Академический Документы
Профессиональный Документы
Культура Документы
Research Article
Velocity Regulation in Switched Reluctance Motors under
Magnetic Flux Saturation Conditions
Received 4 July 2017; Revised 8 November 2017; Accepted 10 December 2017; Published 2 January 2018
Copyright © 2018 Victor M. Hernández-Guzmán et al. This is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
We propose a controller for velocity regulation in switched reluctance motors under magnetic flux saturation conditions. Both
hysteresis and proportional control are employed in the internal electric current loops. A classical PI velocity controller is employed
in the external loop. Our control law is the simplest one proposed in the literature but provided with a formal stability proof. We
prove that the state is bounded having an ultimate bound which can be rendered arbitrarily small by a suitable selection of controller
gains. Furthermore, this result stands when starting from any initial condition within a radius which can be arbitrarily enlarged
using suitable controller gains. We present a simulation study where even convergence to zero of velocity error is observed as well
as a good performance when regulating velocity in the presence of unknown step changes in external torque disturbances.
definite, with 𝑟 > 0 representing the stator winding resistance, m1+, dL1(q)/dq
and scalars 𝐽 > 0, 𝑏 > 0 represent rotor inertia and 1
0.5
viscous friction coefficient, whereas inductance 𝐷(𝑞, 𝐼) = 0
diag{𝐷1 (𝑞, 𝐼1 ), 𝐷2 (𝑞, 𝐼2 ), 𝐷3 (𝑞, 𝐼3 )} is a positive definite diag- −0.5
onal matrix, where −1
5 5.2 5.4 5.6 5.8 6 6.2
q (rad)
𝜓𝑠 𝛽𝐿 𝑖 (𝑞)
𝐷𝑖 (𝑞, 𝐼𝑖 ) = , 𝑖 = 1, 2, 3, (5) m2+, dL2(q)/dq
1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖2 1
0.5
0
and 𝐶(𝑞, 𝐼) = diag{𝐶1 (𝑞, 𝐼1 ), 𝐶2 (𝑞, 𝐼2 ), 𝐶3 (𝑞, 𝐼3 )}, where −0.5
−1
5 5.2 5.4 5.6 5.8 6 6.2
𝜓𝑠 𝛽 𝑑𝐿 𝑖 (𝑞)
𝐶𝑖 (𝑞, 𝐼𝑖 ) = 2 2 2
, 𝑖 = 1, 2, 3. (6) q (rad)
1 + 𝛽 𝐿 𝑖 (𝑞) 𝐼𝑖 𝑑𝑞 m3+, dL3(q)/dq
1
Finally, some important properties of the Euclidean and 0.5
0
spectral norms are −0.5
−1
5 5.2 5.4 5.6 5.8 6 6.2
±𝑦𝑇 𝐷 (𝑥) 𝑤 ≤ ‖𝐷 (𝑥)‖ 𝑦 ‖𝑤‖ ,
q (rad)
2
𝑦𝑇 𝐸 (𝑥) 𝑦 ≥ 𝜆 𝑚 (𝐸 (𝑥)) 𝑦 , Figure 2: A graphical example of functions 𝑚𝑖 (𝑞) and 𝑑𝐿 𝑖 (𝑞)/𝑑𝑞
for a three-phase and four-pole SRM. The latter are represented by
2
𝑦𝑇 𝐸 (𝑥) 𝑦 ≤ 𝜆 𝑀 (𝐸 (𝑥)) 𝑦 , sinusoidal waveform signals and the former are the other waveform
(7) signals.
𝐷 (𝑥) 𝑦 ≤ ‖𝐷 (𝑥)‖ 𝑦 ,
𝑇
𝑦 𝑤 ≤ 𝑦 ‖𝑤‖ ,
Define 𝑚𝑖+ (𝑞) and 𝑚𝑖− (𝑞) as bounded functions with bounded
‖𝐷 (𝑥) 𝐺 (𝑥)‖ ≤ ‖𝐷 (𝑥)‖ ‖𝐺 (𝑥)‖ , and continuous first derivatives such that
N 1 − cos (𝜔𝑓 𝑇∗ )
= 2𝑇∗ , (17)
𝜔𝑓 sin (𝜔𝑓 𝑇∗ )
s
which can be solved numerically; 𝛼𝑓 > 0 is chosen as
−
√𝑇∗
Ii∗ − Ii 𝛼𝑓 = (18)
1 − cos (𝜔𝑓 𝑇∗ )
s
and 𝑇∗ is an arbitrarily small positive constant. Finally, we have
−N 𝛽1 = 1,
𝛽2 = 0,
if 𝜁𝑖 > 𝑇∗ ,
Figure 3: Hysteresis nonlinearity. Slop 𝑠 is assumed to be large but (19)
finite. 𝛽1 = 0,
𝛽2 = 1,
if 𝜁𝑖 ≤ 𝑇∗ .
4. Main Result
There always exist positive scalars 𝐾𝑝 , 𝐾𝑖 , 𝑁, 𝛿, and 𝛼, such
The following proposition establishes our main result. that the whole state (which includes the velocity error 𝜔 ̃ =
𝜔 − 𝜔∗ ) remains bounded and it has an ultimate bound which
Proposition 1. Consider model (2) and (3), in closed loop with can be rendered arbitrarily small by using a suitable choice
the following controller: of controller gains. This result stands when starting from any
initial condition within a ball whose radius can be arbitrarily
enlarged provided that suitable controller gains are chosen.
