International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 12, No. 3, September 2021, pp. 1390∼1404

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v12.i3.pp1390-1404 ❒ 1390

Real time emulator for parallel connected dual-PMSM

sensorless control
Khaldoune Sahri1 , Maria Pietrzak-David2 , Lothfi Baghli3 , Abdelaziz Kheloui4
1,4
Laboratoire Commandes des Machines Electriques, Ecole Militaire Polytechnique, 16111 Bordj El Bahri, Algeria
3
Universite de Tlemcen, LAT, 13000 Tlemcen, Algeria
2
Université de Toulouse; INPT, UPS, LAPLACE ENSEEIHT, CNRS, 2 rue Charles Camichel, F-31071 Toulouse cedex 7,
France

Article Info ABSTRACT

Article history: This paper presents a real-time emulator of a dual permanent magnet synchronous mo-
tor (PMSM) drive implemented on a field-programmable gate array (FPGA) board for
Received Dec 9, 2019
supervision and observation purposes. In order to increase the reliability of the drive, a
Revised Mar 30, 2021 sensorless speed control method is proposed. This method allows replacing the phys-
Accepted Jul 2, 2021 ical sensor while guaranteeing a satisfactory operation even in faulty conditions. The
novelty of the proposed approach consists of an FPGA implementation of an emulator
Keywords: to control the actual system. Hence, this emulator operates in real-time with actual
system control in healthy or faulty mode. It gives an observation of the speed rotation
Emulator in case of fault for the sake of continuity of service. The observation of the rotor posi-
FPGA tion and the speed are achieved using the dSPACE DS52030D digital platform with a
Multi-machine system digital signal processor (DSP) associated with a Xilinx FPGA.
PMSM
Sensorless control
This is an open access article under the CC BY-SA license.

Corresponding Author:
Khaldoune Sahri
Laboratoire Commandes des Machines Electriques
Ecole Militaire Polytechnique
P.O. Box 17, 16111 Bordj El Bahri, Algeria
Email: [email protected]

1. INTRODUCTION
Nowadays we can observe very intensive development of AC variable speed and torque drives. Fre-
quently, their topologies use an association of multi-converters and multi-machines systems. These drives
operate in several embedded systems such as vehicles, aircraft, ships, trains, and other industrial applications
where weight and volume reduction are essential.
To assure the best quality of PMSM vector control, the rotation speed or the rotor position have to
be precisely known. Generally, the principal system variables, such as the currents and rotation speed, are
measured using physical sensors. To increase reliability and reduce costs, mechanical sensors have to be
suppressed. Consequently, the speed information must be reconstituted from measures of electrical quantities
using deterministic state observers. Hence, we reconstruct one or several non-measurable or non-accessible
system states, in fact, an observer is an estimator which operates in closed loop and has its own dynamics
according to the real system. These dynamics are modified by a gain matrix to cancel the estimation error and
ensure convergence. To observe the speed and / or position of a PMSM at different speed levels is always a
challenge. For PMSM drive two modes of operation must be taken into consideration of course when operating
without mechanical sensors [1]-[3]: i) high and medium speed operation, ii) low speed operation.

