Applied Mathematical Modelling 51 (2017) 686–697

Contents lists available at ScienceDirect

Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm

Energy cycle of brushless DC motor chaotic system

Guoyuan Qi
Tianjin Key Laboratory of Advanced Technology of Electrical Engineering and Energy, Tianjin Polytechnic University, Tianjin 300387, China

a r t i c l e i n f o a b s t r a c t

Article history: The vector field of the brushless DC motor (BLDCM) chaotic system is regarded as the
Received 25 March 2017 force field of a pure mechanical system via the transformation of Kolmogorov system. The
Revised 13 July 2017
BLDCM force field is decomposed into four types of torque: inertial, internal, dissipative,
Accepted 17 July 2017
and generalized external torque. The forcing effect of each term in the force field is iden-
Available online 26 July 2017
tified via the analogue of the electrical and mechanical system. The BLDCM energy trans-
Keywords: formation of four forms of energy—kinetic, potential, dissipative, and generalized external
BLDCM is investigated. The physical interpretation of force decomposition and energy exchange
Bound is given. The rate of change of the Casimir energy is equivalent to the power exchanged
Casimir function between the dissipative energy and the energy supplied to the motor, and it governs the
Chaos different dynamic modes. A simple and optimal supremum bound for the chaotic attractor
Exchange power is proposed using the Casimir function and optimization.
Lie–Poisson bracket © 2017 Elsevier Inc. All rights reserved.

1. Introduction

Chaos applies not only to physical problems such as the trajectories of objects in turbulent media, but also to problems
in biology, mathematics, chemistry, engineering, medicine, astronomy [1], music, business, geoscience and environmental
science.
The existence of chaos in motor systems was first discovered by Kuroe in 1989 [2]. The mathematical model for a
permanent magnet synchronous motor was first derived fitting for analysis of chaos and bifurcation [3]. The brushless DC
Motor (BLDCM) is a type of synchronous permanent-magnet motor. In recent years, the BLDCM has achieved a brilliant
expansion in the automotive, aerospace and household-appliance industries, robotics, food and chemical industries, electric
vehicles, medical instruments, and computer peripherals [4,5]. The BLDCM is noted for its high efficiency, long life, low
noise, and good speed-torque characteristics [4].
The BLDCM chaotic system was found in 1994 by Hemati [6]. In regard to performance, the occurrence of chaos in
motors is highly undesirable in most engineering applications [7]. Thus, studies of chaos control have been performed on
several types of motor drive systems [8,9]. The synchronization of the BLDCM chaotic system was also studied [10].
Regarding the analysis of chaotic models, the main research focus is on their dynamical analysis. Although some of the
literature has reported on the application of chaotic systems in meteorology [11] and celestial chaos analog [1], most re-
search has focused on the dynamics of the Lorenz system. The research themes usually are numerical calculation, aperiodic
solutions, sensitivity to initial conditions, bifurcation theory [3,12], power spectra, circuit implementation, fractional order
[13], chaos-based communication [14], chaos control and synchronization [15], generation of chaos [16], and existence of
chaos [17]. However, these aspects of research aforementioned cannot reveal the mechanism or reason for the production of
dynamic modes. To explore this aspect, the mechanics of chaotic systems must be investigated. The lines of study include

E-mail addresses: guoyuanqisa@qq.com, qiguoyuan@tjpu.edu.cn

http://dx.doi.org/10.1016/j.apm.2017.07.025
0307-904X/© 2017 Elsevier Inc. All rights reserved.
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 687

force analysis and the energy transformation between internal energy and supplied energy. If chaos is investigated within
mechanics, more of the fundamentals can be uncovered.
Arnold [18] presented a Kolmogorov system describing a dissipative-forced dynamic system or hydro-dynamic insta-
bility with a Hamiltonian function. For instance, in geoscience and environmental sciences, especially fluid dynamics,
the Navier–Stokes equations (the Galerkin approximation) can be simplified as a Kolmogorov system in fluid dynamics
[19]. Pasini and Pelino gave a unified view of the Kolmogorov and Lorenz systems [20,21], thereby providing the forcing
analysis of the Lorenz system. Qi and Liang studied the mechanics of a four-wing chaotic system [22] and a 3-D chaotic
system [23] through the transformation of Kolmogorov system to perform a forcing analysis. Furthermore, Qi and Zhang
gave the analysis of energy cycling and bound for the 3-D chaotic system [24]. In this regard, the Hamiltonian function
and the Kolmogorov system provide a starting point in studying the mechanism underlying these chaotic systems. The
Casimir function, like enstrophy or potential vorticity in the context of fluid dynamics, is very useful in analyzing stability
conditions and the global description of a dynamical system [25,26]. Both the four-wing chaotic system and the 3-D chaotic
system were built from numerical simulation instead of physical derivation [27,28]. Even if the Lorenz system describing the
atmospheric convection, but the Lorenz model was derived through a process [29]. However, the BLDCM model describes a
real physical process of electromagnetic motor [6,9]. Therefore, the mechanical research result of the BLDCM is applicable
and practical to the design and control of the system.
The number of parameters of BLDCM system is reduced for the Hemati model [6]. However, the modified parameters and
variables lose their physical dimensions. Hence research outcomes using this model cannot match accordingly the dynamics
of the original BLDCM system. This paper investigates the original physical model instead of the dimensionless one.
The presence of chaos in the BLDCM physically causes the system to oscillate and creates acoustic noise and mechanical
vibration, thereby consuming electrical energy and reducing service life. On the other hand, chaotic mixing has been
proposed to improve the energy efficiency and degree of homogeneity by using mechanical means that are essentially
based on the design of permanent-magnet DC motor which serves as the agitator to produce chaotic motion [30–32].
Therefore, the studying of chaos for the system helps to avoid the generation of chaos, and also to create chaotic regulation
so as to improve the configuration at the design stage.
In this paper, the original BLDCM is transformed into a Kolmogorov-type system, which decomposes the vector field of
the function into inertial, internal, dissipative, and external torque. The kinetic energy and potential energy of the system
are identified. We show that the rate of change of the Casimir energy is the exchange power between the dissipative energy
and the supplied energy, and the exchange power determines the dynamic modes of the BLDCM and the supremum bound
of the chaotic attractor. Different dynamical modes are revealed through combinations of energy.
Normally, it is difficult to find the bound of a chaotic attractor. Two important methods of finding the bound have been
developed: one is Lyapunov-based method finding a positive definite matrix solve in the Lyapunov stable equation [33],
another one is localization of compact invariant sets which is efficient to find ellipsoid bound [34,35]. However, we propose
a new and simpler method based on rate of change of the Casimir function.
The paper is organized as follows: The mechanics of the BLDCM chaotic system is analyzed in Section 2. The energy
cycling of the system and the boundary of the chaotic attractor are proposed in Section 3. Section 4 reveals the energy
cycling mechanism underlying the different dynamic modes. A conclusion is given in Section 5.