𝑈 = 𝐻 (𝐼∗ − 𝐼) − 𝛼𝜉 − 𝑘1 |𝜔| 𝜉 + 𝐶 (𝑞, 𝐼) 𝐼∗ 𝜔,
(12) Remark 2. In Figure 4 we depict control scheme in
𝑇
𝐼∗ = [𝐼1∗ , 𝐼2∗ , 𝐼3∗ ] , Proposition 1. Note that it is composed of a classical
proportional-integral (PI) external loop intended to regulate
𝑑𝐿 𝑖 (𝑞)
{ velocity when motor is submitted to unknown but constant
{𝛽1 √𝜁𝑖 + 𝛽2 𝑓 (𝜁𝑖 ) , if 𝑑𝑞 ≠ 0
{
𝐼𝑖∗ ={ external torque disturbances. An inner electric current loop
{
{0, 𝑑𝐿 𝑖 (𝑞) (13) is driven by both a hysteresis controller and a proportional
if = 0,
{ 𝑑𝑞 controller. The former is intended to dominate some
terms which do not vanish as the desired equilibrium
𝑖 = 1, 2, 3, point is reached. The latter is intended to dominate some
cross terms existing between the mechanical and the
exp (2𝛽𝐿2𝑖 (𝑞) 𝑚𝑖 (𝑞) 𝜏∗ /𝜓𝑠 (𝑑𝐿 𝑖 (𝑞) /𝑑𝑞)) − 1 electrical subsystems. This will become clear in the proof of
𝜁𝑖 = ,
𝛽2 𝐿2𝑖 (𝑞) Proposition 1 which is presented next.
(14) We stress that both hysteresis and proportional con-
𝜓 𝛽 𝑑𝐿 (𝑞)
𝑘1 > 𝑠 , trollers are commonly used in practice [11] which shows
2 𝑑𝑞 𝑀 feasibility of the proposed controller. This is important to
stress because this means that insight into how the proposed
𝑡
control scheme can be implemented in practical applications
𝜏∗ = −𝐾𝑝 𝜔
̃ − 𝐾𝑖 ∫ 𝜔
̃ (𝑡1 ) 𝑑𝑡1 , (15)
0 is explained by standard literature on SRM. The main advan-
tage of our approach is that, contrary to standard open-loop
operation of SRM, the trigger angles for switching on or
where 𝜔 ̃ = 𝜔 − 𝜔∗ , with 𝜔∗ the constant desired velocity, 𝜉 = switching off the power devices are determined automatically,
∗ ∗
𝐼−𝐼 , 𝜏 represents the desired torque, and symbol ‖⋅‖𝑀 stands thus simplifying motor operation at high and low velocities.
for the supreme value over the norm. We define the vectorial Finally, the nonlinear terms −𝑘1 |𝜔|𝜉 and 𝐶(𝑞, 𝐼)𝐼∗ 𝜔 are added
function 𝐻(𝐼∗ − 𝐼) = [ℎ(𝐼1∗ − 𝐼1 ), ℎ(𝐼2∗ − 𝐼2 ), ℎ(𝐼3∗ − 𝐼3 )]𝑇 , where at the output of the electric current controllers just for
ℎ(⋅) stands for the hysteresis nonlinearity depicted in Figure 3. stability proof purposes.
Functions 𝑚𝑖 (𝑞) are defined in (11) and
Remark 3. In [19] it is shown that function defined in (16)
∗ and the choice of parameters described after this expression
𝑓 (𝜁𝑖 ) = 𝛼𝑓 (1 − cos (𝜔𝑓 𝜁𝑖 )) , ∀𝜁𝑖 ≤ 𝑇 , (16) ensure that by replacing √𝜁𝑖 with 𝑓(𝜁𝑖 ), ∀𝜁𝑖 ≤ 𝑇∗ , we obtain
Mathematical Problems in Engineering 5
k1 i || L
− −
+ ∗ Ii∗ + + + Ui Ii
∗ PI (13) Electrical
(4) Mechanical
subsystem
subsystem +
− − + +
Ci (q, Ii ) Ii∗
respectively. On the other hand, adding and subtracting term where 𝐼 = [𝐼1 , 𝐼2 , 𝐼3 ]𝑇 with 𝐼𝑖 being a point belonging to
∑3𝑖=1 ((𝜓𝑠 /2𝛽𝐿2𝑖 (𝑞))(𝑑𝐿 𝑖 (𝑞)/𝑑𝑞)) ln(1 + 𝛽2 𝐿2𝑖 (𝑞)𝐼𝑖∗2 ) in (4) and line joining 𝐼𝑖 and 𝐼𝑖∗ . Note that there always exists a positive
replacing 𝐼𝑖∗ from (13), we have constant 𝐾𝐶𝐼 such that
𝑇
3
𝜓𝑠 𝑑𝐿 𝑖 (𝑞) 𝐼 𝐶 (𝑞, 𝐼) ≤ 𝐾 , ∀𝑞 ∈ R, 𝐼 ∈ R3 . (28)
𝜏=∑ 2
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖2 ) 𝐶𝐼
𝑖=1 2𝛽𝐿 𝑖 (𝑞) 𝑑𝑞
3
𝜓𝑠 𝑑𝐿 𝑖 (𝑞) 5. Stability Analysis
−∑ 2
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖∗2 )
𝑖=1 2𝛽𝐿 𝑖 (𝑞) 𝑑𝑞 In this section we present the proof of Proposition 1 employ-
∗ (22) ing Lyapunov stability analysis. First, a positive definite and
+ 𝜏 + 𝛽2 𝜙,
decrescent function is introduced in Section 5.1. Then, time
3 derivative of this function along the closed loop system
𝜓𝑠 𝑑𝐿 𝑖 (𝑞)
𝜙=∑ 2 (𝑞)
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝑓2 (𝜁𝑖 )) trajectories is computed in Section 5.2. There, it is also shown
𝑖=1 2𝛽𝐿 𝑖 𝑑𝑞
that the domain of attraction can be enlarged arbitrarily
− 𝜏∗ , by suitably selecting the controller gains. Finally, the above
findings are exploited in Section 5.3 in order to complete the
where |𝜙| ≤ 𝜙 for some 𝜙 which decreases to zero as 𝑇∗ > proof of Proposition 1 by invoking Theorem 4.18, pp. 172, in
0 approaches zero (see [19], Remark 1). Replacing (15), last [21]. Hence, the basic idea behind this procedure is simple.