Journal homepage: http://ijpeds.iaescore.com

Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1391

Using the current measurement and the direct current (DC) bus voltage, various algorithms are used to
observe the speed of the machine but they does not show satisfactory fonctionnement in a wide range of speed
of the system drive. For example, observers who use the return of the electromagnetic field (EMF) and when
the machine is driven with insufficient speed to produce a measurable EMF (low speed), in this case the current
sensors cannot detect the position of the rotor which cause a bad observation of the speed [1], [4], [5]. Another
point to report is that most estimation and observation algorithms of the speed presented in the literature, use
an adjustable model with currents measure. In case of a current sensor fault these algorithms become useless.
In addition, the sensitivity of these observers to the parametric variations [5]-[7], and the static con-
verter dead time (inverter) limits the proper functioning of the observer’s and hence the system drive. A dual
PMSM system have more degrees of freedom than a single one, (structural actuator and sensors redundancy).
So, when a sensor fault appears, a solution that appears very simple is to switch to the actual speed sensor
[8]-[10]. The main challenge is to obtain a high dynamic sensorless-drive operation. Taking into account the
physical limitations in the real time control of the sensorless synchronous speed drive, we achieve excellent
dynamic performances at high and low speeds. The authors interest in this paper, at an experiments of a new
control strategy without a speed sensor and without the need to use current sensors it means we observe the
rotation speed using the virtual current obtained from the emulator. For a dual PMSM system fed by a dual
VSI. We apply an emulator for each machine with its load torque observer and the reference voltages required
for the operation of the emulator it means Two individual real time vector controls are applied to drive this
global system.
In this way, the obtained deterministic observer replaces the real physical measures, necessary to real-
time control system, by corresponding observed variable. Two possibilities exist for working with dSPACE
DS52030D digital platform; this emulator function can be implemented on the processor part or on the FPGA
target [11]-[13]. To respect its tolerant fault operating two Xilinx FPGA emulators are introduced to increase
the system reliability and to admit total physical measure suppression. Certainly, the natural and structural
redundancy exists in the DPMSM/DVSI system permitting to switch the speed sensor from first PMSM to
second one, or vice versa, when a fault occurs (offset and gain). In the case of current sensor fault its time
sharing operation is impossible and it is necessary to dispose three sensors for each PMSM to choose two of
them in vector control strategy.
The emulator concept with load torque definition, with ”virtual mechanical coupling” in the case of
common loads, corresponds to an introduction of analytic redundancy in parallel to real system. In this way
the virtual reconstituted variables replace the real physical measures in the control of this system [14]. So, the
permanent monitoring detects the sensor failure and generates the decision signal to replace the real measure by
corresponding virtual one. The experimental validation of this tolerant sensor fault vector control is made on
Laplace DPMSM/DVSI experimental bench and confirms its successful operation and its satisfactory reliability.
We present firstly, the different methods for the speed observation that exist in the literature. In the second part,
we will put the light on the proposed method which is valid only in the case of a multi-machine system. The
studied system is shown in Figure 1.

Figure 1. General synopsis of studied system

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1392 ❒ ISSN: 2088-8694

2. SYSTEM PRESENTATION
2.1. PMSM modelling
The electrical equations expressed in d,q Park frame can be written is being as [15]:
dId


 Vd = Rs .Id + Ls . − ωr .Ls .Iq

 dt
dIq

Vq = Rs .Iq + Ls . + ωr .Ls .Id + ωr .ϕf (1)
 dt
dω P P B 1

r

= . (Ld − Lq ) .Id .Iq + .ϕf .Iq − .ωr − .TL


dt J J J J
Where Id and Iq are the direct-axis and quadrature-axis currents (A), Vd , and Vq are the direct-axis and
quadrature-axis voltage (V), ωr is the velocity of the rotor of the motor (rad/s), TL is the load torque (N-m)
We will investigate the control problem of the PMSM with smooth air gap; we will let Ld = Lq = L
In our study we chose a d,q vector control strategy based on a proporsional-integral (PI) cascade control for
each PMSM. Thus, the q axis control is made of two loops. The outer loop is the for speed control and the inner
loop is for the control of the Iq component. The d axis control has only one loop allowing the direct current
regulation Id .
2.2. State observer
The equations detailed the adjustable model can be rewritten is being as [16]:
  

 (R s + s.L) . ˆ
Id = ω̂r .L.Iˆ q + ê d + V̂d + L.l11 . Id − ˆ
Id +

  
L.l12 . Iq − Iˆq





  
 (Rs + s.L) .Iˆq = ω̂r .L.Iˆd − êq + V̂q + L.l21 . Id − Iˆd +


  (2)

 L.l 22 . Iq − ˆq
I

  
ˆ

s.ê = L.l . I − I

d 31 d d



  
ˆ

s.êq = L.l42 . Iq − Iq



where Iˆd and Iˆq are the direct-axis and quadrature-axis estimated currents (A), V̂d and V̂q are the direct-axis
and quadrature-axis estimated voltage (V), ω̂r is the observed velocity of the rotor of the motor (rad/s).l11 , l12 ,
l21 ,l22 , l31 ,l42 are adjustable gains used to ensure a good stability of the model. This adjustable model Figure
2 will be used for all the observer versions proposed in this paper such as:
− Rotation speed observation with EMF observation
− Observation of the speed with control of the cross current error.

Figure 2. Adjustable model structure

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1393

The following section describe the experimental validation of the two method for the observation of
the rotation speed of the PMSM.

