EasyChair Preprint

№ 5215

Dynamic Surface Control of the Axial-Flux

Permanent Magnet Motor with Speed Sensorless
Algorithm

Manh Tung Ngo, Quang Dang Pham, Huy Phuong Nguyen and
Tung Lam Nguyen

EasyChair preprints are intended for rapid

dissemination of research results and are
integrated with the rest of EasyChair.

March 24, 2021

 Dynamic Surface Control of the Axial-Flux Permanent
Magnet Motor with Speed Sensorless Algorithm

Ngo Manh Tung1,2, Pham Quang Dang1, Nguyen Huy Phuong1, Nguyen Tung
Lam1[0000-0003-4108-8275]
1Hanoi University of Science and Technology, 1 Dai Co Viet Road, Viet Nam
2 Hanoi University of Industry, No.298, Cau Dien Street, Bac Tu Liem District, Viet Nam
[email protected]

Abstract. The permanent synchronous motor drive system incorporates magnetic

bearings to perform speed control and balance rotor control between the two sta-
tors. The paper designed a system to perform adjusting motor speed sensorless
based on measured current and voltage components. The electromotive force
(back-EMF) generated in the stator is estimated by a High-Gain observer. The
angular position and velocity rotor is calculated through the α-β components of
the back-EMF. The motor drive is built in a vector control structure based on the
rotor flux. In which based on the Lyapunov stabilization function, the dynamic
surface control method is used to calculate the axial position and speed controller
of the motor. Simulation results show that the proposed controller and observer
are fully capable of meeting the system control requirements.

Keywords: Magnetic self-bearing motor, Back-EMF, High-gain observer, dy-

namic surface control.

1 Introduce

In recent years, the motor with integrated magnetic bearing has been increasingly stud-
ied and applied due to its advantages compared to traditional ball bearing [1]. The motor
has the magnitude of the air gap between the stator and the rotor is constantly varying,
affecting the motor parameters and increasing the nonlinearity of the target. To deal
with this problem, the control process needs to separate the channel between the torque
generating the rotation and the force acting the axial position. Hence, based on the prin-
ciple of control based on the rotor flux, the current id controls the axial force, while the
radial current iq controls the torque [2]. Dynamic surface control method is used to
design the axial position controller and the motor speed control. This is a technique
applied to nonlinear system objects, applied to grip control and stability for the system
[3]–[5]. Therefore, this method is very suitable for the subject in the paper. The Dy-
namic Surface Control (DSC) technique features Backstepping and sliding multi-sur-
face control (MSS) [6]. However, it has added a first order filter to overcome the prob-
lem of explosion of term.
During the control process, the control loops need to convert the coordinate system [7].
This leads to a need for accurate angular position information of the rotor by sensors
2

that measure the rotational angle. This increases costs, increases size, affects the me-
chanical strength and maintains equipment maintenance. Therefore, methods of replac-
ing speed sensors with calculation techniques are studied and published more and more.
In the published works on non-measuring rotation speed control for synchronous mo-
tors, two approaches for estimating the angular position and rotor speed can be given
[8], [9].
This article presents a High-Gain observation based on the back-EMF electrodynamic
reactivity estimation, it has the advantage that the rule of observation is simple, easy to
design, and there is no project using HG observers for magnetic motors. Although [10]
uses a similar observational structure, the hydraulic system applies. The paper proposes
a drive system using a vector control structure, in which the HG observer estimates the
back-EMF value from the stator current and the reference voltage value, thereby calcu-
lating the Rotor angle position and rotation speed.
The feasibility of the proposed methods is demonstrated through simulation of the con-
trol system without speed sensor on the software. Simulation results show that both
position response and speed response follow the set trajectory in fast time. The system
is also stable to the coupling effects of the two control loops.

2 Mathematical Model

The permanent magnet synchronous motor incorporating the axial magnetic bearing
has the structure shown in Fig 1. The motor structure consists of a rotor and two stators
[1]. The torque T that rotates the rotor is the sum of the two component moments and
the F axial force that stops the rotor from moving in the z direction is calculated by the
difference of the two axial forces. To design the control system, the mathematical mo-
del of the motor is shown on the dq rotating coordinate or αβ static coordinate system.
For L’sd0 and L’sq0: the d- axis and q- axis inductance per gap unit; Lsl: leakage induc-
tance; g = g0 ± z: clearance between stator and rotor; g0: clearance at equilibrium posi-
tion; z: displacement. Voltage vector equation of a stator on the αβ coordinate system
is given as:
U s = is Rs + Ls dis / dt + Es (1)
The inductance is presented as [1], [11]:
Lsd = 3L'sd 0 / 2g + Lsl ; Lsq = 3L'sq 0 / 2g + Lsl (2)
Where is the stator current vector, Us is the stator voltage vector. Es is a sine function
and is expressed as components on the αβ coordinate system as follows:
Es = - m e sin  e ; Es = m e cos  e (3)
With ωe is the rotor speed, θe is the position of the rotor flux vector. From (1) we have:
dis / dt = − Rs is / Lsd + us / Lsd - Es / Lsd
 (4)
dis / dt = − Rs is / Lsq + us / Lsq - Es / Lsq
 3