2. Mechanics of BLDCM chaotic system

The equations describing the non-salient-pole (or called round pole or smooth air gap) BLDCM can be written via a Park
transformation as [6,9]

Li˙q = −Riq − nLωid − nkt ω + uq ,

Li˙d = −Rid + nLωiq + ud ,
J ω˙ = nkt iq − bω − TL , (1)
where dot ‘.’ denotes the derivative with respect to time t; iq is the quadrature-axis current, id the direct-axis current, and ω
the rotor velocity; R is the winding resistance, La self-inductance of the winding with L = 32 La , n the number of permanent-

magnet pole pairs, ke the coefficient of the motor torque with kt = 3/2 ke , J the inertia of moment, b the damping coeffi-
cient of bearing; uq and ud are the voltage across quadrature-axis and direct-axis, and TL the external torque due to load.
To reduce the number of parameters and make the system like the Lorenz system, Hemati [6] developed a model
through a non-dimensionalization and obtaining the form [6]

x˜˙ 1 = −x˜2 x˜3 + ρ x˜3 − x˜1 + u˜q ,

x˜˙ 2 = x˜1 x˜3 − x˜2 + u˜d ,
x˜˙ 3 = σ (x˜1 − x˜3 ) − T˜L . (2)
Because of this non-dimensionalization, the parameters and variables lose their original physical dimension. The vari-
ables in (2) do not physically correspond to those in the original system (1). ρ is a free parameter losing physical meaning
in (2). σ = τJb is a combination of three parameters, where τ = RL denotes the reciprocal of damping ratio, bJ is the damping
688 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697

ratio, and therefore σ = τJb becomes a non-dimensional unit. System (1) is a model of a real BLDCM, but the Hemati model
is not. Hence outcomes from mechanical research on the Hemati model cannot directly help in the analysis and design of
the BLDCM.
To investigate the mechanics of BLDCM system, we start from original system (1) because it is derived from physical
process and each variable or parameter has its physical sense. Arnold [18] presented a Kolmogorov system describing
dissipative-forced dynamical systems given in the form

x˙ = {x, H } − x + f, (3)
where x = [x1 x2 x3 ]T ,
{ , } represents the Lie-algebraic structure for the Hamiltonian function of a system denoted by H,
 is a positive definite diagonal matrix, − x represents the dissipation, and f represents the generalized external torque
(or force). The Lie–Poisson bracket on functions of x takes the form (see Ref. [36])

{F , G}(x ) = −x · (∇ F × ∇ G ), (4)
where F, G ∈ C∞ (g∗ ) and g is a Lie algebra.

Remark 1.

(1) System x˙ = {x, H } is an Euler equation describing the dynamics of a free rigid body. Hence the Kolmogorov system is
a generalized Euler equation with dissipation and generalized external torque.
(2) The variable x is analogous to angular momentum.
(3) The torque {x, H} in the Euler equation for the free rigid body is the inertial torque.
(4) {x, H} satisfies skew-symmetry of the Lie–Poisson bracket.

To satisfy the skew-symmetric property of {x, H} for (1), we perform the transformation
√ √ 
x1 = Li q , x 2 = Li d , x 3 = J ω , (5)
which transforms system (1) to
R n nkt 1
x˙ 1 = − x1 −  x2 x3 −  x3 + √ uq ,
L J JL L
R n 1
x˙ 2 = − x2 +  x1 x3 + √ ud ,
L J L
nk b 1
x˙ 3 =  t x1 − x3 −  TL , (6)
JL J J
Because (6) can be regarded as a forced-dissipative rigid body, we set the principle moment of inertia as
  
J J J
[J1 , J2 , J3 ] = , , (7)
2n 2n n

and xi as the angular momentum of each axis of the rigid body satisfying xi = Ji ωi , where ωi is the angular velocity
corresponding to Ji . The kinetic part is expressed as (see Ref. [36])
   