expression can be written as However, as stressed in Remark 4, the algebraic manipulation
becomes elaborated because we do not feed back the complex
3
𝜓𝑠 𝑑𝐿 𝑖 (𝑞) term 𝐼∗̇ and we have to show how these noncancelled
𝜏=∑ 2 (𝑞)
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖2 ) terms are dominated. Moreover, showing that the domain of
𝑖=1 2𝛽𝐿 𝑖 𝑑𝑞
attraction can be arbitrarily enlarged introduces additional
3 algebraic steps.
𝜓𝑠 𝑑𝐿 𝑖 (𝑞)
−∑ 2
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖∗2 ) (23)
𝑖=1 2𝛽𝐿 𝑖 (𝑞) 𝑑𝑞
5.1. A Positive Definite and Decrescent Function. The scalar
𝑡 function
̃ − 𝐾𝑖 ∫ 𝜔
− 𝐾𝑝 𝜔 ̃ (𝑡1 ) 𝑑𝑡1 + 𝛽2 𝜙.
0 1 2 𝐾 𝛾
̃ 𝑧, 𝜉𝑇 ) = 𝐽𝜔
𝑉 (𝜔, ̃ + 𝑖 𝑧2 + (𝑏 + 𝐾𝑝 ) 𝑧2
̃ + 𝛾𝐽𝑧𝜔
Hence, from (3) it is obtained that 2 2 2
(29)
1
̃̇ + 𝑏𝜔
𝐽𝜔 ̃ − 𝐾𝑖 𝑧 + 𝛽2 𝜙,
̃ = Φ − 𝐾𝑝 𝜔 (24) + 𝜉𝑇 𝐿 (𝑞) 𝜉,
2
where
where 𝛾 > 0 is a constant and 𝐿(𝑞) = diag{𝐿 1 (𝑞), 𝐿 2 (𝑞),
3 𝐿 3 (𝑞)}, can be bounded as
𝜓𝑠 𝑑𝐿 𝑖 (𝑞)
Φ=∑ 2
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖2 )
2𝛽𝐿 (𝑞) 𝑑𝑞
𝑖=1 𝑖
(25) 𝑐1 ‖𝑥‖2 ≤ 𝑉1 (𝜔,
̃ 𝑧, 𝜉) ≤ 𝑐2 ‖𝑥‖2 ,
3
𝜓𝑠 𝑑𝐿 𝑖 (𝑞) 𝑇
−∑ 2 (𝑞)
ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖∗2 ) , ̃ 𝑧, 𝜉𝑇 ] ,
𝑥 = [𝜔,
𝑖=1 2𝛽𝐿 𝑖 𝑑𝑞 (30)
𝐽 𝛾𝐽
𝑡
𝑏𝜔∗ + 𝜏𝐿 𝑃=[ ],
̃ (𝑡1 ) 𝑑𝑡1 +
𝑧=∫ 𝜔 . (26) 𝛾𝐽 𝐾𝑖
0 𝐾𝑖
Thus, the closed loop dynamics is given by (20), (24), and 1
𝑐1 = min {𝜆 𝑚 (𝑃) , 𝐿 𝑚1 } > 0,
(26). Finally, let us say that using the mean value theorem we 2
can write 1
𝑐2 = max {𝜆 𝑀 (𝑃) + 𝛾 (𝑏 + 𝐾𝑝 ) , 𝐿 𝑀2 } > 0,
3
𝜓𝑠 𝑑𝐿 𝑖 (𝑞) 2
Φ=∑ 2
[ln (1 + 𝛽2 𝐿2𝑖 (𝑞) 𝐼𝑖2 ) (31)
𝑖=1 2𝛽𝐿 𝑖 (𝑞) 𝑑𝑞 𝐿 𝑚1 = inf {𝜆 𝑚 (𝐿 (𝑞))} ,
𝑞∈𝑅
