17
Original scientific paper
Journal of Microelectronics, 
Electronic Components and Materials
Vol. 52, No. 1(2022), 17 – 27
https://doi.org/10.33180/InfMIDEM2022.103
How to cite:
V. Ambrožič et al., “Operation of Permanent Magnet Synchronous Motor after Open-circuit Battery Supply Fault" , Inf. Midem-J. Microelec-
tron. Electron. Compon. Mater., Vol. 52, No. 1(2022), pp. 17–27
Operation of Permanent Magnet Synchronous 
Motor after Open-circuit Battery Supply Fault 
Vanja Ambrožič
1
, Mitja Breznik
2
, Mitja Nemec
1
1
University of Ljubljana, Faculty of Electrical Engineering, Ljubljana, Slovenia 
2
Kolektor Group d.o.o., Idrija, Slovenia
Abstract: This paper presents a method for post-fault operation with reduced performance of a permanent magnet synchronous 
motor following a battery supply open circuit fault. The approach upgrades the previously developed fast model-oriented supply fault 
detection algorithm based on comparing actual and estimated values of a DC-link voltage. The latter is determined by the model of 
supply circuit and observer of battery open-circuit voltage. Implementing a post-fault concept does not require additional hardware, 
as it uses quantities from an already existent vector control algorithm. After the fault is detected, the algorithm is upgraded by a 
DC-link voltage cascade control loop.  An analytical approach to parametrization of the controller, based on the linearization and the 
reduction of the system’s transfer function, is also proposed. Simulations and experimental results have validated the performance of 
the post-fault algorithm.
Keywords: power supply fault; model-based detection algorithm; control loop parametrization; DC-link voltage regulator; fault-
tolerant design
Delovanje sinhronskega motorja s trajnimi 
magneti po izpadu akumulatorskega napajanja 
Izvleček: Članek predstavi metodo za nadaljevanje obratovanja sinhronskega stroja s trajnimi magneti z zmanjšano zmogljivostjo po 
izpadu akumulatorskega napajanja.  Pristop nadgrajuje že razviti modelsko orientirani algoritem za ugotavljanje izpada napajanja, ki 
sloni na primerjavi dejanske in ocenjene vrednosti napetosti vmesnega tokokroga. Slednjo ugotavljamo prek modela napajalnega 
vezja in opazovalnika napetosti odprtih sponk akumulatorja. Implementacija koncepta obratovanja stroja po napaki ne zahteva 
dodatne aparaturne nadgradnje, ker uporablja že obstoječe količine, potrebne za normalno delovanje. Regulacijski algoritem po 
detekciji napake nadgradimo s kaskadno zanko za regulacijo napetosti vmesnega tokokroga. V članku je predstavljeno tudi analitično 
parametriziranje regulatorja, ki sloni na linearizaciji in redukciji sistemske prenosne funkcije. Simulacije in eksperimentalni rezultati 
potrjujejo učinkovitost obratovanja po napaki. 
Ključne besede: izpad napajanja, modelno zasnovani detekcijski algoritem, parametriziranje regulacijske zanke, regulator napetosti 
vmesnega tokokroga, zasnova odporna na izpade 
* Corresponding Author’s e-mail: vanja.ambrozic@fe.uni-lj.si
1 Introduction
Battery-supplied permanent magnet synchronous 
motor (PMSM) often appears as the main or auxiliary 
component in various mobility applications due to its 
high efficiency and high power density [1]. Therefore, 
research on fault mechanisms and post-fault operation 
of systems that enable key operational features in a ve-
hicle is the basis for providing appropriate safety and 
reliability. The probability of a supply system failure has 
already been extensively investigated [2].  However, 
the development of diagnostic and post-fault opera-
tion methods for a drive in case of power supply failure 