1 x21 x2 x2 1 2nLi2q 2nLi2d nJ ω2
K= + 2 + 3 =  +  +  . (8)
2 J1 J2 J3 2 J J J

Setting the potential part to

nk nk
U=  t x2 =  t id , (9)
JL J
the Hamiltonian function is then

H = K + U. (10)
Eq. (6) can be written as a Kolmogorov system,

x˙ = {x, H } − x + f
= {x, K } + {x, U } − x + f
⎡ ⎤ ⎡ ⎤ ⎡R ⎤ ⎡ ⎤
−√
n
x2 x3 −√
nkt
x3 x √1 uq
J JL L 1 L
⎢ n ⎥ ⎢ ⎥ ⎢ R ⎥ ⎢ √1 ud ⎥
= ⎣ √J x1 x3 ⎦ + ⎣ 0 ⎦−⎣ x
L 2 ⎦+⎣ L ⎦, (11)
0 √
nkt
x1 b
x −√
1
TL
JL J 3 J
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 689

Table 1
Analogous quantities of electrical and mechanical systems [37].

Electrical quantity Mechanical translational analog (force voltage) Mechanical rotational analog (torque voltage)

Voltage, u Force, f Torque, T

Current, i Velocity, v Angular velocity, ω
Resistance, R Friction b Friction, b
Inductor, L Mass, m Moment of inertia, J
Flux, ϕ = Li Momentum, P = mv Angular momentum, x = Jω
u = Ri f = bv T = bω
u = ddtφ = L dt
di
f = dP
dt
= m ddtv T = dxdt
= J ddtω
i = C du
dt
v = K1 df
dt
ω = K1 dTdt
Consuming Power, ui = Ri2 Consuming, fv = bv2 Consuming, Tω = bω2
Inductor energy, 12 Li2 Spring energy, 12 mv2 Rotational energy, 12 J ω2

where
   T
R R b 1 1 1
 = diag , f= √ uq √ ud −  T .
L L J L L J

Similar to the BLDCM chaotic system in structure, the Lorenz system is also a Kolmogorov-type system with a Lie–Poisson
bracket (referring to [20]). Qi et al. studied the transformations from four-wing chaotic system and 3-D chaotic system
to Kolmogorov system [22–24]. Note that the investigated objects in these references are the transformed models. These
original systems in Refs. [22–24] are purely mathematically built and the Lorenz system in Ref. [20] is a dimensionless one,
therefore, both the original models and the transformed ones do not have physical sense, although the mechanical analysis
of these systems have been investigated. However, in this paper, Eq. (11), the transformation of Kolmogorov system, is just a
bridge to facilitate the mechanical analysis of the BLDCM. System (1) is always the object of study, in addition, the original
physical meaning are always interpreted for the analysis results in the whole paper.
Substituting (5) into (11), we have
⎡√ ⎤ √ ⎡ ⎤ ⎡ nkt ⎤ ⎡ √R ⎤ ⎡ 1 ⎤
Li˙q −n Lid ω − √L ω i
L q
√ uq
L
⎢√ ˙ ⎥ ⎢ √ ⎥ ⎢ 0 ⎥ ⎢ √R id ⎥ ⎢ √1 u ⎥
⎣ Li d ⎦ = ⎣ n Li q ω ⎦ + ⎣ ⎦ − ⎣ L ⎦ + ⎣ L d ⎦. (12)
 √
nkt √b ω −√1
TL
Jω˙ 0 iq
J J J

Eq. (12) is exactly equivalent to (1). By comparing (1), (11), and (12) for system (1), we interpret that:

Remark 2.

(1) Vector [Liq Lid Jω]T represents the generalized angular momentum, and [Li˙q Li˙d J ω˙ ]T the generalized reaction torque
or angular acceleration.
(2) Vector [−nLid ω nLiq ω 0]T in the quadratic term represents the inertial torque transferred from kinetic energy, and
[−nkt ω 0 nkt iq ]T the internal torque released from the potential energy; vector −[Riq Rid bω]T represents the dis-
sipative torque, and [uq ud −TL ]T the generalized external torque.

Remark 2 clarifies the force action of each term in the vector field of the BLDCM using the transformation between the
BLDCM and the Kolmogorov system. It can also be interpreted with respect to analogues of electrical and mechanical vari-
ables [37]. The analogous quantities are given in Table 1 [37]. System (1) is an electro-mechanical system, but it is analogous
to a pure mechanical system or a pure electrical circuit. From the sixth row in Table 1, the magnetic fluxes, Liq and Lid , are
analogues to angular momentum Jω, so from a pure mechanical perspective, [Liq Lid Jω]T represents the generalized angular
momentum. Term − nkt ω represents the back electromotive force acted by the motor rotation in windings, and nkt iq is the
electromagnetic torque acted upon by the current in the windings. Hence both the EMF and the electromagnetic torque are
the interaction results between the electrical and mechanical parts. Both are torques released from potential energy, U =
√ id . Therefore, vector [−nLid ω nLiq ω 0]T is called an internal torque. Vector −[Riq Rid bω]T is called a dissipative torque
nkt
J
(seventh row of Table 1). The supplied voltages, uq and ud , correspond to the external torque TL , so vector [uq ud −TL ]T is
called a generalized external torque (second row). The definition of kinetic energy in (8) has physical grounds (11th row).

