Professional Documents
Culture Documents
The BLDC motor has Y– connected windings on the stator In a BLDC motor with Y– configuration, the windings
with a non salient rotor with permanent magnets. The are connected to a central point and power is applied to
assumptions of this mathematical model are: the remaining end of each winding, which means that the
• Negligible saturation effects. stator currents are related by the expression
• Constant winding inductances. ia + i b + i c = 0 . (5)
• Same parameters in all phases. Therefore, the only available voltage are the differential
The mathematical model, proposed by Chiasson [2005], voltages, i.e., known as line-to-line voltages.
then reads, To obtain a model based on the measured variables, the
line-to-line voltages are substituted in equation (1a). Thus,
di the line-to-line electrical dynamics are given by
D = u − Ri + Ep E R ω (1a)
dt dil
dθ Dl = ul − Ril + Ep E Rl ω , (6)
=ω (1b) dt
dt where
dω 1 1 d
= τe − τL (t) − ω , (1c) L−M 0 0
dt J J J Dl = 0 L−M 0 , (7)
0 0 L−M
where u ∈ R3 is the input voltage vector, i ∈ R3 is the
stator current vector, θ ∈ R is the angular rotor position, il = [iab ibc ica ]
T
= [ia − ib ib − ic ic − ia ]
T
ω ∈ R is the angular rotor velocity, J ∈ R+ is the rotor T T
inertia, d ∈ R+ is the friction coefficient, Ep is the voltage ul = [uab ubc uca ] = [ua − ub ub − uc uc − ua ]
constant, R ∈ R3×3 + is a diagonal matrix of the stator T
E Rl = [ab bc ca ] = [a − b b − c c − a ] .
T
645
2018 IFAC MICNON
646
Guadalajara, Mexico, June 20-22, 2018Alejandra de la Guerra et al. / IFAC PapersOnLine 51-13 (2018) 644–649
646
2018 IFAC MICNON
Guadalajara, Mexico, June 20-22, 2018Alejandra de la Guerra et al. / IFAC PapersOnLine 51-13 (2018) 644–649 647
τd
and Ip = 2τ p
. This implies that only two phases, with system trajectories (37) will be finally bounded with a
the same magnitude but opposite sign, are active at any bound given by
time. The remaining desired phase currents are obtained λm
as x12d = Ip is2 and x13d = Ip is3 taking into account the y(t0 ) < ymax . (40)
λM
phase shift for each phase as follows,
4. SIMULATIONS
2π
is2 (θ) = is1 θ −
3 To evaluate the performance of the proposed scheme are
presented numerical simulations of mathematical model
2π
is3 (θ) = is1 θ + . (1) using the parameters of PMAC ME0913 BLDC motor
3 given in Table 1. The simulations were made in MATLAB
SIMULINK with a fixed step time T = 0.1 [ms] and the
3.3 Electrical control loop Runge–Kutta solver. Also, it was simulated noise in the
measured phase current using an AWGN Gaussian noise
The internal control loop objective is to regulate the stator block with a signal–to–noise ratio of 1 [dB] and input signal
phase currents, for this, is used a PBC power of 0.1 [mW].
dx1d
ul = Dl + Rx1d − Ep (ê3 + x3d )E Rl − Ke1 , (33) Table 1. ME0913 BLDC motor parameters
dt
where the stator current error is defined as
Parameter Value
e1 = x1 − x1d ,
with x1d the desired line current. Thus, the electrical error J 0.0045 [Kgm2 ]
dynamics are defined as d 0.001 [Kgm2 /s]
ė1 = −D−1 −1
l (R + K)e1 − Ep Dl E Rl ẽ3 (34) R 0.0125 [Ω]
L 1 [mH]
or in matrix form Ep 0.1909 [V/rad/s]
ė1 = Ae e1 + Be (t) (35) τp 0.1808 [Nm/A]
where Max Voltage 96 [V]
Rated torque 32.53 [Nm]
Ae = −D−1
l (R + K) In both cases the total inertia is the sum of the wheel
Be (t) = −Ep D−1 inertia Jw and the motor inertia J, where Jw = 12 mR2
l E Rl ẽ3 .