18
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27
lags behind the successfully implemented solutions 
that cope with the faults in PMSM, converter, and/or 
appropriate sensors [3]–[10]
The interruption of the supply system can be caused 
by damage or change of operational state (e.g., discon-
nection due to safe stop) of individual drive subsys-
tems. One of these is an intermediate DC/DC converter 
whose failure may prevent the proper energy flow. 
Method for fault detection of switching elements in a 
step-down converter feeding a brushless DC machine 
is proposed in [11]. Different fault scenarios in a paral-
lel DC/DC converter for interconnecting electrical drive 
and various power sources are analyzed in [12].
The fault mechanism and consequent detection meth-
ods for battery voltage sources have been thoroughly 
investigated. The number of connections between 
individual battery cells increases the fault probability, 
which is reflected through galvanic interruptions, an 
increase of impedance, or a decrease in power capac-
ity. The fault/breakage detection of welds between 
battery cells in electric vehicles, based on a statistical 
energy capacity determination, has been investigated 
in [13]. A comprehensive overview of supply systems 
fault diagnosis based on Li-Ion technology has been 
presented in [14].
A high-voltage battery system in hybrid and electric 
vehicles is connected to the rest of the drive through 
a contactor. The interruption of the contactor, with 
subsequent disconnection of a supply circuit, is imple-
mented in case of a safe stop or emergency shutdown 
following an accident. In case of an emergency shut-
down, a safe level of DC-link voltage has to be provided 
in conformity with ECE R94 [15]. A method for reduc-
ing the voltage without a braking resistor has been 
presented in [16]. Operation is based on upgrading the 
field-oriented control (FOC) with a DC-link voltage con-
troller and modulation index controller. The detection 
method and guidelines for the determination of con-
troller parameters are not presented. 
Supervision of the energy flow and DC-link voltage 
level is crucial in drives with diode rectifiers. Dynamic 
braking of the PMSM can be achieved by directing the 
mechanical energy into additional Joule losses in the 
machine’s windings. The DC-link voltage level is limited 
by keeping machine losses equal to braking power 
through stator current d and q components in the field 
reference frame [17]. Control algorithm for braking of 
switched reluctance machine is also proposed [18]. 
Limited braking capability and consequent regenera-
tive performance are also present in battery-fed elec-
trical drives that are close to energy capacity limits. In 
[19], the approach for limiting the energy flow into the 
DC supply through the increase of losses by injecting 
high-frequency current components into the d axis of 
an induction motor is presented.
Operating conditions, similar to those caused by the 
fault of the supply system, can be found in FOC dou-
bly-fed induction generators in island operation [20]. 
The system provides for the DC load voltage and sup-
plies the rotor windings with AC voltage oscillating at 
slip frequency. A separate control loop controls DC-link 
voltage, adapting it to the reference. A cascade volt-
age control of a three-phase rectifier DC-link voltage 
has been studied in [21]. The output power of the load 
is determined by the observer that allows for the im-
plementation of the inner control loop based on the 
power reference.
Finally, it can be concluded that due to the complexity 
of supply systems, several sources and causes can inter-
rupt the energy flow. If the PMSM drive operates in the 
regenerative mode or the field weakening region, this 
fault may produce secondary damage caused by the 
failure of electronic components due to overvoltage [9].
In this paper, a detection method and appropriate con-
trol algorithm that prevents secondary damage with-
out the need for additional hardware modifications 
(e.g., braking resistors, electronic components with 
higher values of the maximum permitted voltages) is 
presented. In order to ensure a fast and robust transi-
tion to the post-fault operation, a model-oriented algo-
rithm for real-time detection of supply interruption has 
been applied. As a result, the diagnostic method and 
post-fault algorithm constitute a fault-tolerant system, 
based on the most commonly used configuration of 
the supply system and control method for PMSM (FOC).
Post-fault control approach and analytical guidelines 
for controllers’ parametrization provide for:
- maintaining the DC-link voltage level inside 
predetermined operational boundaries. The 
maximum voltage is defined by the maximum 
permitted voltage of the converter’s electronic 
components. On the other side, the minimum 
voltage level is defined by the minimum supply 
voltage required by the converter’s peripheral 
and logic units in the case of controllers with a 
single supply.
- control of PMSM electrical quantities, even in field 
weakening region.
- potential for upgrading the control scheme that 
enables the drive’s deceleration by injecting the 
currents that cause additional losses in the sys-
tem.
- sufficient time for the electronics to enter the 
safe-stop mode.

19
2 System definition
The proposed method is developed for a standard and 
straightforward system of battery-fed PMSM drive. 
The system consists of Li-Ion batteries, supply cables, 
inverter with accompanying capacitor bank, and a 
PMSM. Block scheme of individual parts of the drive 
with potential failure locations is depicted in Fig. 1. 
Although this paper presents a method designed for 
a specific configuration, the approach could be easily 
adapted to different types of machines and topologies 
of supply circuits, since it is based on modelling of the 
supply circuit and well-established control principle.
Figure 1: Graphic representation of the analyzed sys-
tem.
Block scheme showing FOC of PMSM drive, capable of 
operating in field weakening region, is shown in Fig. 2 
[22]. The scheme already includes proposed fault de-
tection and a post-fault approach (inside a dotted rec-
tangle), described in detail in the following sections. 
The symmetrical voltage model of PMSM in d-q rotor 
field coordinates is determined by (1), where v
d
, v
q
, i
d
 
and i
q
 represent stator voltages and currents in a rotor 
reference frame. R
s
, L
d
, L
q,
 and ψ
m
 denote stator resist-
ance, inductances, and flux of the rotor’s permanent 
magnet, respectively. Electromagnetic torque T
e
 in case 
of surface-mounted permanent magnet motor (SPM) 
or interior permanent magnet motor (IPM) with similar 
L
d
 and L
q
 can be defined by (2), where p
p
 is the number 
of pole pairs. Electric power is defined by (3).
 
0
ds dq d
qd sq q m
vR sL Li
vL Rs Li
ω
ω ωψ
+−  
=+
 
+
  
  (1)
 3
 
2
ep qm
Tp iψ =     (2)
 3
()
2
ep dd qq
pp vi vi =+    (3)
3 Fault detection
The dynamics of post-fault response depends on fast 
and reliable fault detection to avoid false positives or 
false negatives. In this paper, the post-fault algorithm 
relies on the model-based detection that has been 
proven effective [23]. Possible open-circuit fault lo-
cations depicted in Fig. 1 can be quickly detected by 
monitoring the change of DC-link voltage. However, 
only a specific change to this phenomenon must be 
considered, while the changes due to regular opera-
tion should be ignored. This task can be achieved by 
comparing the measured DC-link voltage (measure-
ment performed in almost every drive) to the one ob-
tained through the model.
Furthermore, the model and corresponding control cir-
cuit must be robust and self-adaptable to slow chang-
es. These are caused by battery state-of-charge and 
parasitic effects, which occur during regular operation 
and react only to fast changes, owing to an open cir-
cuit. The model is based on a Thévenin’s circuit shown 
in Fig. 3, where i
batt
 and v
batt
 denote battery current and 
voltage, i
DC
 and v
DC
 are DC-link current and voltage. v
oc
 