𝑉(𝑥), defined in (29), along the trajectories of the closed loop where Δ 𝑖 is the 𝑖−th diagonal entry of diagonal matrix Δ,
system (20), (24), and (26), is given as since, according to Figure 3, we have ℎ(𝐼𝑖∗ − 𝐼𝑖 ) = 𝑁 sign(𝐼𝑖∗ −
𝑇
𝐼𝑖 ) = −𝑁 sign(𝜉𝑖 ), if |𝜉𝑖 | > 𝛿; that is, 𝜉𝑖 ℎ(𝐼𝑖∗ − 𝐼𝑖 ) = −𝑁|𝜉𝑖 |, if
𝑉̇ = − (𝑏 + 𝐾𝑝 − 𝛾𝐽) 𝜔
̃ 2 + (𝜔
̃ + 𝛾𝑧) 𝐼 𝐶 (𝑞, 𝐼) 𝜉 |𝜉𝑖 | > 𝛿, and 𝜉𝑖 ℎ(𝐼𝑖∗ − 𝐼𝑖 ) ≤ 𝑁𝛿, for all 𝜉𝑖 ∈ R. Also note that,
according to the global Lipschitz property of right hand of
1 𝑑𝐿 (𝑞) (24), expressions for 𝐼𝑖∗̇ in (20), and the fact that 𝜎𝑖 , 𝑖 = 1, 2, 3,
̃ + 𝛾𝑧) 𝜙 − 𝛾𝐾𝑖 𝑧2 − 𝜉𝑇
+ 𝛽2 (𝜔 𝜉𝜔
2 𝑑𝑞 (32) depend on 𝜏∗ which is an unbounded function of 𝜔 ̃ and 𝑧,
there is a positive continuous function 𝜂(𝑟1 ) which grows as
𝑇 ∗ ∗
+ 𝜉 Δ [−𝑅𝜉 − 𝑅𝐼 − 𝛼𝜉 + 𝐻 (𝐼 − 𝐼) − 𝑘1 |𝜔| 𝜉 𝑟1 = √𝜔 ̃ 2 + 𝑧2 grows, such that we can write
̇ ≤ 𝜂 (𝑟1 ) 𝜉 [|𝜔|
−𝜉𝑇 𝐿 (𝑞) 𝐼𝐴𝐵
∗
− 𝐷 (𝑞, 𝐼) 𝐼∗̇ ] , ̃ + |𝑧| + 𝜉 + 𝛽2 𝜙] . (39)
Thus,
where Δ is a diagonal positive definite matrix given as
𝑉̇ ≤ −𝑦𝑇 𝑄𝑦 + 6 max {Δ 𝑖 } 𝑁𝛿
1 + 𝛽2 𝐿21 (𝑞) 𝐼12 𝑖
(40)
Δ = 𝐿 (𝑞) 𝐷−1 (𝑞, 𝐼) = diag { ,
𝛽𝜓𝑠 + 𝛽2 (1 + 𝛾 + 𝜂 (𝑟1 )) 𝜙 𝑦 ,
(33)
1 + 𝛽2 𝐿22 (𝑞) 𝐼22 1 + 𝛽2 𝐿23 (𝑞) 𝐼32 ̃ , |𝑧| , 𝜉] .
𝑇
, }. 𝑦 = [|𝜔| (41)
𝛽𝜓𝑠 𝛽𝜓𝑠 Entries of matrix 𝑄 are given as
Replacing 𝑘1 from (14), we can write (32) as 𝑄11 = (𝑏 + 𝐾𝑝 − 𝛾𝐽) ,
𝑇
𝑉̇ ≤ − (𝑏 + 𝐾𝑝 − 𝛾𝐽) 𝜔
̃ 2 + (𝜔
̃ + 𝛾𝑧) 𝐼 𝐶 (𝑞, 𝐼) 𝜉 𝑄22 = 𝛾𝐾𝑖 ,
̃ + 𝛾𝑧) 𝜙 − 𝛾𝐾𝑖 𝑧2 − 𝜉𝑇 Δ (𝑟 + 𝛼) 𝜉
+ 𝛽2 (𝜔 (34) 𝑟+𝛼
𝑄33 = − 𝜂 (𝑟1 ) ,
𝛽𝜓𝑠
+ 𝜉𝑇 Δ [−𝑅𝐼∗ + 𝐻 (𝐼∗ − 𝐼) − 𝐷 (𝑞, 𝐼) 𝐼∗̇ ] . (42)
𝑄12 = 𝑄21 = 0,
Moreover, we can write
1 1
𝑄13 = 𝑄31 = − 𝐾𝐶𝐼 − 𝜂 (𝑟1 ) ,
−𝜉𝑇 Δ𝐷 (𝑞, 𝐼) 𝐼∗̇ = −𝜉𝑇 Δ𝐷 (𝑞, 𝐼) 𝐼𝐴𝐵
̇
∗
2 2
̇ , 𝛾 1
− 𝜉𝑇 Δ𝐷 (𝑞, 𝐼) 𝐼𝐶𝐷
∗
𝑄23 = 𝑄32 = − 𝐾𝐶𝐼 − 𝜂 (𝑟1 ) .