with m = 3.17515 [Kg] and a tire radius R = 0.2578 [m],
3.4 Closed loop dynamics thus Jw = 0.10552 [Kgm2 ].
To tune this controller Λm2 , Λm3 and K gains must
If we define the state vector be chosen to minimize the steady state tracking error,
T T
y = eT T
1 em eO , (36) while the observer poles must be chosen as far from
the closed loop dynamics can be written as the imaginary axis of the complex plane as the system
bandwidth allows it. In this case, the gains for the ADRC
ẏ = Ay + B(t) (37) controller are Λm2 = 0.5/J and Λm3 = 1.5/J and the
where
electrical gain is K = 0.1910I with I ∈ R3×3 . For the
Ae O O Be (t) observer design, p = 5 and the poles of the polynomial
A = O Am O B(t) = Bm r̄1 . (38) approximation are p1 = p2 = p3 = p4 = p5 = −0.025. In
O O AO BO r̄2 the figures, the speed tracking error is identified as eω , the
control input is denoted as u and the output as ADRC.
Using (37), an analysis of the closed-loop dynamics of
the controller–observer scheme can be carried out, in a 4.1 Case 1: Trapezoidal reference
similar way to that presented in de la Guerra et al. [2016]
for the SRM. In this case, for reasons of space, we only In this case the load torque is shown in Figure 1 where the
present an outline of this analysis. It consists of three maximum torque load is equal to 50% of the rated torque,
parts: 1) It must be verified that the signals of interest with a maximum desired speed of 55 [rad/s]. Figure 2
(including ẽ2 , ẽ3 , z̃1 , ..., z̃p , e1 , e2 , e3 ) are bounded in a shows the speed tracking where the proposed scheme
region D. 2) Positive functions should be proposed for the can compensate the constant load torque in steady state.
mechanical dynamics, Vm , the estimation dynamics, VO , While, Figure 3 shows a constant speed estimation error
and for electrical dynamics, Ve . 3) Finally, it must proved and a speed tracking error that remains close to zero in
that the state y is finally bounded if the initial condition steady state. Figure 4 shows the input which is far below
y(t0 ) is small enough, using the positive functions defined the maximum voltage for all the simulation as in case 1;
in the previous point to form the function the control input form can be observe in the zoom window.
V (y) = Vm + VO + Ve ,
for which there exist positive gains λm and λM such as
4.2 Case 2: Driving cycle
λm y2 ≤ V (y) ≤ λM y2 , (39)
given Am and AO are Hurwitz, L > M and K = KT > In this case the reference is a test emission cycle
Ke |ω|I, with I ∈ R3×3 the identity matrix. Therefore, the US06 [U.S., 2008], designed to represent an aggressive
647
2018 IFAC MICNON
Guadalajara, Mexico, June 20-22, 2018
648 Alejandra de la Guerra et al. / IFAC PapersOnLine 51-13 (2018) 644–649
20
10
15
5
τL [Nm]
u [V]
10
0
5 -5
0 -10
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t [s] t [s]
Fig. 1. Load Torque for Case 1. Fig. 4. Input voltages for Case 1.
shows the tracking error and the bottom one shows the
60 estimation error for the proposed scheme. In Figure 8 it
can be observed that the input is below the maximum
50 voltage for all the route, also a zoom window shows control
input waveform. To evaluate the ADRC performance, the
40
ω [rad/s]
25
30
20
20
10 15
τL [Nm]
0
10
0 10 20 30 40 50 60
t [s]
5
Fig. 2. Speed tracking Case 1: ADRC (—), x3d (- - -).
0
2 0 100 200 300 400 500 600
t [s]
0
eω
0 10 20 30 40 50 60
150
t [s]
30
20
ê3
100
10
ω [rad/s]
0
0 10 20 30 40 50 60
t [s]
50
648
2018 IFAC MICNON
Guadalajara, Mexico, June 20-22, 2018Alejandra de la Guerra et al. / IFAC PapersOnLine 51-13 (2018) 644–649 649
0
Chiasson, J. (2005). Modeling and high performance
-5
control of electric machines, volume 26. John Wiley
0 100 200 300 400 500 600 & Sons.
t [s] Cortés-Romero, J., Luviano-Juarez, A., Sira-Ramı́rez, H.,
40 et al. (2010). Fast identification and control of an
uncertain brushless dc motor using algebraic methods.
20 In Power Electronics Congress (CIEP), 2010 12th In-
ê3
649