is the battery’s open-circuit voltage, and R
i
 is its internal 
resistance. Supply cable is represented by resistance R
c
 
and inductance L
c
 of the transmission path, as calculat-
ed from [24]. Total capacitance C
DC
 and series resistance 
R
esr
 represent DC-link capacitor bank.
Since linear elements form the electrical circuit, the 
model estimated DC-link voltage (
 
ˆ
DC
v ) could be cal-
culated as the sum of two terms that depend on i
DC
 and 
v
oc
, respectively
 ˆˆˆ
DC oc
iv
DC DC DC
vvv =+     (4)
Figure 3: Lumped element electric circuit model of DC 
supply.
Figure 2: The control scheme of PMSM drive with add-
ed detection and post-fault control blocks.
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

20
The transfer functions of the two terms are defined as
 
()
()
()
()
2
2
ˆ
()
1
==
++ +++
=
−− ++ −
DC
i
DC
iDC
DC
ic DC esrc cc DC esr
DC icesrc DC
i
vs
Fs
is
RRsC RRRs Ls LC R
sC RRRs LC
 (5)
and
 
()
()
()
2
ˆ
()
ˆ
1
1 
==
+
=
++ ++
oc
v
DC
voc
oc
DC esr
DC esricc DC
vs
Fs
vs
sC R
sC RR Rs LC
  (6)
as represented in a block scheme in Fig. 4.
Figure 4:  Block diagram of a DC supply model.
As previously mentioned, the main idea of the diag-
nostic approach to battery open circuit detection is 
to monitor the difference between the estimated and 
measured v
DC
 (Fig. 5). Consequently, in an open circuit, 
the measured and the estimated value will differ. How-
ever, the erroneous estimation may also generate the 
difference, thus leading to a false positive. The causes 
for this effect are either altered battery voltage (v
oc
) or 
erroneously estimated/modified lumped circuit pa-
rameters.
The change of battery open circuit voltage v
oc
 during 
operation is normal and relatively slow. Thus its esti-
mation should be incorporated into the diagnostic 
scheme (Fig. 5) that now has to fulfill two tasks: 
- 
 
ˆ
oc
v estimation: 
 
ˆ
oc
v is needed for 
 
ˆ
DC
v estimation in 
(4) and (6). Fortunately, during normal operation 
 
ˆ
oc
v , and consequently the v
DC
 estimation error e, 
change slowly; thus the slower and stable inner 
control algorithm should adapt the estimated 
value 
 
ˆ
oc
v and so minimize the error. A possible so-
lution could be using the sliding mode observer 
[23,25]. 
- Battery open-circuit fault detection: Sudden er-
ror between measured and estimated DC-link 
voltage e, after 
 
ˆ
oc
v has been adequately tracked, 
could be caused only by some other fault source 
– open circuit – as will be explained next.
At steady-state (5) and (6) can be simplified, leading to 
the equation for DC link voltage error from Fig. 5
 ()
()
()
ˆˆ
3
ˆ
2
mD CD CD Co cD Ci C
DC oc dd qq iC
DC
iR R
RR
vv vv
vv vi vi
v
εε
ε
εε
εζ ζ
ζ
−+ +
−+ +
=−±= ±=
−±
    
 (7)
The term consists of:
- e – the difference between measured (v
DC
) and 
estimated (
 
ˆ
DC
v ) DC-link voltage. Note that i
DC
 in 
 
ˆ
DC
v is calculated from control variables instead of 
being measured. 
- z e
 – the error due to erroneous modeling (param-
eter mismatch, discretization, measurement er-
ror, etc.)
The maximum possible error of z e
 (denoted as z e_max
) 
can be analytically determined and serves as a diag-
nostic threshold. Hence, as depicted in Fig. 5, if the er-
ror e m
 exceeds z e
 = z e_max
, the open circuit has undoubt-
edly occurred.
Such diagnostic method proves to be reliable and fast, 
allowing for fast post-fault operation.
Figure 5: Block diagram of the estimation model.
4 Post fault operation
Once the fault is detected, post fault operation can be 
performed in two ways. The first one is to shut down 
the inverter. While the simplest to implement, this op-
tion is only viable when the drive is not in a field weak-
ening mode or the inverter’s DC link does not supply 
the electronics. During field weakening, current control 
has to stay active. Also, if the inverter’s DC link should 
supply the electronics, rotational energy can be used 
to provide the necessary power, at least for some time.
The DClink voltage can change significantly during the 
fault transient, as the magnetic energy stored in ma-
chine windings dissipates to the DC-link regardless of 
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

21
the post fault operation mode. Therefore, for the in-
verter to survive such a fault, it has to be designed ac-
cordingly. Only then can the inverter take appropriate 
action to control the DC-link voltage.
4.1 Inverter design considerations for open circuit 
battery supply fault
Immediately following the fault, i
d
 should remain un-
changed. It is already zero in a constant torque mode, 
while in field weakening mode, it prevents back emf 
from rising above a safe level. The only active required 
power flow is for supplying the control electronics and 
the motor losses. Hence, the active power current com-
ponent i
q
 after the fault transient will be almost zero.
After the fault, the energy circulates only between the 
machine and the inverter. After inserting (1) into (3), 
three distinct power flow components can be identi-
fied: winding losses (P
loss
), reactive power (P
r
), and me-
chanical power (P
mech
) (8).
 