3. Energy of BLDCM system

3.1. Energy representation of the BLDCM system

Normally, a type of force or torque corresponds to a form of energy. We manage to find out the energy forms involved
in the BLDCM in terms of the four types of torques.
690 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697

Define
1
Dp = x, x, (13)
2
as the dissipative power, i.e., the rate of change of dissipative energy, where  ,  is the inner product. Because x represents
the angular momentum, and x represents dissipative torque, the definition makes physical sense. Likewise, define
G p = x, f, (14)
as the generalized external power. Then Kolmogorov Eq. (11) can be written as
x˙ = {x, H } − ∇ D p + ∇ G p . (15)
Therefore, dissipative energy and generalized external energy are also involved in the Kolmogorov system.

Remark 3. The total energy of the chaotic BLCCM system contains four forms of energy: kinetic, potential, dissipative, and
generalized external, which correlate with the four types of torque: inertial, internal, dissipative and generalized external.
The four forms of energy circulate in a cycle, and the input energy is the generalized external energy and output energy is
the dissipative energy.

From Eqs. (13), (14) and (5), we have

1 1 
3
1
Dp = x, x = i x2i = Ri2q + Ri2d + bω2 , (16)
2 2 2
i=1

and

3
G p = x, f = fi xi = uq iq + ud id − TL ω. (17)
i=1

Eqs. (16) and (17) can be interpreted as follows:

As listed in the 10th row of Table 1, terms Ri2q , Ri2d , and bω2 are the loss of electrical power of the motor. Therefore
2D p = (Ri2q + Ri2d + bω2 ) is really the internal dissipative power. Term − TL ω related to another consuming mechanical power
due to load, called the loss in external power which differs from the internal dissipative power. Term iq uq + id ud is the
supplied power. Therefore, (17) contains both the supplied power and the external losing power.
From (4), we have
∇ K · {x, H } = {K, H }. (18)
From (15), we have
K˙ =∇ K, x˙ 
= ∇ K, ({x, H } − x + f )
= {K, H } − ∇ K, x + ∇ K, f. (19)
Therefore, the rate of change of the kinetic energy relates to its exchange with potential energy, dissipative energy, and
generalized external energy. Likewise, we have
U˙ = −{K, U } − ∇ U, x + ∇ U, f. (20)
If we assume that ud is constant, and uq = 0, TL = 0. The equilibria of Eq. (1) are obtained as follows:
 T  T
1 b √ kt bR √
S1 = 0, u , 0 , S2 = η, − − 2 , η ,
R d nkt L n Lkt
 T
b √ kt bR √
S3 = − η, − − 2 ,− η (21)
nkt L n Lkt
2 kt ud Rkt2
where η = −( nR2 L2 + bL
+ bL2
).

3.2. Casimir function

The Casimir function C, like enstrophy or potential vorticity in the context of fluid dynamics, is very useful in analyzing
stability conditions and global descriptions of a dynamical system. C is defined by the kernel of the bracket (4), i.e., {C,
G} = 0, ∀G ∈ C∞ (g∗ ) and therefore is a constant of the motion of the Hamiltonian system, C˙ = {C, H } = 0. Moreover it
defines a foliation of the phase space [22,25]. Define the Casimir energy function for the BLDCM chaotic system as
1 1 
C= x, x = x21 + x22 + x23 ,
2 2
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 691

From (4), for ∀x = 0, C > 0, the Casimir function commutes with every function in the Lie-Poisson bracket. From
Eq. (5), the Casimir function can be interpreted as follows:
1 2  1 2 
C= x + x22 + x23 = Li + Li2d + J ω2 . (22)
2 1 2 q
For an electromechanical system, terms Li2q and Li2d in (22) represent the electrical inductor energy of windings, and
term Jω2 is the mechanical rotational energy (11th row in Table 1).

Remark 4. The Casimir function represents the existing internal energy of the BLDCM system. The rate of change of the
Casimir function is
C˙ = ∇C, x˙ 
= {C, H } − ∇ C, x + ∇ C, f
= −x, x + x, f
= −2D p + G p
 
= − Ri2q + Ri2d + bω2 + (iq uq + id ud − TL ω ). (23)

Remark 5. The rate of change of the internal energy, C˙ , is determined by the exchange power between generalized external
energy and dissipative energy of the motor system.

Equation C˙ = 0 determines the extremal points of the Casimir function of BLDCM chaotic system. Given the triaxle
ellipsoid
 uq
2  ud
2  TL
2 u2q u2d TL2
e : R iq − + R id − +b ω+ = + + . (24)
2R 2R 2b 4R 4R 4b
We assume all the input variables uq , ud , TL are constant.

Theorem 1.

(1) The equilibria of system (1) belongs to the extremal ellipsoid, e .

(2) When C˙ > 0, the orbits of the BLDCM system stay inside the extremal ellipsoid, the internal energy is stored up, and the
orbits diverge. When C˙ < 0, the orbits of system run outside the extremal ellipsoid, the internal energy is released, and the
orbits converge.
(3) When C˙ = 0, the system orbits intersect the extremal ellipsoid. These orbits exit the extremal ellipsoid at max (C); they
reenter the extremal ellipsoid symmetrically at min (C).
(4) The BLDCM chaotic attractor has the supremum ellipsoid,
L 2 L 2 J
s : Li2q + Li2d + J ω2 = u + u + T 2. (25)
R2 q R2 d b2 L
Proof.