3. EXPERIMENTAL VALIDATION
The parameters of the PMSM are being as:
− Stator resistance Rs = 2.281ohms .
− Direct and quadrature inductance Ld = 23.173mH .
− Magnetic Flux ϕf = 0.241W b .
− Number of pairs of poles P = 4 .
− PMSM inertia J = 22, 1.10−4 kg.m2 .
− Viscous friction coefficient f = 1.10−4 N ms/rad .
3.1. Experimental bench
This experimental bench is composed of dual PMSM speed controlled drives fed by dual two level
PWM VSIs connected in parallel to DC bus as shown in Figures 3 dan 4. The vector control strategy associated
with sinusoidal SVPWM is chosen to control this experimental bench. The implementation is illustrated in
Figure 5. The common or separated load torque for each studied PMSM is generated thanks to industrial
PMSM controlled torque drives. The torque references are given by the DS1005 processor. In the FPGA target,
the following parts have been implemented:
− ADC, DAC and the two incremental encoders.
− Space Vector PWM strategy (SVPWM).
− Direct and reverse transformations 3-phases 123/αβ and αβ/dq for the two PMSMs .
− Computing the estimated voltages from the duty cycle of pulses and the DC bus voltage.

Figure 3. Experimental bench

Figure 4. General diagram power part of the experimental bench

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1394 ❒ ISSN: 2088-8694

Figure 5. Different control parts implemented in FPGA and PPC card

3.2. Observation of the rotation speed with EMF observer status

The Figure 6 shows the principle of the observation of the rotation speed based on the electromagnetic
field (EMF) observer status. The EMF ed is imposed to zero and this quantity is obtained from the adjustable
system model Figure 2. Next, the obtained control error must be linearized to assure a good observation of the
rotational speed ωˆr [16].

Figure 6. Rotation speed observation with EMF observer status

− Experimental results
In the Figure 7 the rotation speed responses of two M1 and M2 PMSM. The transient and steady state
response performances are satisfactory during all speed reference cycles and under different load torques. The
responses of dual-PMSM electrical variables, i.e. Id, Iq, Vd, and Vq , are presented in Figures 8 and 9.

200 Speed of M1. Speed of M2 Observed speed.

Rotation speed (rad/s)

100
140 −100
120
−120
0 100
4.7 4.8 4.9
100 x 10
5 −140
9.6 9.7 9.8
0 5
x 10
−100 −100
−200
7 7.2 7.4
5
x 10
−200
2 4 6 8 10
Time (s) 5
x 10
Figure 7. Rotation speed control

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1395

5
M1 Iq current.
M2 Iq current.
M1 Id current.

Currents (A)
0

−5
0 2 4 6 8 10
Time (s) 5
x 10
Figure 8. Current Id and Iq responses

500
Estimated Vd. Estimated Vq.
Estimated Voltage (V)

−500
0 2 4 6 8 10
Time (s) 5
x 10
Figure 9. Estimated voltages Vd and Vq

3.3. Observation of the speed control by controlling the cross currents error (analytical redundancy)
We can observe the position or the speed thanks to a PI controller using the crossed error between the
measured and the estimated currents Figure 10. The principle of this observer is given in the Figure 10. The
error is obtained by evaluating:
ϕf
(Iqest .Id − Idest .Iq ) − (Iqest − Iq ) . (3)
L

Figure 10. Rotation speed estimation with regulation of error cross currents

A robust and stable correction can be based on approach MRAS synthesized using through Popov’s
hyper stability criteria [17], [18].
− Experimental results
We analyze the PMSM system behavior with the same rotation speed reference cycles. So, the speed
responses of M1 and M2 PMSMs are satisfactory during all reference cycles, small oscillations are observed
in the steady state after load torque application as shown in Figures 11 and 12. The actual and observed
mechanical positions of PMSM are shown in the Figure 13. We notice the effective precision in both rotation
directions. In order to test the robustness of the Dual PMSM speed observation, both velocity and load profile

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1396 ❒ ISSN: 2088-8694

were applied using the different proposed methods. For a wide operation range (low, medium and high speed),
these algorithms are not satisfying. To address this issue, each algorithm must be used in one speed range.
EMF based observers are not used at low speed because of the low EMF and are better in medium and high
speed. Observation of the speed by controlling the cross currents error, is used for low and medium speed
range. Several others approaches are used for speed estimation. These methods require huge memory and
involve high computational complexity and memory storage [15], [19].