22
3
2
()
q
d
es ds qd dq q
mq dq dq
loss
r
P
mech
P
P
di
di
pR iR iL iL i
dt dt
ii iL L ωψ ω


=+ ++






++ −




  
  
  (8)
In order to assess the under-/overvoltage, the energy 
transferred from fault occurrence until the end of fault 
transient (t
1
) should be calculated .
 
() () ()
1
1
0
2
0 22
0
3
3
()
24
t
e
t
qq
sd qq md dq
Wp dt
Li
Ri ii iL Ld t ωψ
Δ= =
++ +− −
∫
∫
 (9)
The amount of transferred magnetic (reactive) energy 
does not depend on the transient duration but only on 
the current values at the beginning of the transient i
d0
, 
i
q0
 (t=0). Additionally, since i
d
 remains almost unaltered, 
its derivative can be neglected.
The total duration of the fault transient (t
1
) consists of 
fault detection time t
det
 and current transient time Dt 
(assuming local linearity, see (13))
 
1 det
tt t =+ Δ                  (10)
The final equation for the total energy transferred dur-
ing the transient  is
 
2
0 2
00
3
() ()
22 2
q
sq mq detq
i
t
WR ii tL ωψ

Δ
Δ= ++ − 


             (11)
With a shorter total time t
1
, the energy transferred be-
tween the machine and inverter will be lower; thus, the 
under-/overvoltage on DC-link will be reduced.
The main parameters determining the duration of the 
current transient (D t) are machine inductances L
d
, L
q,
 
and the voltage difference between maximum applica-
ble inverter voltage  and back emf.
 
22
3
DC
dq
v
vv +=                  (12)
Insertion of (1) into (12), neglecting the terms with re-
sistance R
s
 (being significantly smaller than other con-
tributions) and di
d
/dt (since i
d
 remains constant), and 
also assuming linear transition of i
q
 (di
q
/dt = –i
q0
/D t) 
results in .
 
2
0 22
00
() ()
3
q
DC
qd dq q m
i
v
LL iL i
t
ψ ωω ω
−
++=− −
Δ
            (13)
The solutions of the quadratic equation  for D t with dif-
ferent i
d
, i
q
 pairs in i
d
 < 0 half-plane (which corresponds 
to normal operation/field weakening), show that the 
transient time Dt mainly depends on i
q
 (Fig 6).
Figure 6: Transient duration.
The total energy transferred will alter the energy level 
in the DC-link capacitor (14), resulting in either an in-
creased or decreased DC voltage (v
DC_1
) compared to 
the voltage at the instant of the fault v
DC_0
. Using (11) 
and (14), the relation between final DC-link voltage 
v
DC_1
 and DC-link capacitance is obtained (15).
 
22
_1 _0
()
2
DC DC DC
Cv v
W
−
Δ=                 (14)
-30- 20 -100 10 20 30
i
d
 (A)
-30
-20
-10
0
10
20
30
i
q
 (A)
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

22
 
2
_1 _0 2
3
α =−
DC DC
DC
vv
C
                 (15)
where
 
()
2
0 2
00
2 2
αω ψ

=+
Δ 
+


−

det
qq
sq mq
Li
it Ri
t
Equation (15) serves as a design criterion for assessing 
the drive’s tolerance during the supply fault to over-
voltage, when in generator mode, and undervoltage, 
in motoring mode. 
4.2 Post-fault DC-link voltage control
After the supply fault detection, the DC link voltage 
has to stay under control, especially in field weakening 
mode, to prevent further faults. Since the field weaken-
ing controller still sets the direct current id in this case, 
the voltage can be controlled by quadrature current i
q
. 
The sign of i
q
 has to account for the direction of rotation 
(dotted area in Fig. 2).
4.2.1 Tuning DC-link voltage controller
The selection of voltage controller parameters is espe-
cially critical for achieving fast transients. In order to set 
the parameters analytically, the control loop (Fig. 7) has 
to be analyzed.
Electric power drawn from the DC-link depends on the 
speed and torque of the machine, with losses in ma-
chine and inverter neglected (16). 
 3
 