(1) The extremal ellipsoid can be obtained by solving

C˙ = ∇ C, x˙  = 0,
and the equilibria of system (1) are obtained by solving x˙ = 0, and hence the equilibria belong to e.
(2) When C˙ > 0, from (23) and (24), we have
 uq
2  ud
2  
T 2 u2 u2 T2
R iq − + R id − +b ω+ L < q + d + L ,
2R 2R 2b 4R 4R 4b
which implies that the orbits
 of the system stay inside e , and the Casimir energy increases. The radius of orbit
from the origin is r (t ) = 2C (t ). When the Casimir energy increases, the orbit expands. Likewise, when C˙ < 0, the
trajectories of the system run outside the ellipsoid. The Casimir energy then decreases and the orbits converge.
(3) For C˙ = 0, the system orbits intersect the extremal ellipsoid. When trajectories of the system exit the extremal ellip-
soid, the Casimir energy reaches its maximum. Likewise, when the orbits enter the ellipsoid, the Casimir energy drops
to its minimum, i.e., ∩ e = {max (C), min (C)}, where denotes the chaotic attractor.
(4) From (23),
C˙ = −Ri2q + iq uq − Ri2d + id ud − bω2 − TL ω. (26)
The following inequalities hold,
2R − α L 2 1
iq u q ≤ iq + u2 ,
2 2 ( 2R − α L ) q
2R − α L 2 1
id u d ≤ id + u2 ,
2 2 ( 2R − α L ) d
692 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697

2b − α J 2 1
−ωTL ≤ ω + T 2, (27)
2 2 ( 2b − α J ) L
where α is a factor used in adjusting the space of the boundary ellipsoid. Substituting (27) into (26), we have
1
1  2  1

C˙ (t ) ≤ −αC (t ) + u + u2d + T2 . (28)
2 2R − α L q 2b − α J L
Hence
    
−α t
t
1 1  2  1
C (t ) ≤ e C0 + eατ u + u2d + T 2 dτ . (29)
0 2 2R − α L q 2b − α J L
Because uq , ud and TL are constants, when t → ∞, we have
1
1  2  1

C (t ) ≤ u + u2d + T 2 + ε, (30)
2α 2R − α L q 2b − α J L
where lim ε = 0. Setting the boundary ellipsoid as
t→∞

1
 1  2  1

: Li2q + Li2d + J ω2 = uq + u2d + T2 , (31)
1
α 2R − α L 2b − α J L
Note that α is a factor used in adjusting the space of the boundary ellipsoid, larger the quadratic polynomial terms
α (2R − α L) and α (2b − α J), smaller bound in Eq. (31). To find the minimal bound, we maximize quadratic polynomial terms
in terms of α as follows:
R2 b2
max (α (2R − α L ) )|α = R = , max(α (2b − α J ) )|α = b = .
L L J J
Therefore, the supremum of the chaotic attractor is obtained as in (25). The proof is complete.

4. Mechanism and energy cycling of different dynamical modes

We now investigate the dynamic modes corresponding to the different forms of energy for the BLDCM system (1) to find
the forcing effect of each type of torque and energy, and to uncover the key factors producing the various dynamic modes.
Following the parameters of the BLDCM chaotic system given in [6], we take Kt = 0.031Nm/A, n = 4, L = 14.25 × 10 − 3 H,
R = 0.9 , b = 0.0162 N/rad/s, J = 4.7 × 10 − 5 kg/m2 , and set the generalized external torque uq = 0 V, ud = −50 V, TL = 0 V,
initial conditions iq0 = id0 = 1 A, ω0 = 1 rad/s, and sampling time τ = 0.0 0 01 s.

4.1. BLDCM governed by kinetic energy

Suppose system (15) only contains the term for the inertial torque x˙ = {x, K }. System (1) becomes
i˙q = −nωid , i˙d = nωiq , ω˙ = 0. (32)
Quadratic terms, − nωid and nωid , are terms for inertial torque. The rotor of the motor is frictionless and subject to zero
electrical supply. The rotor runs freely with initial angular velocity according to Newton’s first law. Because ω = ω 0 remains
constant, from the first and second equations, we find the quadrature-axis and direct-axis currents id are periodic (Fig. 1).
However, the angular velocity remains constant.
When the system is only subject to kinetic energy, in another word, the system only contains the inertial torque terms,
the system is conservative, which means the orbit of system is periodic. The conservativeness can be verified by (19),
K˙ = {K, K } = 0, with K = 16.6423. From (23), the Casimir function satisfies C˙ = {C, K } = 0 with the initial internal energy
C = C0 = 0.0143.

4.2. BLDCM system contains kinetic and potential energy

The system is governed by the Lie–Poisson bracket terms for the inertial torque and internal torque in the Hamiltonian
function, given as
x˙ = {x, K } + {x, U }.
System (1) becomes
nkt nkt
i˙q = −nωid − ω, i˙d = nωiq , ω˙ = iq . (33)
L J
Therefore, terms − nLkt ω and nkJ t iq are the terms for internal torque. Compared with (32), an interaction between the
electromagnetic torque and back EMF arises. A periodic orbit is also produced (Fig. 2(a) and (b)). By comparing Fig. 2(a)
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 693

Fig. 1. Time response of the quadrature axis current when the energy of the system is solely kinetic.

Fig. 2. System is governed by both kinetic energy and potential energy.