2 2
𝑇 (35)
̇ = [𝐼∗̇ + 𝐼∗̇ , 𝐼∗̇ + 𝐼∗̇ , 𝐼∗̇ + 𝐼∗̇ ] ,
∗
𝐼𝐴𝐵 Matrix 𝑄 is positive definite if and only if all of its principal
1𝑎 1𝑏 2𝑎 2𝑏 3𝑎 3𝑏
minors are positive; that is,
̇ = [𝐼∗̇ + 𝐼∗̇ , 𝐼∗̇ + 𝐼∗̇ , 𝐼∗̇ + 𝐼∗̇ ]𝑇 .
∗
𝐼𝐶𝐷 𝑏 + 𝐾𝑝 − 𝛾𝐽 > 0,
1𝑐 1𝑑 2𝑐 2𝑑 3𝑐 3𝑑
(43)
Hence, 𝑑2 = (𝑏 + 𝐾𝑝 − 𝛾𝐽) 𝛾𝐾𝑖 > 0,
𝑉̇ ≤ − (𝑏 + 𝐾𝑝 − 𝛾𝐽) 𝜔 ̃ + 𝛾 |𝑧|) 𝐾𝐶𝐼 𝜉
̃ 2 + (|𝜔| det (𝑄) = 𝑑2 𝑄33 + 𝑄13 (𝑄21 𝑄32 − 𝑄31 𝑄22 )
1 (44)
̃ + 𝛾 |𝑧|) 𝜙 − 𝛾𝐾𝑖 𝑧2 −
+ 𝛽2 (|𝜔|
2
(𝑟 + 𝛼) 𝜉 − 𝑄23 (𝑄11 𝑄32 − 𝑄12 𝑄31 ) > 0.
𝛽𝜓𝑠 (36) Assume for a moment that 𝜆 𝑚 (𝑄) > 0. Using some constant
− 𝜉 𝐿 (𝑞) 𝐼 ̇
𝑇 ∗
𝐴𝐵
0 < Θ < 1, we can rewrite (40) as
𝑉̇ ≤ −𝜆 𝑚 {𝑄} ‖𝑥‖2 + 6 max {Δ 𝑖 } 𝑁𝛿 + 𝛽2 (1 + 𝛾 + 𝜂 (𝑟1 )) 𝜙 ‖𝑥‖ ,
𝑇 ̇ ].
+ 𝜉 Δ [−𝑅𝐼∗ + 𝐻 (𝐼∗ − 𝐼) − 𝐷 (𝑞, 𝐼) 𝐼𝐶𝐷
∗ 𝑖
∗̇ + 𝛽2 (1 + 𝛾 + 𝜂 (𝑟1 )) 𝜙 ‖𝑥‖ ,
𝑁 > 𝑟𝐼𝑖∗ + 𝛽𝜓𝑠 𝐿 𝑖 (𝑞) 𝐼𝑖𝐶𝐷 , 𝑖 = 1, 2, 3, (45)
(37)
2 2
≤ −𝑐3 (1 − Θ) 𝜆 𝑚 {𝑄} ‖𝑥‖ − 𝑐4 (1 − Θ) 𝜆 𝑚 {𝑄} ‖𝑥‖
then
𝛽2 (1 + 𝛾 + 𝜂 (𝑟1 )) 𝜙
+ 6 max {Δ 𝑖 } 𝑁𝛿, ∀ ‖𝑥‖ ≥ ,
̇ ]
𝜉𝑇 Δ [−𝑅𝐼∗ + 𝐻 (𝐼∗ − 𝐼) − 𝐷 (𝑞, 𝐼) 𝐼𝐶𝐷
∗ 𝑖 Θ𝜆 𝑚 {𝑄}
Expressions in (43) are always satisfied using large enough 𝛾 + 𝜂(𝑟1 ))𝜙/Θ𝜆 𝑚 {𝑄} can be rendered arbitrarily small by
values for 𝐾𝑝 > 0 and 𝐾𝑖 > 0 and a small 𝛾 > 0. On the other choosing a small enough 𝑇∗ . Thus, 𝜇0 can always be rendered
hand, we remark that any singularity in 𝐼𝑖𝑎∗̇ and 𝐼𝑖𝑐∗̇ as 𝜁𝑖 → arbitrarily small. Since 𝛼1−1 (𝛼2 (⋅)) is a K∞ function, then the
0, 𝑖 = 1, 2, 3, is avoided if 𝑇∗ > 0 is chosen. Define 𝜇 as the ultimate bound in (51) tends to zero as 𝑇∗ > 0 and 𝛿 > 0
radius of an open ball 𝐷 = {𝑥 ∈ R5 | ‖𝑥‖ < 𝜇} around the ori- approach zero.