2
ee qm
p
pT i
p
ω
ωψ = =
                  (16)
Thus, the DC-link voltage is
 1
DC DC esrD C
DC
vi Ri
sC
=+
                  (17)
Figure 7: Block scheme of a post-fault voltage control 
loop.
As can be seen, the control loop is nonlinear and de-
pends on multiple input parameters. However, several 
simplifications can be introduced to obtain an analyti-
cally solvable system.
Nonlinear relationship in DC-link current i
DC
 calculation 
from power p
e
 can be linearized as the loop will operate 
only around the DC-link voltage reference value. Hence 
DC link voltage (v
DC
) can be considered pseudo-con-
stant resulting in the division (far right-hand side on 
Fig. 7) being replaced by multiplication with constant 
v
DC
-1
.
Furthermore, as the mechanical speed w dynamics is 
several orders of magnitude slower than electrical tran-
sients, the mechanical speed can also be considered 
constant during these transients. Consequently, the di-
rect axes current i
d
, which is used only for field weaken-
ing (the latter is a function of w ), can also be considered 
constant. Hence, the i
d
 control loop in Fig. 7 is oblivious. 
Thus the transfer function of the system, including i
q
 
current PI controller with gains K
p_q
, K
i_q
 (dashed in Fig. 
6), is
 
*
2
__ __
32
__
3
2
()
()
m
s
DC DC
pq DC esrp qi qD Ce sr iq
qp qs iq
F
vC
KC Rs KK CRsK
Ls KR sK s
ωψ
=⋅
++ +
⋅
+++
               (18)
The equation can be rewritten in a form where individ-
ual time constants can be easily identified
 
_
_ _1 2
*
12
1( 1)
3
2( 1)(1 )
pq
DC esr
iq iq
sm
DC DC q
K
sC Rs
K K
F
ss s vCL
ττ
ωψ
ττ

++ 


=
++
       (19)
where
 
()
1,2
2
__ _
2
4
q
pq sp qs qi q
L
KRKR LK
τ =
+± +−
               (20)
By setting the current controller parameters K
p_q
/K
i_q
 = 
τ
2
, the order of the transfer function is further reduced 
[26,27]. Additional simplification is possible by neglect-
ing R
esr
. The result of this reduction is a second-order 
system (21).
 
1_
*
1
31
2( 1)
pq
sm
DC DC q
K
F
ss vCL
τ
ωψ
τ
=
+
                 (21)
This system can be further reduced to an integrator 
with a delay (22).
 
'
()
d
s
s
e
Fs k
s
τ −
=                   (22)
with gain defined as
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

23
 
1_
*
3
'
2
pq
m
DC DC q
K
k
vCL
τ
ωψ =
                  (23)
The delay t d
 is calculated as proposed in [27]
 
1
22
s
d
τ τ
τ =+
                   (24)
using a time constant of the initial transfer function 
1
, 
and sampling time t S
. 
Now a voltage PI controller can be parametrized. “Sim-
ple control” or “Skogestad (SIMC)” guidelines offer the 
recipe for setting the (voltage) controller parameters 
for the transfer function in form equal to (25), where 
c
 is 
the desired closed loop time constant.
 
'
11
,4 ()
()
pi c
c
k
k
ττ θ
τθ
== +
+
(25)
The v
DC
 control should function regardless of the opera-
tional point. As the system gain k’ actually depends on 
rotor speed and DC voltage reference setpoint (23), the 
proportional k
p
 gain in (25) has to adapt accordingly. 
5 Simulation results
First, the simulations were performed to test if the sim-
plifications of the DC-link voltage control loop in sec-
tion 4 were justified. The parameters correspond to 
those of the experimental system (Tables 1 and 2).
The transient response of a simplified system based on 
(18) and fully nonlinear system described in (8), were com-
pared. The controller parameters for both systems were 
set using (25). As shown in Fig. 8 and 9, the only signifi-
cant difference between the two systems occurs during 
the initial transient of the DC-link voltage. As the voltage 
controller response is much slower, the transient error 
owed to the simplification has no effect. The agreement 
between the responses of both models during the main 
transient shows that the simplifications were justified and 
that the voltage controller was correctly parameterized.
Additionally, the enlarged transient of i
q
 in Fig. 9 corrob-
orates the duration of the transient obtained from (13).
6 Experimental results
6.1 Experimental setup
Detection algorithm and post-fault operation were 
tested on a laboratory back-to-back setup consisting of 
two identical PMSM drives with independent battery 
power supplies, thus enabling four-quadrant opera-
tion (see Figs. 10-11). Control and detection algorithm 
(running at 6 kHz) was implemented on a motor con-
troller build around ARM Cortex M4 MCU with support 
for floating-point arithmetic. The drive under test is 
25
30
35
40
v
D C
* ,
v
D C
  (V)
v
D C
*
v
D C
 - simp
v
D C
 - nonlin
-40
-20
0
20
40
i
q
* , i
q
 (A)
i
q
*
 - simp
i
q
 - simp
0 0.005 0.01 0.015 0.02 0.025
t 
(s)
-40
-20
0
20
40
i
q
* , i
q
 (A)
i
q
* - nonlin
i
q
 - nonlin
Figure 8: Transient response of simplified (“simp”) and 
fully nonlinear (“nonlin”) model in motoring mode.
Figure 9: Transient response of simplified (“simp”) and 
fully non-linear (“nonlin”) model in generator mode.
30
35
40
45
v
D C
* ,
v
D C
  (V)
v
D C
*
v
D C
 - simp
v
D C
 - nonlin
-40
-20
0
20
40
i
q
* , i
q
 (A)
i
q
*
 - simp
i
q
 - simp
0 0.005 0.01 0.015 0.02 0.025
t 
(s)
-40
-20
0
20
40
i
q
* , i
q
 (A)
i
q
* - nonlin
i
q
 - nonlin
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