694 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697

with Fig. 1, we find that the frequency of the orbit in (33) is much larger than that in (32), implying that the internal
torque added makes the orbits oscillate much faster. Not only has the frequency increased, but also the orbit shape is
distorted and expanded by the internal torque (figure not given). Hence the internal torque released by the potential energy
is an important factor in changing the dynamics. The angular velocity changes periodically (Fig. 2(b)), while it is constant
in the last case. From (19) and (20), we obtain
−n2 kt
K˙ = {K, U } =  iq ω, U˙ = −{K, U },
J
which means differing from the situation in (32) both kinetic energy and potential energy oscillate with bound and
exchange energy (Fig. 2(c)). However, the Hamiltonian energy is also conserved since H˙ = K˙ + U˙ = 0 with H = 34.7296
(Fig. 2(c)). The increase in kinetic energy is gained from the potential energy, upholding the conservative law of energy,
which determines the orbit is still periodic.
The Casimir function satisfies C˙ = {C, H } = 0 with the initial C = C0 = 0.0143. The internal energy in both situations has
the same value, although the Hamiltonian energy changes with H = 16.6423 in (32) and H = 34.7296 in (33). As dissipative
energy and generalized energy have not been involved in either case, the system does not exchange energy with the
outside. Hence, the rate of internal energy reflects how much power is exchanged.

4.3. BLDCM governed by energy and dissipative energy

The system is described as

x˙ = {x, K } + {x, U } − ∇ D p (34)
which is equivalent to the following system
Li˙q = −nLωid − nkt ω − Riq ,
Li˙d = nLωiq − Rid ,
J ω˙ = nkt iq − bω. (35)
The phase space volume of the system satisfies
 
R R b
V˙ = − + + V < 0,
L L J

with solution V = V0 exp(−( RL + RL + bJ )). Therefore, the system is dissipative, i.e., the volume in phase space shrinks
exponentially to zero.
We now analyze how the energy shrinks. From (23), we have
 
C˙ = −2D p = − Ri2q + Ri2d + bω2 < 0. (36)
2Dp is the dissipative power of system (35). From (36), the losses in the electric motor can be separated as a loss due to
mechanical friction, bω2 , and a loss in the winding circuit, Ri2q + Ri2d . In practice, the main energy loss of the electric motor
is due to external torque TL . However as its source lies outside the motor, it is not discussed further.
Because the system has the dissipative torque term but no external torque term, from Eq. (36), system (35) is globally
asymptotically stable. Hence, it produces a sink converging toward the origin of the phase space of system (1). Therefore,
system (35) decays in terms of both phase space volume and internal energy.
Note that a normal operating motor, the three situations aforementioned rarely take place. The reason that we list these
situations is to analyze the effect of each term in structure from mechanical point of view.
If the phase space volume of the system decays but internal energy oscillates with bound, the system produces chaos.
To make a chaotic system, supplied energy is needed to dynamically counteract the loss of dissipative energy.

4.4. BLDCM governed by full energy

The BLDCM chaotic system is

x˙ = {x, K } + {x, U } − ∇ D p + ∇ G p . (37)
which is equivalent to system (1). Setting uq = 0, ud = −50 and TL = 0, the supplied voltage then becomes effective in pro-
ducing chaos for the system with positive Lyapunov exponent L = 34.51849 and Lyapunov dimension Ld = 2.07683 < 3 (see
Fig. 3(a)). There are two regions in general: the high energy (cold color) region on the right and low energy (warm color)
region on the left. Differing from systems (32) and (33) with C˙ = 0, and system (34) with C˙ < 0, the internal energy oscil-
lates frequently, randomly, and rapidly in the range [0.0069, 12.0477]; see first plot of Fig. 3(b). From (22), the exchange
power is
 
C˙ = −2D p + G p = − Ri2q + Ri2d + bω2 + (iq uq + id ud − TL ω );
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 695

Fig. 3. System is under full torque.

see third plot of Fig. 3(b). The external torque term ∇ Gp in (37) balances the dissipative torque term ∇ Dp dynamically, and
the system dissipates and absorbs energy abruptly, causing the system orbits to stretch and fold rapidly and chaotically. The
chaotic degree of orbit is determined by how oscillated the exchange power C˙ is, as illustrated in Fig. 3(b).
Note that, the supplied energy is just a necessary condition to produce chaos, which means even if the four types of
torque are all existent, we cannot guarantee the system must produce chaotic orbit. A steady angular speed can also be
generated, which is desired in design in most practice. In chaos application, for instance fluid mixing, chaos generation
is also desired. Normally, a steady speed or chaotic speed needs a controller to implement. If we know when it produces
chaos, periodic orbit or chaos under full types torque, the controller can be exempted or easily implemented. This work
needs further investment of mechanics and bifurcation analysis for parameter values, which is beyond the paper research
scope. However, we can point out a constantly balance between dissipative torque, external torque and internal torque, the
system will produce steady speed, otherwise other dynamical modes will be generated.

4.5. Energy cycling and bound of BLDCM chaotic system

To gain further insight into chaos, we investigate further the Casimir energy. We now illustrate aspects of each of the
three parts of Theorem 1.