200 Observed speed. Speed of M1. Speed

speed of M2.
Rotation speed (rad/s)

100 150

100 −100

4.7 4.8 4.9

0 x 10
5
−150
9.6 9.7 9.8 9.9
100 5
x 10
0
−100
−200
−100 7 7.2 7.4
5
x 10

−200
4 6 8 10 12
Time (s) 5
x 10
Figure 11. Rotation speed control

10 Actual position of M2. Actual position of M1 Observed position of M1

10 10

5 5
8
0 0
4.1 4.105 4.11 4.115 4.12
Position (rad)

9.6 9.605 9.61 9.615 9.62

5 5
x 10 x 10
6

0
0 2 4 6 8 10
Time (s) 5
x 10
Figure 12. Actual and observed position

3.4. Natural structural redundancy

In the Dual PMSM system, there is a natural physical redundancy, because each machine has the
mechanical or electrical sensors to achieve a closed loop control. This redundancy will be called structural
with natural redundancy. The principle of this method is explained on the following figure: The Figure 13
show the proper functioning of the system with physical speed sensor, and when fault sensor operation occurs,
we can continue satisfactory operating with only one speed sensor shared by the two machine to achieve a
closed loop control with satisfactory operation.

Figure 13. Fonctionnement of Dual PMSM with physical sensors

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1397

The following method was adopted in order to validate our idea. The state of the signal select change
in function of fault in speed sensor . Without fault; the signal select takes the value of zero and when we have
fault in the first sensor the signal select passes to 1, then it passes to 2 when we have fault in the second sensor
(Figure 14). In our case the decision on faulty speed sensor if the sensor give a value of 80% for the reference
speed. For the decisional organ the following algorithms has been implemented:
(
1 if speed M1 ¡ 80 % of ref speed.
select =
2 if speed M1 ¡ 80 % of ref speed.

Figure 14. Decision device

− Experimental results
As shown in the Figures 15 and 16, we start the operation with the physical speed sensors that is to
say that the system has not yet suffered a defect, then a defect is caused, in our case it is indeed a defect in the
second sensor of Speed of the second motor and then we continuing the operation of the system with a single
speed sensor shared for the closed loop vector control of the two motors. It should be noted that the natural
switching between the physical sensor does not allow the system to be started (it is not possible to start with
one speed sensor).
Rotation speed (rad/s)

100
110
120 110
100
100
100
90 90
5.2 5.25 5.3 6.9 6.95 7
50 80
3.8 4 x 10
5
x 10
5

5
x 10
fault in the speed continuity of service with
sensor of M2. sensor of M1.

0
3 4 5 6 7 8
Time (s) 5
x 10
Figure 15. Rotation speed control

6 4
2
2
Current Iq of M2.
1
Currents Id,Iq (A)

4 0 0 Current Iq of M1.
−1 5 5.5
3.5 4
5
5 x 10
x 10

0
Current Id of M2. Current Id of M1.
−2
3 4 5 6 7 8
Time (s) 5
x 10
Figure 16. Currents Id and Iq

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1398 ❒ ISSN: 2088-8694

4. THE PROPOSED SENSELESS SPEED FOR THE SYSTEM

We describe in this section the design, implementation, and experimental validation of a real-time
hardware-in-the-loop emulation of a dual PMSM. The PMSM machine are modelled using a flexible piecewise
linear state space approach [20]; and are simulated in hard real-time with 10 ns time step, which enables high-
fidelity modelling of PMSM dynamics. We validate the real-time PMSM drive emulator by making real-time
comparisons with a reference hardware model of a PMSM drive. The validation of this real-time DPMSM
emulator is made by real-time comparison with a reference hardware model of a DPMSM drive. In this way
we want analyze experimentally the operation of the HIL-DPMSM drive emulation under different operating
scenarios, in health and faulty conditions. Therefore we experimentally demonstrate the capability of the
hardware in the loop PMSM drive emulation under various operating scenarios, including fault conditions
[21]-[23].
The implementation of the different control configurations is done using the dSPACE DS52030D
digital platform containing a PowerPC DS1005 and a Xilinx FPGA DS5203. The task dispatching between
PPC and FPFA can be modified during the final optimized final implementation of global DPMSM drive system
and also its load in common or separated configurations. The computing time on the FPGA processor is 10
ns. Thus, we can use the implementation of the continuous mathematical model of the PMSM directly on the
FPGA Target. The computing time on the PPC DS1005 is 100 µs. So the mathematical model of the PMSM
should be discretized.
In the FPGA target, the following parts have been implemented Figure 17:
− PWM, ADC, DAC and the two incremental encoders.
− Modulation (PWM).
− d,q /αβ transformations for the two PMSM.
− 123 /αβ and αβ/d,q transformations for the two PMSM.
− Mathematical model of the PMSM in this way a more complex PMSM model can be implemented on
FPGA target.