Fig. 1. The structure of the motor

We have the equation for calculating the total torque and total axial force as follows
[11], [12]: F 4 K Fd i f id ; T 2 KT iq (5)

3 Design dynamic surface controls

The motor's rotation equation can be written as follows:

T TL J .d / dt or can be written as  = (T − TL ) / J (6)
According to (6) we have:  = iq .2 KT / J − TL / J = M .iq − N (7)

The definition of a sliding surface is: z1 =  - d (8)

The derivatice of z1 is: z1 =  − d = Miq − N − d (9)
According to DSC, virtual control signal: iq = −c z1 / M + ( N + d ) / M (10)
Choose a factor c > 0. The signal controls iq through the filter which is the first-order
filter to obtain the reference signal iqref : iqref = (iq − iqref ) /   (11)

Choose time constant τꞷ>0. Set: iq = iqref − iq (12)

Suppose a Lyapunov candidate function is as follows: V = z12 / 2 + iq 2 / 2 (13)

In order for the speed loop to be controlled asymptotically, then the derivative of V:
V = z1 z1 + iq iq  0 (14)
According to (14), the time constant satisfies the following inequality:
    / B2 (15)
4

where B >0 is a maximum constant that iq  B .

According to (5): z = ( F − FL ) / m = 4 K Fd i f id / m − F / m = P.id − Q (16)

Definition of sliding surface: S = z +  z (choose factor   0 ) (17)

Calculations similar to the speed controller, virtual control signal are defined [10]:

id = Q / P −  z / P − cz .sat (S ); idref = (id − idref ) /  z (18)

Choose cz > 0, τz > 0 and the time constant should be satisfied:  z   z / A 2
(19)

Where A is a maximum constant that id  A . The position control loop will stabilize
after a finite time with z approaching sliding surface S.

4 High-gain observer design

From (3) and (4) we have:

dis
es = - m e sin  e = − Ls − Rs is + us h
dt
(20)
dis
es = m e cos  e = − Ls  − Rs is  + us  h
dt
The derivative property equation for e s and e s  according to the HGO set is written
as follows [10], [13], with 1/  and 1/   are the coefficients of HGO:
eˆs = (− Ls is − Rs is + us − eˆs ) /   = −eˆs /   + h /  
(21)
eˆs = (− Ls is − Rs is + us  − eˆs  ) /   = −eˆs  /   + h /  
The definition of the error between the actual and the estimated value is as follows:
es = es − eˆs ; es = es − eˆs (22)
According to [13], we have the following values satisfying the inequality:
( −1  ) t
es  e( −1  )t es (0) +  . (t ); es  e  es (0) +   . (t ) (23)
From (20), the values esα and esβ are harmonic functions, so the coefficients hmax and

hmax exist such that: h  h max and h  hmax (24)

Thus, the smaller the coefficients ɛα and ɛβ, the smaller es and es are blocked by
the smaller bounding regions es () and es () . Then, the smaller the value ɛα and
ɛβ, the faster the rate of convergence of the estimated value follows the real values esαvà
esβ of the back-EMF. From the output signal of High-gain observer, the values of the
angular position and rotor speed are calculated as follows:
ˆe = arctan(− Eˆ s / Eˆ s ) (25)
 5

ˆ e = Eˆ s2 + Eˆ s2 / m (a ) or: ˆ e = dˆ / dt (b) (26)

Note that if the speed is calculated according to (35a), the value of the speed depends
on the linkage flux λm, which is influenced by the ambient temperature. If used accord-
ing to (35b), the speed is more reliable, especially in the medium and high-speed range,
but is affected by process noise.

5 Simulation and result

The parameters of the motor include: phase resistance is 2.6 Ω; the air gap between the
stator and the rotor is 1 mm; the rotor mass is 0.28kg; the moment of inertia is 10.6x10-
6 kgm2; the amplitude of the flux flux generated by the permanent magnet is λm = 0.022
Wb.
When the set speed is 4500 rpm, at the moment of 0.65 s the load torque affects the
system. Fig. 3 shows that the estimation speed is very fast with real speed value and the
error is negligible.