gin of the state space where 𝜆 𝑚 (𝑄) > 0. The simplest estimate Note that, as was explained in paragraph after (46),
of the domain of attraction Ω𝑐 ⊂ 𝐷 is given as [21], pp. 317, conditions (43) and (44) can be satisfied by finite controller
gains only if 𝑇∗ > 0, that is, to keep (46) valid. This means
that the closed loop system has an ultimate bound which
Ω𝑐 = {𝑥 ∈ R5 | 𝑉 (𝑥) < 𝑐 = min 𝑉 (𝑥)} . (47)
‖𝑥‖=𝜇 cannot be reduced to zero but can be rendered arbitrarily
small by a suitable choice of controller gains. Recall that
Hence, according to (30), we can compute 𝑐 = min‖𝑥‖=𝜇 𝑉(𝑥) = this result stands semiglobally. This completes the proof of
𝑐1 𝜇2 . A more conservative estimate of the domain of Proposition 1. Finally, we emphasize that the conditions to
attraction Ω𝑐 , represented by a ball of radius 𝑟2 , can be guarantee Proposition 1 are summarized by 0 < 𝛾 < √𝐾𝑖 /𝐽,
found by bounding 𝑉(𝑥) ≤ 𝑐2 ‖𝑥‖2 < 𝑐1 𝜇2 , which results in (37), (43), (44), and some small constants 𝑇∗ > 0, 𝛿 > 0.
Ω𝑐 = {𝑥 ∈ R5 | 𝑐2 ‖𝑥‖2 < 𝑐1 𝜇2 }. Thus, this new estimate of
the domain of attraction contains the set Remark 5. Note that use of the desired electric current
𝑇
5 𝑐 𝐼∗ = [𝐼1∗ , 𝐼2∗ , 𝐼3∗ ] ,
Ω𝑐 = {𝑥 ∈ R | ‖𝑥‖ < 𝑟2 = √ 1 𝜇} . (48)
𝑐2 𝑑𝐿 𝑖 (𝑞)
{
{ √𝜁𝑖 , if ≠ 0
{ 𝑑𝑞
(52)
Replace 𝜂(𝑟1 ) with 𝜂(𝜇) in (44). From the resulting expression 𝐼𝑖∗ = {
{
{0, 𝑑𝐿 𝑖 (𝑞)
we obtain if = 0,
{ 𝑑𝑞
𝑟+𝛼
𝑑2 exp (2𝛽𝐿2𝑖 (𝑞) 𝑚𝑖 (𝑞) 𝜏∗ /𝜓𝑠 (𝑑𝐿 𝑖 (𝑞) /𝑑𝑞)) − 1
𝛽𝜓𝑠
𝜁𝑖 = , (53)
> 𝑑2 𝜂 (𝑟1 ) 𝛽2 𝐿2𝑖 (𝑞)
(49)
− 𝑄13 (𝜂 (𝜇)) [𝑄21 𝑄32 (𝜂 (𝜇)) − 𝑄31 (𝜂 (𝜇)) 𝑄22 ] for 𝑖 = 1, 2, 3, renders the corresponding time derivative
unbounded when 𝜏∗ = 0 and 𝜏∗̇ ≠ 0, a common condition
+ 𝑄23 (𝜂 (𝜇)) [𝑄11 𝑄32 (𝜂 (𝜇)) − 𝑄12 𝑄31 (𝜂 (𝜇))] . when a change in the sense of rotation is commanded. This
can be seen from the product of the last term between
Since 𝑄33 is the only entry of 𝑄 depending on 𝛼 and 𝜂(𝜇) does brackets and factor at the left in the following expression:
not depend on 𝛼, it is clear that we can enlarge arbitrarily 𝜇
by enlarging 𝛼. Using this and the fact that 𝑐1 and 𝑐2 do not 1 exp (𝜎𝑖 ) − 1
−1/2
60
14
40 12
20 10
(rad/sec)
I1 (A)
0 6
−20 4
2
−40
0
−60 −2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
t (sec) t (sec)
∗
(a) Motor velocity and desired velocity (b) Electric current in phase 1
14 14
12 12
10 10
8 8
I2 (A)
I3 (A)
6 6
4 4
2 2
0 0
−2 −2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
t (sec) t (sec)
(c) Electric current in phase 2 (d) Electric current in phase 3
of the state such as 𝐼∗̇ which we only have to dominate. Aside experimentally identified by [8]. Sharing functions 𝑚𝑖 (𝑞)
from an important simplification of the control law this is were designed similarly as described by [10, 19]; that is, they
also instrumental to successfully present stability proof of are composed of a constant unitary value, a constant zero
Proposition 1. value, a raising polynomial 𝑝𝑟 (𝑥), and a falling polynomial
𝑝𝑓 (𝑥) = 1 − 𝑝𝑟 (𝑥). As the rising polynomial we used 𝑝𝑟 (𝑥) =
̃ 𝑧, 𝜉𝑇 ), introduced in (29),
Remark 7. Note that function 𝑉(𝜔, 35𝑥4 −84𝑥5 +70𝑥6 −20𝑥7 , 𝑥 = ℎ/𝑞𝑚𝑖 (see [10, 19] for definitions
contains the quadratic term (1/2)𝜉𝑇 𝐿(𝑞)𝜉 in order to cope of ℎ and 𝑞𝑚𝑖 and further details). This ensures that all of the
with the motor electric dynamics. This is a clever choice
requirements in Section 3 are satisfied. The controller gains
since 𝐿(𝑞) does not represent the motor inductance matrix
were set to 𝐾𝑝 = 0.6, 𝐾𝑖 = 20, 𝑘1 = 5, 𝛼 = 10, 𝑁 = 30,
but simplifies the stability proof. As a matter of fact, if
𝛿 = 0.02, 𝑇∗ = 0.1, 𝛼𝑓 = 0.1632, and 𝜔𝑓 = 27.85. All of the
the quadratic function (1/2)𝜉𝑇 𝐷(𝑞, 𝐼)𝜉 is employed, because
matrix 𝐷(𝑞, 𝐼) does represent the motor inductance matrix, initial conditions were set to zero.