Figure 17. Different parts of real time control implemented in FPGA and PPC card

All the parts implemented are resumed in Figure 17. The mathematical model of the PMSM Figures
18 requires Luenberger observer to reconstruct the torque load. For the supply voltages Vd and Vq necessary
for the proper functioning of the emulator, it can be obtained from the references voltages of the actual system.
We can obtain them using the DC bus voltage and the pulses of the inverters [24]:

Vdq = [d]dq Vdc

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1399

3
Idq = [d]dq Idc
2

dd .Vdc = Vd

dq .Vdc = Vq

Figure 18. The mathematical model of the PMSM

d: is the matrix of control. Unfortunately, if we have a fault in physical sensors the references voltage
calculated by this method becomes instable and we do not have enough time to replace the faulty sensor by the
emulator variables. To fix this problem, the reference voltages are obtained by a closed loop vector control of
the mathematical model implemented on FPGA or PPC in Figure 19. So the voltage are calculated in closed
loop and we obtain a robust and reliable observer. For the torque load TL it can be observed by the following
algorithm (Figure 20).

Figure 19. Closed loop vector control of the mathematical model

Figure 20. Observed torque load

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1400 ❒ ISSN: 2088-8694

− Experimental results
The model of the PMSM machine is implemented on the FPGA target in order to validate it, in real
time conditions. Firstly, the laws are used for the controls implemented on the card PPC DS1005, PWM, and
calibration of inputs/outputs. The results presented in Figures 21-24 allow the comparison of the measurable
variables (stator currents and speed) of the PMSM, with the ones of the model running, in real time, on the
FPGA. This approach validates the proper functioning of the model, the parameter identification and the vector
control of the machine. Next, this model is implemented in the FPGA, in order to use the emulated model
for the control of the actual system.We validated the real-time emulation by making steady state and transient
operation comparisons with the actual PMSM .
Additionally, we demonstrate the ability of the real-time PMSM emulation, as a hardware in the loop
prototyping platform, to control the actual system in a wide speed range; from low to high speed. The key
advantages of the proposed the real-time emulation platform are twofold. First, the piecewise linear state space
modelling approach enables comprehensive modelling of PMSM, including the dynamics that occur on the
control [25], [26]. Second, the flexible modelling implementation enables the platform to be used as a powerful
tool for rapid prototyping and validation [8]. In the next section, we will present the results obtained during
a fault in the speed sensor, the observed torque variable must be reconstructed from the healthy machine as
shown in Figures 25-28.

200 Ref speed. M1. M2. Emul Proc. Emul FPGA.

Rotation speed (rad/s)

100 −100
140
120
−120
100 −140
0 4.5 4.52 4.54
5
9.16 9.18 9.2
x 10 5
50
x 10
0
−50
−100 −100
6.6 6.8 7
5
x 10

−200
0 2 4 6 8 10
Time (s) 5
x 10

Figure 21. Comparison of actual and emulated speed

5
Actual and emulated currents (A)

−5
2 3 4 5 6 7 8 9 10 11
Time (s) x 10
5
Actual and emulated current (A)

5
Actual current.
Current of emulator on FPGA.

−5
4.95 4.96 4.97 4.98 4.99 5
Time (s) 5
x 10

Figure 22. Actual current of the PMSM and their emulators

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1401

10
Actual Iq of M2.
Observed Iq.
5

Currents Iq (A)
Iq with emulator on FPG.
Iq with emulator on Proc.
0 6

0
−5 2
−2
0 −4
4.85 4.9 4.95
5 9 9.05 9.1 9.15 9.2
x 10
−10 5
x 10
4 5 6 7 8 9 10
Time (s) 5
x 10