Fig. 3. Speed response when set value is 4500 rpm under the influence of load torque
Fig. 4 shows the response of the two current id and iq, where the iq current at the begin-
ning has a large value to accelerate the motor to quickly reach the set value, then the
value drops very small at the end of the transition mode. Fig.5 shows that when the load
is applied, the back-EMF component on of the observer adheres to the back-EMF com-
ponent of the motor very fast and few errors. Thus, in this case the HG observer esti-
mates that the components esα and esβ have a sinusoidal form, with the same amplitude
and following the observed value.
When changing the speed setting value from 2500 rpm-3500 rpm-1500 rpm in Fig. 7.
Estimated speed is still capable of adhering to the real speed. The components es at the
output of the HGO continue to be the same sinusoidal form and follow the motor Back-
EMF. Changing the speed setting value does not change the estimation time of the
monitor.
6

Fig. 4. Current response id and iq

Fig. 5. esα and esβ component of the motor and esα and esβ set of the observation

Fig. 6. Speed response

a) Response of e_alpha b) Response e_beta

Fig. 7. Back-EMF components on α-β trục axis when speed setting value is changed

6 Simulation and result

The paper presented the control system without measuring the rotation speed for syn-
chronous motor with integrated magnetic bearing, calculating the dynamic surface con-
trollers and designing the HGO unit in the system. By taking the input signal to the
observer, the reference voltage and stator current perform an estimate of the back-EMF
electromotive force as the basis for calculating the position and rotor speed. The system
works stably at medium speed range upwards, in which the interaction between axial
position control and speed control has also been limited. However, the peak phenome-
non caused by the visual disturbance affects the quality of the output speed.
 7

References
1. Q. Nguyen and S. Ueno, “Salient pole permanent magnet axial-gap self-bearing motor,” no.
Im, 2009, doi: 10.5772/intechopen.83966.
2. R. He and Q. Han, “Dynamics and Stability of Permanent-Magnet Synchronous Motor,”
Math. Probl. Eng., vol. 2017, 2017, doi: 10.1155/2017/4923987.
3. B. S. J. K. Hedrick, Dynamic Surface Control of Uncertain Nonlinear Systems. .
4. Y. H. Lan and Lei-Zhou, “Backstepping Control with Disturbance Observer for Permanent
Magnet Synchronous Motor,” J. Control Sci. Eng., vol. 2018, 2018, doi:
10.1155/2018/4938389.
5. Z. J. Yang, T. Nagai, S. Kanae, and K. Wada, DYnamic surface control approach to adaptive
robust control of nonlinear systems in semi-strict feedback form, vol. 16, no. 1. IFAC, 2005.
6. T. D. Nguyen, K. J. Tseng, S. Zhang, and H. T. Nguyen, “A novel axial flux permanent-
magnet machine for flywheel energy storage system: Design and analysis,” IEEE Trans. Ind.
Electron., vol. 58, no. 9, pp. 3784–3794, 2011, doi: 10.1109/TIE.2010.2089939.
7. N. P. Quang and J.-A. Dittrich, Power Systems Vector Control of Three-Phase AC Ma-
chines. .
8. Q. D. Nguyen and S. Ueno, “Sensorless speed control of inset type axial gap self-bearing
motor using extended EMF,” 2010 Int. Power Electron. Conf. - ECCE Asia -, IPEC 2010,
pp. 2260–2264, 2010, doi: 10.1109/IPEC.2010.5542012.
9. T. D. Nguyen and G. Foo, “Sensorless control of a dual-airgap axial flux permanent magnet
machine for flywheel energy storage system,” IET Electr. Power Appl., vol. 7, no. 2, pp.
140–149, 2013, doi: 10.1049/iet-epa.2012.0048.
10. D. Won, W. Kim, D. Shin, and C. C. Chung, “High-Gain Disturbance Observer-Based Back-
stepping Control With Output Tracking Error Constraint for Electro-Hydraulic Systems,”
no. m, pp. 1–9, 2014.
11. Q. Dich, Nguyen, and S. Ueno, “Axial position and speed vector control of the inset perma-
nent magnet axial gap type self bearing motor,” IEEE/ASME Int. Conf. Adv. Intell. Mech-
atronics, AIM, pp. 130–135, 2009, doi: 10.1109/AIM.2009.5230025.
12. Q. D. Nguyen and S. Ueno, “Analysis and control of nonsalient permanent magnet axial gap
self-bearing motor,” IEEE Trans. Ind. Electron., vol. 58, no. 7, pp. 2644–2652, 2011, doi:
10.1109/TIE.2010.2076309.
13. H. Khalil and L. Praly, “High-gain observers in nonlinear feedback control,” Int. J. Robust
Nonlinear Control, vol. 24, no. 6, pp. 993–1015, 2014, doi: 10.1002/rnc.v24.6.