(1) The system has three equilibria S1 ,S2 , and S3 as given in (21), where S1 is a saddle-node with eigenvalues of [559.7195,
− 967.5583, − 63.1579], S2 and S3 are saddle-foci with the same eigenvalues: [12.83 + j370.91, 12.83 − j370.91,
− 496.66]. The BLDCM chaotic orbits run around the two saddle-foci. They intersect the extremal ellipsoid e in
(24) (Fig. 4(a)), and the three equilibria are on the extremal ellipsoid. The blue star marks S1 , and the two black ‘+’
mark S2 and S3 , which cannot be seen clearly because they are hidden by dense chaotic orbits. Note that Fig. 4(a) has
been rotated. The chaotic attractor intersects the extremal ellipsoid e symmetrically. The intersections are marked by
red stars representing the maxima of the internal energy and black stars representing the minima. To display clearly
the change of the internal energy along the system orbits, the ellipsoid is removed but the energy color is marked,
as shown in Fig. 4(a). Furthermore, to display clearly the energy cycling process, a 0.2-s orbit is shown in Fig. 4(b).
Equilibria S2 and S3 (black ‘+’) are located in the two centers of the wings of the chaotic attractor (Fig. 3(a) and Fig.
4(b)).
(2) Fig. 4(a) shows the intersections between the invariant extremal ellipsoid and the attractor. After the orbits enter the
extremal ellipsoid, internal energy accumulates, the process of which is marked by red arrows in Figs. 3(a) and 4(b)
in a few orbits. The energy of the orbits gradually increases (color warms up) after the red arrow sign. The orbits
expand, and run away from one of the two saddle-foci of the chaotic attractor.
(3) When the energy reaches its maximum, the orbits exit from inside the ellipsoid (red star in Fig. 4(b)). Thereafter, the
internal energy is used up because the supplied power is less than the dissipative power. The orbital energy of the
system gradually decreases (marked by black arrows), and the orbits shrink and run toward one of the two saddle-foci
of the chaotic attractor. The orbit then runs inside the ellipsoid to restore energy, and continue to repeat the process.
The maxima are the folding points of the orbits, and the minima are the expanding points of the orbits. The analysis
above further explains the mechanism of generation of different dynamical modes, especially chaos.
(4) The chaotic attractor is bounded as formulated in (25). The bound is shown in Fig. 4(c), where the blue region is the
chaotic attractor.

The chaotic attractor is bounded, so an unstable trajectory grows in fractal dimension with folding. Therefore, the
property of the boundary solution is important for chaos. Finding the bound of a chaotic attractor is difficult [34] as the
696 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697

Fig. 4. System impacted by the Casimir function and its derivative. (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)

positive definite matrix is exceedingly difficulty to solve in the equation for Lyapunov stability. In addition, the bound in
[34] was found numerically instead of analytically because optimization software is needed in solving procedures, such as
LMI control toolbox. However, the bound of (25) is found analytically using the rate of change of the Casimir function, and
its construction is quite simple.

5. Conclusion

The original BLDCM was transformed into the Kolmogorov type of system. Four types of torque have been revealed for
vector field of the BLDCM chaotic system. Accordingly, four forms of energy are identified for the system—kinetic, potential,
dissipative, and supplied. The transforming relationship between kinetic energy and potential energy has been clarified for
the system. Through an analogue of electrical and mechanical system, the electromechanical BLDCM was regarded as a pure
mechanical system. The rate of change of the Casimir function is the exchange power between dissipative energy and the
supplied energy of the motor, which determines the behavior of orbit of the BLDCM. The optimal, simple and analytical
supremum bound of the BLDCM chaotic attractor was identified through the Casimir function as well. Four cases analysis
of the BLDCM chaotic system in terms of combinations of four forms of energy further have been performed uncovering
the insight and contributing factors of dynamics of periodic orbit, sink, and the chaotic attractors. The function of each
term in the vector field of the BLDCM and its mechanism of generation of different dynamic modes were revealed, helping
accordingly in BLDCM design to produce the relevant dynamics. If the system needs to be managed so as to avoid chaos
for reduce vibration or to generate chaos for mixing fluid or other application, the parameter space associated with chaos
generation needs to be investigated. The studying of the mechanism underlying energy dissipation identifies sources of
 G. Qi / Applied Mathematical Modelling 51 (2017) 686–697 697

energy consumption of system during the design stage and can mitigate their affects. Identifying the critical parameters via
Hopf and Pitchfork bifurcations needs to be conducted to serve the design work.

Acknowledgment

This work is supported by the Tianjin Natural Science Foundation (17JCZDJC38300).