Figure 23. Actual currents Iq of the PMSM and their emulator

10 10
10

5 5
8
0
Position (rad)

0 9 9.01 9.02
4 4.01 4.02 5
6 x 10
5 x 10

0
2 3 4 5 6 7 8 9 10 11
Time (s) x 10
5

Figure 24. Emulated and actual position

− Experimental results
− Using the emulator FPGA as observer

Figure 25. Rotation speed control

6
M2 Iq current.
Id and Iq currents (A)

4 M1 Iq current.

0
M1 Id current. M2 Id current.
3 4 5 6 7 8 9 10 11
Time (s) 5
x 10

Figure 26. Iq and Id currents

− Using the emulator PPC as the observer

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1402 ❒ ISSN: 2088-8694

Rotation speed (rad/s)

100
110
120
100
100
90
80 5.8 5.85 5.9
4.8 4.9 5
50 5
x 10
x 10
5

Fault in the speed Continuity of service with

sensor of M2. Emulator.

0
3 4 5 6 7 8 9 10 11
Time (s) 5
x 10

Figure 27. Rotation speed control

6
Iq M2 current .
Id and Iq currents (A)

Iq M1 current .
4

0
Id M2 current . Id M1current .
3 4 5 6 7 8 9 10 11
Time (s) 5
x 10

Figure 28. Iq and Id currents

The results obtained shows clearly that the mathematical model of the machine replaces accurately
the physical speed sensor. So, when we impose the rotation speed cycle for this DPMSM we observe good
performances of real rotation speeds controlled with emulated one. The load torque modification don’t modify
the system precision, the zooms shown in the Figures 16-18 proves the satisfying behavior of this DPMSM
system.
The results show that it is possible to continue the operation of the system in the case of a current
sensor fault, because the currents Id and Iq required for the vector control can be obtained directly from the
emulator which is implemented on the FPGA or the PPC. The results obtained also show that the emulator
can be used for the diagnosis and detection of the various existing mechanical fault by the use of the different
method of frequency analysis. It means that we can used the emulated mathematical model as a reference
system and when we have a mechanical fault, we can easily see this fault in the actual currents by comparing
with the emulated currents . Consequently, we can easily conclude on the nature of fault.
The Figure 29 as shown presents a comparison between the reconstitution of speed on FPGA and
processor. The results of comparison show that the implemented emulator on the FPGA or the PPC has the
same dynamics; so we can distribute the tasks efficiently on FPGA and processor. Therefore, the resources of
FPGA can be minimized in a very effective way.
Rotation speed (rad/s)

100
110
120
100
100
90
80 5.8 5.85 5.9
4.8 4.9 5
50 5
x 10
x 10
5

Speed with emulator Speed with emulator

on FPGA. on proc.

0
3 4 5 6 7 8 9 10 11
Time (s) 5
x 10

Figure 29. Emulated speed

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404
Int J Pow Elec & Dri Syst ISSN: 2088-8694 ❒ 1403

5. CONCLUSION
FPGA is used for the control of two PMSMs with operation at variable speeds and variable torque
loads. The objective of this work is to validate the performance of the real time dual-PMSM drive by using the
vector control law. We compare the reference system (which is the actual physical system) with the emulated
system implemented on FPGA and on the PPC. We also demonstrate the ability of the Hardware-in-the loop
emulation in real time to control the real system under fault of speed sensor conditions and we will demonstrate
also other faulty condition for example by injecting open phase fault, and current sensor fault in the emulator
drive system. In fact, this emulator can be considered as an adjustable model that does not use gains to adjust the
mathematical model; but it uses much more the observation of the load torque as parameter for the adjustment
of the model contrary to the existing methods in the literature.

ACKNOWLEDGEMENT
We thank all those who contributed to make this work, including my supervisors in the EMP school
and the LAPLACE laboratory in toulouse University and all my friends for their support.