24
torque-controlled, which is the usual operation mode 
in traction applications. A second coupled drive emu-
lates the mechanical load by running at a controlled 
speed. Power electronics is integrated into both motors’ 
housings. Fault emulation was performed by switching 
off the contactor of the torque controlled drive.
Inverter and power supply parameters are shown in Ta-
ble 1, while the machine parameters are given in Table 2. 
Table 1: DC Supply parameters
Symbol Value Unit
v
oc
34.5 (V)
R
i
86 (mΩ)
R
c
94 (mΩ)
L
c
3.82·10-6 (H)
r
c
0.6 (mm)
d
c
0.1 (m)
l
c
1.8 (m)
C
DC
 1.4 (mF)
R
esr
94.1 (mΩ)
k
i
0.5 /
Table 2: PMSM parameters
Symbol Value Unit
R
s
 28.5 (mΩ)
L
d
 164 (mH)
L
q
 186 (mH)
Ψ
m
 0.022 (Wb)
p
p
 4 /
Figure 10: Experimental setup with back-to-back con-
figuration.
6.2 Fault detection
Firstly, the performance of the DC-link voltage observer 
for fault detection, described in section 3, is shown un-
der normal operation [23]. As we can see (fig. 12), the 
observer tracks the actual voltage during the drive’s 
operating point changes without any significant devia-
tions.
Figure 12: DC supply model response in case of peri-
odic changes of drive’s operating point.
The response of the fault detection algorithm during 
the fault in generator mode is shown in fig. 13. At the 
instant of the fault, the error between the actual DC-
Figure 11: Schematic of the experimental setup.
32
34
36
v
DC
, v
DC
 (V)
v
DC
v
D C
-2
0
2
, 
 max
 (V)
 max
-20
0
20
i
D C
 (A)
0 123 4 5
t 
(min)
33
34
35
v
oc
 (V)
30
40
50
v
DC
, v
DC
 (V)
v
DC
v
D C
-10
0
10
, 
 max
 (V)
 max
0.01 0.012 0.014 0.016 0.018 0.02
t 
(s)
-50
0
50
i
d
, i
q
 (A)
i
d
i
q
Figure 13: Transient response during a fault in generator 
mode without voltage controller (1500 rpm, i
q
* = - 30 A).
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

25
link voltage and observer output almost immediately 
exceeds the threshold resulting in instantaneous fault 
detection. Note that this test is only for demonstration 
purposes, as no pos-fault action is undertaken.
6.3 Post fault operation
Results in this section show the system response with 
both the fault detection and post-fault control meth-
od activated. A fast voltage control loop response can 
be observed, as the DC-link voltage stabilizes quickly 
to the reference value (Fig. 14-15) in both the motor 
and generator mode. In motor mode (Fig. 14), a short-
lasting and minor undervoltage occurs. Hence, the 
drives with electronics supplied from the DC-link can 
continue to operate even after the fault. Even more 
important is the behavior in generator mode (Fig. 15). 
Here the voltage controller prevents more extended 
overvoltage, which would occur without a post-fault 
voltage controller (Fig. 13). Additionally, the overvolt-
age is very close to the theoretically calculated value 
from (15), where during generator operation, 46,8 V 
was measured. In comparison, 46,5 V was calculated 
from the analytical equation based on system param-
eters. The combined detection and activation time of 
four sampling intervals (three for fault detection and 
one for starting the post-fault mode) was considered 
for the analytical approach. This confirms the practical-
ity of (15), when designing the drive, thus providing 
the basis for the maximum voltage ratings of electronic 
components.
Safe DC-link voltage level with a disconnected power 
supply can be provided only by the fading kinetic ener-
gy of the system’s rotating parts. Its duration depends 
on the actual speed at the instant of the fault and the 
system’s inertia. However, this time should be more 
than enough for the electronics to enter the safe-stop 
mode, including the storage of relevant parameters. 
Hence, almost all of the (kinetic) energy is supporting 
a low-consuming operation of the control electronics.
Figure 15: Transient response during fault in generator 
mode (1500 rpm, i
q
* = - 30 A).
7 Conclusion
This paper proposed an approach to post-fault opera-
tion following the malfunctioning of the battery power 
supply of a PMSM drive. The main goal is to control the 
level of DC-link voltage and provide for a temporary 
supply of the controller’s electronics before entering 
the safe-stop mode. The functionality of a post-fault 
operation, following successful fault detection, has 
been confirmed by the simulation and experimental 
results. These also confirm the validity of the analytical 
terms to determine limit values of DC-link voltage dur-
ing the transition to the post-fault operation. Further 
work will focus on applying the concept to more com-
plex configurations of the supply system and methods 
for dynamic braking.
8 References
1. D. Cabezuelo, E. Ibarra, E. Planas, I. Kortabarria, 
and J. I. Garate, ‘Rare-Earth Free EV and HEV Mo-
tor Drives: State of the Art’ , in PCIM Europe 2018; 
International Exhibition and Conference for Power 
25
30
35
40
v
DC
, v
DC
 (V)
v
DC
v
D C
v
D C
*
-5
0
5
, 
 max
 (V)
 ma x
-20
0
20
i
D C
 (A)
0.01 0.02 0.03 0.04 0.05
t 
(s)
-50
0
50
i
d
, i
q
 (A)
i
d
i
q
Figure 14: Transient response during a fault in motor 
mode (1500 rpm, i
q
* = 30 A).
30
40
50
v
DC
, v
DC
 (V)
v
DC
v
D C
v
D C
*
-5
0
5
, 
 max
 (V)
 ma x
-20
0
20
i
D C
 (A)
0.01 0.02 0.03 0.04 0.05
t 
(s)
-50
0
50
i
d
, i
q
 (A)
i
d
i
q
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