References

[1] G. Qi, G. Chen, A spherical chaotic system, Nonlinear Dyn 81 (2015) 1381–1392.
[2] Y. Kuroe, S. Hayashi, Analysis of bifurcation in power electronic induction motor drive systems, in: Proceedings of IEEE Power Electronics Specialists
Conference, 1989, pp. 923–930.
[3] Z. Li, J. Park, B. Zhang, G. Chen, Bifurcations and chaos in a permanent-magnet synchronous motor, IEEE Trans. Circuits Syst. I 49 (2002) 383–387.
[4] C. Xia, Permanent Magnet Brushless DC Motor Drives and Controls, John Wiley & Sons Singapore Pte. Ltd., Singapore, 2012.
[5] M.A. Jabbar, H.N. Phyu, Z. Liu, C. Bi, Modeling and numerical simulation of a brushless permanent-magnet DC motor in dynamic conditions by
time-stepping technique, IEEE Trans. Ind. Appl. 40 (3) (2004) 763–770.
[6] N. Hemati, Strange attractors in brushless DC motors, IEEE Trans. Circuits Syst. I Fundam. Theory Appl. 41 (1) (1994) 40–45.
[7] C. Lia, W. Lia, F. Li, Chaos induced in brushless DC motor via current time-delayed feedback, Optik 125 (2014) 6589–6593.
[8] A. Zaher, A nonlinear controller design for permanent magnet motors using a synchronization-based technique inspired from the Lorenz system, Chaos
18 (2008) 013111.
[9] A. Mohammad, K. Arash, G. Behzad, Control of chaos in permanent magnet synchronous motor by using optimal Lyapunov exponents placement, Phys.
Lett. A 374 (2010) 4226–4230.
[10] Z. Ge, G. Lin, The complete, lag and anticipated synchronization of a BLDCM chaotic system, Chaos Solitons Fract. 34 (2007) 740–764.
[11] J.L.D. Rodríguez, M. Thompson, A balanced atmospheric model of Lorenz, Nonlinear Anal.: RWA 11 (2010) 3251–3271.
[12] Y. Wang, Bifurcation and chaos analysis for multi-freedom gear-bearing system with time-varying stiffness, Appl. Math. Model. 40 (2016) 9656–9674.
[13] A.M.A. El-Sayed, H.M. Nour, A. Elsaid, A.E. Matouk, A. Elsonbaty, Dynamical behaviors, circuit realization, chaos control, and synchronization of a new
fractional order hyperchaotic system, Appl. Math. Model. 40 (2016) 3516–3534.
[14] A.M.A. El-Sayed, H.M. Nour, A. Elsaid, A.E. Matouk, A. Elsonbaty, Dynamical behaviors, circuit realization, chaos control, and synchronization of a new
fractional order hyperchaotic system, Appl. Math. Model. 40 (2016) 3516–3534.
[15] Z. Kang, J. Sun, L. Ma Y. Qi, S. Jian, Multimode synchronization of chaotic semiconductor ring laser and its potential in chaos communication, IEEE J.
Quant. Electron. 50 (2014) 148–157.
[16] G. Qi, Z. Wang, Y. Guo, Generation of an eight-wing chaotic attractor from Qi 3-D four-wing chaotic system, Int. J. Bifurc. Chaos 12 (2012) 1250287–1-9.
[17] Q. Yuan, X. Yang, Horseshoe chaos and topological entropy estimation in a palaeoclimate dynamical model, Appl. Math. Model. 40 (2016) 2705–2710.
[18] V. Arnold, Kolmogorov hydrodynamic attractors, Proc. R. Soc. Lond. 434 (1991) 19–22.
[19] A. Pasini, V. Pelino, S. Potestà, Torsion and attractors in the Kolmogorov hydrodynamical system, Phys. Lett. A 241(1998) 77–83.
[20] A. Pasini, V. Pelino, A unified view of Kolmogorov and Lorenz systems, Phys. Lett. A 275 (20 0 0) 435–446.
[21] V. Pelino, A. Pasini, Dissipation in Lie–Poisson systems and the Lorenz-84 model, Phys. Lett. A 291(2001) 389–396.
[22] G. Qi, X. Liang, Mechanical analysis of Qi four-wing chaotic system, Nonlinear Dyn 86 (2016) 1095–1106.
[23] G. Qi, X. Liang, Force analysis of Qi chaotic system, Int. J. Bifurc. Chaos 26 (2016) 1650237–1-13.
[24] G. Qi, J Zhang, Energy cycle and bound of Qi chaotic system, Chaos Solitons Fract. 99 (2017) 7–15.
[25] V. Pelino, F. Maimone, A. Pasini, Energy cycle for the Lorenz attractor, Chaos Solitons Fract. 64 (2014) 67–77.
[26] V.I. Arnold, B.A. Khesin, Topological Methods in Hydrodynamics, Springer, Berlin, 1998.
[27] G. Qi, G. Chen, S. Du, Z. Chen, Z. Yuan, Analysis of a new chaotic system, Phys. A: Stat. Mech. Its Appl. 352 (2005) 295–308.
[28] G. Qi, G. Chen, M.A. van Wyk, B.J. van Wyk, Y. Zhang, A Four-wing chaotic attractor generated from a new 3-D quadratic chaotic system, Chaos Solitons
Fract. 38 (2008) 705–721.
[29] E. Lorenz, Deterministic nonperiodic flow, J. Atmos. Sci. 20 (1963) 130–141.
[30] S. Ye, K.T. Chau, Chaoization of DC motors for industrial mixing, IEEE Trans. Ind. Electron. 54 (2007) 2024–2032.
[31] R. Reyes, C. Cruz, M. Nakano-Miyatake, H. Perez-Meana, Digital video watermarking in DWT domain using chaotic mixtures, IEEE Latin Am. Trans. 8
(2010) 304–310.
[32] K.T. Chau, Z. Wang, Chaos in Electric Drive Systems—Analysis, Control and Application, John Wiley & Sons (Asia) Pte Ltd., Singapore, 2011.
[33] P. Wang, D. Li, X. Wu, J. Lü, X. Yu, Ultimate bound estimation of a class of high dimensional quadratic autonomous dynamical systems, Int. J. Bifurc.
Chaos 21 (2011) 2679–2694.
[34] A.P. Krishchenko, K.E. Starkov, Localization of compact invariant sets of the Lorenz system, Phys. Lett. A 353 (2006) 383–388.
[35] L.N. Coria, K.E. Starkov, Bounding a domain containing all compact invariant sets of the permanent-magnet motor system, Commun. Nonlinear Sci. 14
(2009) 3879–3888.
[36] J.E. Marsden, T.S. Ratiu, Introduction to Mechanics and Symmetry: A Basic Exposition of Classical Mechanical Systems, second ed., Springer, 2002.
[37] E. Cheever, in: Analogous electrical and mechanical systems, in: http://lpsa.swarthmore.edu/Analogs/ElectricalMechanicalAnalogs.html, 2016.