when computing the time derivative of this function, the fact The desired velocity was designed as follows. A ramp
that 𝐷(𝑞, 𝐼) depends on 𝐼 complicates the stability analysis. takes the desired value from 0 [rad/s] at 𝑡 = 0 [s] to 50 [rad/s]
at 𝑡 = 0.15 [s]. It remains constant at 50 [rad/s] from 𝑡 =
0.15 [s] to 𝑡 = 0.4 [s]. Then, a ramp takes it from 50 [rad/s]
6. Simulation Results at 𝑡 = 0.4 [s] to −50 [rad/s] at 𝑡 = 0.7 [s]. Finally, the desired
In this section we present a simulation example intended velocity remains constant at −50 [rad/s] for 𝑡 ≥ 0.7 [s]. The
to give some insight into performance achievable with con- load torque was selected to be constant at 𝜏𝐿 = −4 [Nm] from
troller in Proposition 1. For this, we use the SRM whose 𝑡 = 1 [s] to 𝑡 = 1.4 [s] and 𝜏𝐿 = 0 elsewhere. This allows
numerical parameters are given by [10, 14] where it is assumed to study performance when step changes in the load torque
that 𝐿 𝑖 (𝑞) = 𝑙0 +𝑙1 cos(𝑁𝑟 𝑞−(𝑖−1)(2𝜋/3)): 𝑁𝑟 = 8, 𝑅 = 5 [Ω], are applied. We stress that the above specified load torque
𝑙0 = 0.03 [H], 𝑙1 = 0.02 [H], 𝐽 = 0.001 [kg-m2 ], and 𝑏 = is unknown for the controller and, despite this, its effect is
0.02 [Nm/(rad/s)]. Additionally, the flux model parameters successfully rejected by the integral part of the controller, as
were chosen as 𝜓𝑠 = 0.5 [Wb], 𝛽 = 1.8, which were observed in Figure 6. We consider that this feature represents
10 Mathematical Problems in Engineering
150 150
100 100
50 50
0 0
−50 −50
U1 (V)
U2 (V)
−100 −100
−150 −150
−200 −200
−250 −250
−300 −300
−350 −350
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
t (sec) t (sec)
(a) Applied voltage in phase 1 (b) Applied voltage in phase 2
150
100
50
0
−50
U3 (V)
−100
−150
−200
−250
−300
−350
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
t (sec)
(c) Applied voltage in phase 3
an important improvement with respect to results by [13, 14] this load torque smoothly passes through zero when velocity
where any information is not given on the load torque that changes direction. We realize that, despite this, any voltage
they have considered. Moreover, a zero load torque due to (or electric current) does not become large when velocity
friction is considered by [14]. On the other hand, although passes though zero. This proves that singularity referred to
a nonzero constant (for all time) load torque is reported in Remark 5 is not present in controller in Proposition 1.
in simulations presented by [12], performance during the We observe that large negative voltage spikes appear all
transient response cannot be evaluated from such a test. the time, especially when load torque is increased. These
The simulation results obtained when using controller in voltage spikes are naturally produced by the controller as
Proposition 1 are presented in Figures 6 and 7. We can see that a means of reducing to zero electric current through those
velocity 𝜔 and the desired velocity 𝜔∗ remain almost identical stator phases which were required to produce a nonzero
for most of the time. Important differences between these torque but, then, they are suddenly required not to produce
variables only appear when step changes in the load torque are torque. This happens when sign of 𝑑𝐿 𝑖 (𝑞)/𝑑𝑞 changes for
applied. However, such differences disappear very fast which those phases, which would generate torque in the opposite
proves the good disturbance rejection properties of controller direction if electric currents through such phases were not
in Proposition 1. We stress that a zero steady state velocity forced to be zero very fast. SRM are provided with electronic
error is achieved when the desired velocity is constant drivers which are able to provide such negative voltage values
although Proposition 1 only ensures an ultimate bound in as described by [22]. These voltage spikes are also present in
velocity error. Reason for this is that the presence of friction [14] although the voltage spikes in Figure 7 of the present
and a load torque ensures that |𝜁𝑖 | > 𝑇∗ > 0 in steady state. paper are larger. This is because we are using a large load
In Figures 6 and 7, we also present the applied voltages torque 𝜏𝐿 = −4 [Nm] (load torque in [14] is not specified).
and electric currents through all the three motor phases. As a second test, in Figures 8, 9, and 10 we present
Note that these variables have important changes when step some simulation results when two-step desired velocities
changes in the load torque appear and when the desired are commanded. The first one is applied at 𝑡 = 0 and
velocity is described by ramps. We stress that the desired changes desired velocity from 0 to 25 [rad/s] and the second
torque 𝜏∗ becomes zero at some point of time which is close one changes desired velocity from 25 [rad/s] to 0 at 𝑡 =
to time when velocity is zero. This is because the only load 0.9 [s]. As expected, large electric current peaks appear
torque applied before 𝑡 = 1 [s] is that due to friction, and which are required to accelerate and deaccelerate motor in
Mathematical Problems in Engineering 11
30 25
25
20
20
15
15
10
10
5
5
0 0
−5 −5
−10 −10
−15 −15
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0.89 0.895 0.9 0.905 0.91 0.915 0.92
t (sec) t (sec)
(a) Motor velocity and generated torque (b) Motor velocity and generated torque (zoom)
Figure 8: Additional simulation results when using controller in Proposition 1. 𝜔 is in rad/s and 𝜏 is in Nm.