26
Electronics, Intelligent Motion, Renewable Energy 
and Energy Management, Jun. 2018, pp. 1–8.
2. D. Wanner, Controlling over-actuated road vehicles 
during failure conditions. Stockholm: Engineer-
ing Sciences, KTH Royal Institute of Technology, 
2015. Accessed: Sep. 29, 2018. [Online]. Available: 
http://urn.kb.se/resolve?urn=urn:nbn:se:kth:di
va-166819
3. J. Zheng, Z. Wang, D. Wang, Y . Li, and M. Li, ‘Review 
of fault diagnosis of PMSM drive system in electric 
vehicles’ , in 2017 36th Chinese Control Conference 
(CCC), Jul. 2017, pp. 7426–7432. 
 https://doi.org/10.23919/ChiCC.2017.8028529.
4. Y. Da, X. Shi, and M. Krishnamurthy, ‘Health moni-
toring, fault diagnosis and failure prognosis 
techniques for Brushless Permanent Magnet Ma-
chines’ , in 2011 IEEE Vehicle Power and Propulsion 
Conference, Sep. 2011, pp. 1–7. 
 https://doi.org/10.1109/VPPC.2011.6043248.
5. A. M. El-Refaie, ‘Fault-tolerant permanent magnet 
machines: a review’ , IET Electr. Power Appl., vol. 5, 
no. 1, pp. 59–74, Jan. 2011, 
 https://doi.org/10.1049/iet-epa.2009.0117.
6. T. A. Najafabadi, F. R. Salmasi, and P. Jabehdar-
Maralani, ‘Detection and Isolation of Speed-, DC-
Link Voltage-, and Current-Sensor Faults Based on 
an Adaptive Observer in Induction-Motor Drives’ , 
IEEE Trans. Ind. Electron., vol. 58, no. 5, pp. 1662–
1672, May 2011, 
 https://doi.org/10.1109/TIE.2010.2055775.
7. J. O. Estima, N. M. A. Freire, and A. J. M. Cardoso, 
‘Recent advances in fault diagnosis by Park’s vec-
tor approach’ , in 2013 IEEE Workshop on Electrical 
Machines Design, Control and Diagnosis (WEM-
DCD), Mar. 2013, pp. 279–288. 
 https://doi.org10.1109/WEMDCD.2013.6525187.
8. W. Zhang, D. Xu, P . N. Enjeti, H. Li, J. T. Hawke, and 
H. S. Krishnamoorthy, ‘Survey on Fault-Tolerant 
Techniques for Power Electronic Converters’ , IEEE 
Trans. Power Electron., vol. 29, no. 12, pp. 6319–
6331, Dec. 2014, 
 https://doi.org/10.1109/TPEL.2014.2304561.
9. L. Mathe, C. Kopacz, L. Bede, and P . O. Rasmussen, 
‘Sensor-fault tolerant control of PMSM in flux-
weakening operation using LKF observer’ , in Rail-
way and Ship Propulsion 2012 Electrical Systems for 
Aircraft, Oct. 2012, pp. 1–6. 
 https://doi.org/10.1109/ESARS.2012.6387427.
10. S. K. E. Khil, I. Jlassi, J. O. Estima, N. Mrabet-Bellaaj, 
and A. J. M. Cardoso, ‘Detection and isolation of 
open-switch and current sensor faults in PMSM 
drives, through stator current analysis’, in 2017 
IEEE 11th International Symposium on Diagnos-
tics for Electrical Machines, Power Electronics and 
Drives (SDEMPED), Aug. 2017, pp. 373–379. 
 https://doi.org/10.1109/DEMPED.2017.8062382.
11. J. Fang, W. Li, H. Li, and X. Xu, ‘Online Inverter Fault 
Diagnosis of Buck-Converter BLDC Motor Combi-
nations’ , IEEE Trans. Power Electron., vol. 30, no. 5, 
pp. 2674–2688, May 2015, 
 https://doi.org/10.1109/TPEL.2014.2330420.
12. H. Al-Sheikh, O. Bennouna, G. Hoblos, and N. 
Moubayed, ‘Modeling, design and fault analysis 
of bidirectional DC-DC converter for hybrid elec-
tric vehicles’, in 2014 IEEE 23rd International Sym-
posium on Industrial Electronics (ISIE), Jun. 2014, 
pp. 1689–1695. 
 https://doi.org/10.1109/ISIE.2014.6864869.
13. Y. Cai, L. Lu, P . Shen, X. Feng, H. Wang, and M. Ouy-
ang, ‘Online Weld Breakage Diagnosis for the Bat-
tery of Electric Vehicle: A Data-Driven Approach’ , 
in 2016 IEEE Vehicle Power and Propulsion Confer-
ence (VPPC), Oct. 2016, pp. 1–5. 
 https://doi.org/10.1109/VPPC.2016.7791739.
14. C. Wu, C. Zhu, Y. Ge, and Y. Zhao, ‘A review on fault 
mechanism and diagnosis approach for Li-ion 
batteries’ , J. Nanomater., vol. 2015, pp. 1–9, 2015.
15. United Nations Economic Commission for Europe 
(United Nations Economic and Social Council), 
‘Regulation No 94 — Uniform provisions con-
cerning the approval of vehicles with regard to 
the protection of the occupants in the event of a 
frontal collision’ . Sep. 20, 2012.
16. Z. Ke, J. Zhang, and M. W. Degner, ‘DC Bus Capaci-
tor Discharge of Permanent-Magnet Synchro-
nous Machine Drive Systems for Hybrid Electric 
Vehicles’ , IEEE Trans. Ind. Appl., vol. 53, no. 2, pp. 
1399–1405, Mar. 2017, 
 https://doi.org/10.1109/TIA.2016.2636279.
17. V. R. Jevremović and B. Jeftenić, ‘An Efficient Brak-
ing Algorithm for Interior Permanent Magnet 
Synchronous Motors’ , Nov. 2005.
18. S. M. Yang and J. Y. Chen, ‘Controlled Dynamic 
Braking for Switched Reluctance Motor Drives 
With a Rectifier Front End’ , IEEE Trans. Ind. Electron., 
vol. 60, no. 11, pp. 4913–4919, Nov. 2013, 
 https://doi.org/10.1109/TIE.2012.2233696.
19. Y. Wang and R. D. Lorenz, ‘Creative usage of stator 
flux linkage to fast brake electric vehicles’ , in 2016 
IEEE Transportation Electrification Conference and 
Expo (ITEC), Jun. 2016, pp. 1–6. 
 https://doi.org/10.1109/ITEC.2016.7520278.
20. G. D. Marques and M. F. Iacchetti, ‘Sensorless Fre-
quency and Voltage Control in the Stand-Alone 
DFIG-DC System’ , IEEE Trans. Ind. Electron., vol. 64, 
no. 3, pp. 1949–1957, Mar. 2017, 
 https://doi.org/10.1109/TIE.2016.2624262.
21. T. Guo, J. Liu, S. Vazquez, L. Wu, H. Gao, and L. G. 
Franquelo, ‘Predictive direct power control for 
grid-connected power converters with an Ex-
tended State Observer based dc-link voltage 
regulator’, in 2016 IEEE 25th International Sympo-
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27