REFERENCES
[1] A. S. Budden, R. Wrobel, D. Holliday, P. Mellor, and P. Sangha,” Sensorless control of permanent magnet machine
drives for aerospace applications,” 2005 International Conference on Power Electronics and Drives Systems, 2005,
pp. 372-377, doi: 10.1109/PEDS.2005.1619715.
[2] T. Mueller, P. Thiemann, C. See, A. Ghani, and A. Bati,”Direct flux control–a sensorless control method of
pmsm for all speeds–basics and constraints,” Electronics Letters, vol 53, no. 16, pp. 1110-1111, 2005, doi:
10.1049/el.2017.1772.
[3] J.-S. Hu, D. Yin, and Y. Hori,” Fault-tolerant traction control of electric vehicles,” Control Engineering Practice, vol.
19, no. 2, pp. 204–213, February 2011, doi: 10.1016/j.conengprac.2010.11.012.
[4] B. Gerard, S. Caux, and P. Maussion,” Redundant Position Observer Improvement for Sensorless PMSM
at Low Speed,” 2006 IEEE International Symposium on Industrial Electronics, 2006, pp. 2083-2088, doi:
10.1109/ISIE.2006.295894.
[5] Z. Song, Z. Hou, C. Jiang, and X. Wei,” Sensorless control of surface permanent magnet synchronous motor us-
ing a new method,” Energy Conversion and Management, vol. 47, no. 15, pp. 2451–2460, September 2006, doi:
10.1016/j.enconman.2005.11.009.
[6] B. Akin, S. B. Ozturk, H. A. Toliyat, and M. Rayner,” DSP-Based Sensorless Electric Motor Fault-Diagnosis Tools
for Electric and Hybrid Electric Vehicle Powertrain Applications,” in IEEE Transactions on Vehicular Technology,
vol. 58, no. 6, pp. 2679-2688, July 2009, doi: 10.1109/TVT.2009.2012430.
[7] M. E. H. Benbouzid, D. Diallo, and M. Zeraoulia,” Advanced Fault-Tolerant Control of Induction-Motor Drives
for EV/HEV Traction Applications: From Conventional to Modern and Intelligent Control Techniques,” in IEEE
Transactions on Vehicular Technology, vol. 56, no. 2, pp. 519-528, March 2007, doi: 10.1109/TVT.2006.889579.
[8] S. Bolognani, M. Zordan, and M. Zigliotto,” Experimental fault-tolerant control of a pmsm drive,” in IEEE Transac-
tions on Industrial Electronics, vol. 47, no. 5, pp. 1134-1141, Oct. 2000, doi: 10.1109/41.873223.
[9] R. L. de Araujo Ribeiro, C. B. Jacobina, E. R. C. Da Silva, and A. M. N. Lima,” Fault-tolerant voltage-fed PWM
inverter AC motor drive systems,” in IEEE Transactions on Industrial Electronics, vol. 51, no. 2, pp. 439-446, April
2004, doi: 10.1109/TIE.2004.825284.
[10] D. Bidart, M. Pietrzak-David, P. Maussion, and M. Fadel,” Mono inverter dual parallel PMSM-structure
and control strategy,” 2008 34th Annual Conference of IEEE Industrial Electronics, 2008, pp. 268-273, doi:
10.1109/IECON.2008.4757964.
[11] C. B. Jacobina, R. L. de Araujo Ribeiro, A. N. Lima, and E. C. Da Silva,” Fault-tolerant reversible AC motor drive
system,” in IEEE Transactions on Industry Applications, vol. 39, no. 4, pp. 1077-1084, July-August 2003, doi:
10.1109/TIA.2003.814567.
[12] T.-H. Liu, J.-R. Fu, and T. A. Lipo,” A strategy for improving reliability of field-oriented controlled induction
motor drives,” in IEEE Transactions on Industry Applications, vol. 29, no. 5, pp. 910-918, Sept.-Oct. 1993, doi:
10.1109/28.245714.
[13] O. Wallmark, L. Harnefors, and O. Carlson,” Control Algorithms for a Fault-Tolerant PMSM Drive,” in IEEE Trans-
actions on Industrial Electronics, vol. 54, no. 4, pp. 1973-1980, Aug. 2007, doi: 10.1109/TIE.2007.895076.
[14] R. R. Errabelli and P. Mutschler,” Fault-Tolerant Voltage Source Inverter for Permanent Magnet Drives,”in IEEE
Transactions on Power Electronics, vol. 27, no. 2, pp. 500-508, Feb. 2012, doi: 10.1109/TPEL.2011.2135866.
[15] S. Maiti, C. Chakraborty, and S. Sengupta,” Simulation studies on model reference adaptive controller based speed