25 5
4.5
20 4
3.5
15 3
I1 (A)
I2 (A)
2.5
10
2
1.5
5
1
0 0.5
0
−5 −0.5
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
t (sec) t (sec)
(a) Electric current in phase 1 (b) Electric current in phase 2
16
14
12
10
8
I3 (A)
6
4
2
0
−2
0 0.2 0.4 0.6 0.8 1 1.2
t (sec)
(c) Electric current in phase 3
order to follow the commanded velocity changes. Note that must be rendered zero very fast in phases which, according to
these current spikes are always positive and only appear in sharing functions, must not contribute to generated torque. If
certain motor phases, that is, those phases which, according electric current is not zero in such phases then those phases
to sharing functions, are able to produce suitable torque would contribute with a torque in the opposite direction more
according to the motor position angle. Notice that voltage than required. A positive electric current is rendered zero
may be negative. This is due to the fact that electric current very fast by applying a large negative voltage.
12 Mathematical Problems in Engineering
40 40
20 20
0 0
U1 (V)
U2 (V)
−20 −20
−40 −40
−60 −60
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
t (sec) t (sec)
(a) Voltage applied at phase 1 (b) Voltage applied at phase 2
40
20
0
U3 (V)
−20
−40
−60
Figure 10: Additional simulation results when using controller in Proposition 1 (cont.).
7. Conclusions References
In this paper we have presented a solution for the velocity [1] M. M. Bridges and D. M. Dawson, “Adaptive control of rigid-
regulation in SRM when taking into account magnetic flux link electrically-driven robots actuated with switched reluc-
saturation. The main feature of our proposal is that it includes tance motors,” in Proceedings of the 1995 American Control
an internal electric current loop driven by a controller by Conference. Part 1 (of 6), pp. 1392–1396, June 1995.
hysteresis, which is the common scheme in practice to control [2] D. G. Taylor, “Pulse-width modulated control of electrome-
electric current in SRM. We prove, for the first time, ultimate chanical systems,” Institute of Electrical and Electronics Engi-
boundedness of the state. Moreover, the ultimate bound can neers Transactions on Automatic Control, vol. 37, no. 4, pp. 524–
be rendered arbitrarily small and this result stands when 528, 1992.
starting from any initial condition within a ball whose radius [3] R. S. Wallace and D. G. Taylor, “Low-Torque-Ripple Switched
can be arbitrarily enlarged by choosing suitable controller Reluctance Motors for Direct-Drive Robotics,” IEEE Transac-
gains. It is observed in a simulation study that the velocity tions on Robotics and Automation, vol. 7, no. 6, pp. 733–742, 1991.
error converges to zero if the desired torque in steady state [4] M. Ilic’-Spong, R. Marino, S. M. Peresada, and D. G. Taylor,
is not zero. It is also shown that a good performance is “Feedback Linearizing Control of Switched Reluctance Motors,”
IEEE Transactions on Automatic Control, vol. 32, no. 5, pp. 371–
obtained when rejecting unknown step changes in the load
379, 1987.
torque. We conclude that good performance of the proposed
[5] M. De Queiroz, T. Burg, D. Dawson, and S. Donepudi, “A partial
control scheme is achieved if we compare the obtained results
state feedback controller for srm-rled robot manipulators:
with those presented previously in the literature which have
preliminary experimental validation,” in Proceeding of the IEEE
been referred to in the simulations section. Moreover, it is Conference on Control Applications, pp. 916–921, 1995.
important to stress that performance of our proposal has [6] R. Milman and S. A. Bortoff, “Observer-based adaptive control
been achieved despite the remarkable simplicity of the control of a variable reluctance motor: Experimental results,” IEEE
law when compared to previous controllers proposed in the Transactions on Control Systems Technology, vol. 7, no. 5, pp.
literature. 613–621, 1999.
[7] D. M. Dawson, J. Hu, and T. C. Burg, Nonlinear Control of
Conflicts of Interest Electric Machinery, Marcel Dekker, New York, NY, USA, 1998.
[8] P. Vedagarbha, D. M. Dawson, and W. Rhodes, “An adaptive
The authors declare that there are no conflicts of interest controller for a general class of switched reluctance motor
regarding the publication of this paper. models,” Automatica, vol. 33, no. 9, pp. 1647–1655, 1997.
Mathematical Problems in Engineering 13
International
Journal of Journal of
Mathematics and
Mathematical
#HRBQDSDĮ,@SGDL@SHBR
Sciences