27
sium on Industrial Electronics (ISIE), Jun. 2016, pp. 
1218–1223. 
 https://doi.org/10.1109/ISIE.2016.7745068.
22. B. K. Bose and P. M. Szczesny, ‘A microcomputer-
based control and simulation of an advanced IPM 
synchronous machine drive system for electric 
vehicle propulsion’, IEEE Trans. Ind. Electron., vol. 
35, no. 4, pp. 547–559, Nov. 1988, 
 https://doi.org/10.1109/41.9178.
23. M. Breznik, V. Ambrožič, and M. Nemec, ‘Detec-
tion of open circuit fault in battery power supply 
feeding permanent magnet synchronous motor’ , 
IET Power Electron., vol. 11, no. 14, pp. 2377–2384, 
2018, 
 https://doi.org/10.1049/iet-pel.2018.5214.
24. H. A. Aebischer and B. Aebischer, ‘Improved for-
mulae for the inductance of straight wires’, Adv. 
Electromagn., vol. 3, no. 1, pp. 31–43, 2014.
25. S. K. Kommuri, M. Defoort, H. R. Karimi, and K. C. 
Veluvolu, ‘A Robust Observer-Based Sensor Fault-
Tolerant Control for PMSM in Electric Vehicles’, 
IEEE Trans. Ind. Electron., vol. 63, no. 12, pp. 7671–
7681, Dec. 2016, 
 https://doi.org/10.1109/TIE.2016.2590993.
26. C.-H. Park, D.-Y. Kim, H.-B. Yeom, Y.-D. Son, and J.-
M. Kim, ‘A Current Reconstruction Strategy Fol-
lowing the Operation Area in a 1-Shunt Inverter 
System’ , Energies, vol. 12, no. 8, p. 1423, Jan. 2019, 
 https://doi.org/10.3390/en12081423.
27. S. Skogestad, ‘Simple analytic rules for model re-
duction and PID controller tuning’ , J. Process Con-
trol, vol. 13, no. 4, pp. 291–309, Jun. 2003, 
 https://doi.org/10.1016/S0959-1524(02)00062-8.
Arrived: 05. 10. 2021
Accepted: 28. 01. 2022
V. Ambrožič et al.; Informacije Midem, Vol. 52, No. 1(2022), 17 – 27
Copyright © 2022 by the Authors. 
This is an open access article dis-
tributed under the Creative Com-
mons Attribution  (CC BY) License (https://creativecom-
mons.org/licenses/by/4.0/), which permits unrestricted 
use, distribution, and reproduction in any medium, 
provided the original work is properly cited.