝐶𝐿
𝒑
𝛀∗
1s
O de LPF wi h
adap i e ime cons an
𝝉
Adap i e
gain
Ω∗ ′𝑉
𝑓𝐶𝐿
𝜔𝑠𝑉
𝑓𝐶𝐿
′
ℛ𝑐𝐶𝐿
ℛ𝑐𝐶𝐿
ℛ𝑐𝑂𝐿
𝐴ℛ𝑐𝑂𝐿
𝜔𝑠𝑉
𝑓𝑂𝐿
′
Ω∗ ′𝑉
𝑓𝑂𝐿
𝑘𝑑1(𝑡−𝜏𝑑1)
𝑖𝑓 𝑐𝑛≠3 ⇒𝑜𝑢𝑡𝑝𝑢𝑡=1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡= 0
Con olle
Numbe
𝑖𝑓 𝑐𝑛≠4 ⇒𝑜𝑢𝑡𝑝𝑢𝑡= 1
𝑒𝑙𝑠𝑒 ⇒𝑜𝑢𝑡𝑝𝑢𝑡=0
Fig13
Fig. 12: Modi ied V/ CL con ol o so ansi ion. “Dashed line” s ands o all he modi ica ions on he basic scheme.
whe e ω0
sV
CL is he new elec ic equency o V/
CL con ol. RcCL is he eleasing coe icien . I
equals o “1” as long as V/ CL con olle is no
selec ed. A he ansi ion A he ansi ion mo-
men RcCL goes o “0” g adually h ough a low
pass il e . This means ha he V/ CL con ol
becomes independen a e ansi ion.
•The second modi ica ion consis s in adap ing Ω∗
by Eq. (4), assumed as an an icipa ion ac ion, o
educe all so s o speed de ia ions om he e e -
ence, including speed a ia ion due o ansi ion.
Ω∗0
V
CL = (Adap i e gain)·2nd O de LPF·
Ω∗−Ω) + RcCL ·kd1( −τd1),
(4)
whe e:
–Ω∗0
V
CL is he adap ed speed e e ence o V/
CL con ol.
–kd1( −τd1)pe mi s an ins an aneous ampli-
ica ion o speed e e ence a e he swi ching
momen . I is delayed by τd1 o keep a maxi-
mum alue o a sho ime s a ing om he
swi ching ins an .
–The 2nd o de LPF is conside ed as a e e -
ence sys em. The di e ence be ween i s ou -
pu and Ωis added o he e e ence speed
h ough he adap i e gain. So, he esponses
o he IM d i e and e e ence sys em a e in-
ended o be simila . This simple modi ica-
ion b ings many ad an ages, i does no only
so en he ansi ion bu also imp o es he
dynamic beha io by educing he o e shoo
and he e ec o he load o que.
–The adap i e gain is compu ed in unc ion o
Ω∗using a designed lookup able. This la -
e is o med o ele en-speed e e ence poin s
and he co esponding gain alues allowing
he smoo hes ansi ion om IFOC o V/
CL con ol.
4.3. Smoo hening he T ansi ion o
V/ OL Con ol
The new scheme o V/ OL con ol is shown in Fig. 13.
Same imp o emen s a e made o V/ OL con ol o
smoo hen he ansi ion o i om senso less IFOC o
V/ CL.
•~
V∗
sV
OL is linked o ~
V∗
sV
CL by Eq. (5):
ω0
sV
OL =ωsV
OL +ω0
sV
CL −ωsV
OL ·ARcOL,
(5)
whe e:
–ω0
sV
OL is he new s a o elec ic speed o V/
OL con ol.
–ARcOL is an adap i e eleasing coe icien .
I s alue is always equal o “1” ill he V/
OL con ol is selec ed, hen, i ansi s o “0”
g adually h ough a 1s o de LPF wi h an
adap i e ime cons an “τ”. A lookup able
is composed o ele en speed e e ence poin s
and he co esponding “τ” alues chosen o
allow he smoo hes elease o ~
V∗
sV
OL .
c
2018 ADVANCES IN ELECTRICAL AND ELECTRONIC ENGINEERING 9