Real time emulator for parallel connected dual-PMSM sensorless control (Khaldoune Sahri)
1404 ❒ ISSN: 2088-8694

estimation technique for the vector controlled permanent magnet synchronous motor drive,” Simulation Modelling
Practice and Theory, vol. 17, no. 4, pp. 585–596, April 2009, doi: 10.1016/j.simpat.2008.08.017.
[16] R. P. Burgos, P. Kshirsagar, A. Lidozzi, F. Wang, and D. Boroyevich,” Mathematical Model and Control Design for
Sensorless Vector Control of Permanent Magnet Synchronous Machines,” 2006 IEEE Workshops on Computers in
Power Electronics, 2006, pp. 76-82, doi: 10.1109/COMPEL.2006.305655.
[17] W. Sae-Kok, D. Grant, and B. Williams,” System reconfiguration under open-switch faults in a doubly fed induc-
tion machine,” IET Renewable Power Generation, vol. 4, no. 5, pp. 458–470, September 2010, doi: 10.1049/iet-
rpg.2010.0005.
[18] A. M. Mendes and A. M. Cardoso,” Fault-Tolerant Operating Strategies Applied to Three-Phase Induction-
Motor Drives,”in IEEE Transactions on Industrial Electronics, vol. 53, no. 6, pp. 1807-1817, Dec. 2006, doi:
10.1109/TIE.2006.885137.
[19] D. Campos-Delgado, D. Espinoza-Trejo, and E. Palacios,” Fault-tolerant control in variable speed drives: a survey,”
IET Electric Power Applications, vol. 2, no. 2, pp. 121–134, March 2008, doi: 10.1049/iet-epa:20070203.
[20] J. Poon, E. Chai, I. Čelanović, A. Genić, and E. Adzic,” High-fidelity real-time hardware-in-the-loop emulation
of PMSM inverter drives,” 2013 IEEE Energy Conversion Congress and Exposition, 2013, pp. 1754-1758, doi:
10.1109/ECCE.2013.6646919.
[21] F. Alvarez-Gonzalez, A. Griffo, B. Sen, and J. Wang,” Real-Time Hardware-in-the-Loop Simulation of Permanent-
Magnet Synchronous Motor Drives Under Stator Faults,” in IEEE Transactions on Industrial Electronics, vol. 64, no.
9, pp. 6960-6969, Sept. 2017, doi: 10.1109/TIE.2017.2688969.
[22] Q.-T. An, L.-Z. Sun, K. Zhao, and L. Sun,” Switching function model-based fast-diagnostic method of open-switch
faults in inverters without sensors,”in IEEE Transactions on Power Electronics, vol. 26, no. 1, pp. 119-126, Jan. 2011,
doi: 10.1109/TPEL.2010.2052472.
[23] F. Zidani, D. Diallo, M. E. H. Benbouzid, and R. Naı̈t-Saı̈d,” A Fuzzy-Based Approach for the Diagnosis of Fault
Modes in a Voltage-Fed PWM Inverter Induction Motor Drive,” in IEEE Transactions on Industrial Electronics, vol.
55, no. 2, pp. 586-593, Feb. 2008, doi: 10.1109/TIE.2007.911951.
[24] C.-C. Yeh and N. A. Demerdash,” Induction Motor-Drive Systems with Fault Tolerant Inverter-Motor Ca-
pabilities,” 2007 IEEE International Electric Machines Drives Conference, 2007, pp. 1451-1458, doi:
10.1109/IEMDC.2007.383642.
[25] M. Karpenko and N. Sepehri, ”Hardware-in-the-loop simulator for research on fault tolerant control of electrohy-
draulic actuators in a flight control application,” Mechatronics, vol. 19, no. 7, pp. 1067– 1077, October 2009, doi:
10.1016/j.mechatronics.2009.01.008.
[26] R. D. Turney, C. Dick, D. B. Parlour, J. Hwang,” Modeling and implementation of dsp fpga solutions,” in Proceedings
of International Conference on Signal Processing Applications and Technology (ICSPAT), 1999, pp. 1–4.

Int J Pow Elec & Dri Syst, Vol. 12, No. 3, September 2021 : 1390 – 1404