Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Using COMSOL Multiphysics for the Modelling of a Hybrid Linear

Stepper Motor
Riheb Wislati *,1, Helmut Haase 1
1
Leibniz Universität Hannover, Institut für Grundlagen der Elektrotechnik und Messtechnik
*Dipl.-Ing. Riheb Wislati, Inst. für Grundlagen der Elektrotechnik und Messtechnik, Appelstr. 9A, 30167 Hannover,
wislati@geml.uni-hannover.de

Abstract: In this paper a 2-phase hybrid linear
stepper motor (HLSM) also known as Sawyer
linear motor has been considered as a potential
approach in Variable Valve Actuation (VVA)
Systems for Internal Combustion Engines.
Initially the reluctance network approach (RNA)
with lumped parameters has been used assuming
an infinite permeability of the steel core. The
results have been compared with a finite element
(FEM) simulation. Thereby a 2D FEM model in
COMSOL Multiphysics has been used.
Furthermore the FEM simulation has been
performed taking into account all non-linear
effects. The static force calculation has been
applied using the virtual work method. Finally a
SIMULINK model is presented to simulate the
dynamics of the HLSM using FEM results.
motor is considered in this paper as a potential
valve actuator.
2. General Requirements
The typical valve stroke curve is depicted in
Fig. 1. The maximum lift xˆ = 9 mm has to be
reached during the time T = 3 ms . The required
maximum force depends on the used profiles for
velocity v and acceleration a and on the
mass m of all moving parts.
T = 3ms
xˆ = 9 mm

Keywords: Hybrid linear stepper motor, Sawyer

linear motor, variable valve actuation, reluctance
network approach, finite element method.

1. Introduction

In recent years, one of the main focuses

within the automotive industry was improving Figure 1. Valve stroke curve.
fuel economy and reducing exhaust emissions. In
order to achieve this purpose, variable engine
valve actuation systems are considered. Notice Assuming the profiles in Fig. 2 the equations
that in conventional engines the valve’s open and for the valve stroke, velocity and acceleration
closing timings are fixed relatively to the engine within the interval 0 ≤ t ≤ T are given by
crank angle and cannot be adjusted to engine
load and speed. xˆ ⎛ ⎛ π ⎞⎞
There are several ways to implement variable x (t ) = ⎜ 1 − cos ⎜ t ⎟ ⎟ (1)
2⎝ ⎝ T ⎠⎠
valve trains [1]. Mechanical approaches such as
the Valvetronic and Vanos-System by BMW and dx xˆ π ⎛π ⎞
the VETEC-System by Honda make use of an v (t ) = = sin ⎜ t ⎟ (2)
dt 2 T ⎝T ⎠
adjustable camshaft. However, most research
projects focus on the electromagnetical approach dv xˆ ⎛ π ⎞
2
⎛π ⎞
[2][3][4][5]. a (t ) = = ⎜ ⎟ cos ⎜ t ⎟ (3)
Because of their efficiency, high reliability dt 2 ⎝ T ⎠ ⎝T ⎠
and mechanical simplicity hybrid linear steeper respectively.
motors have been widely applied in high speed
accurate positioning systems. Therefore a
2-phase HLSM also known as Sawyer linear
 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Stroke attenuates it at Pole1’. A tangential force (thrust)

x, v , a
Velocity results at the face of Pole 1 and tends to align its
Acceleration teeth with the corresponding teeth of the mover
x̂
in such a manner as to minimize the air gap
vmax magnetic energy. When coil 1 is switched off
and coil 2 is excited in a way that the magnetic
flux at Pole 2 becomes higher, the mover will be
amax
driven a quarter tooth pitch to the right.
Note that in most applications [6][7][8] the
coils and the permanent magnet are arranged at
the mover. In this paper it is else wise to reduce
0 T t the mass of the mover.

Phase 1 Phase 2
amax Φ PM Coil 1 Coil 2
PM Stator
Φ1
N S
Figure 2. Used profiles for stroke, velocity and PM flux Φ PM
Pole 1 Pole 1' Pole 2 Pole 2'
acceleration. coil 1 flux Φ1

δ Mover
From Eq. (3) the maximum acceleration is
determined by
2 τ
xˆ ⎛ π ⎞ m x
amax = ⎜ ⎟ ≈ 4950 2 (4) 0
Φ PM
2⎝T⎠ s
Φ2
Assuming that m = 100 g the maximum N S
PM flux Φ PM
force required from the actuator is coil 2 flux Φ 2

Fmax = 495 N (5)

The available mounting space sets very

restrictive design requirements. The outer
dimensions of the actuator may not exceed
36 mm (width), 60 mm (depth) and 100 mm
(height). Therefore the most important criterion
in this preliminary study is to reach the
maximum force Fmax considering the given
geometrical dimensions.
3. Principle of Operation
The operating principle of the 2-phase hybrid
linear stepper motor (HLSM) also known as
Sawyer linear motor is shown in Fig. 3. The
HLSM is in fact a variable reluctance, permanent
magnet excited synchronous linear motor. The
mover is a laminated steel bar. The stator
consists of 2 wound pairs of poles (Pole 1 and
Pole 1’ respectively Pole 2 and Pole 2’)
separated by a permanent magnet (PM). When
coil 1 is excited the resulting magnetic field
reinforces the PM magnetic flux at Pole 1 and
x0
Figure 3. Construction of a HLSM.
4. Reluctance Network Approach
In this section the reluctance network
approach is used to analyze the HLSM [9].
Thereby following assumption have been made:
• The permeability of all steel parts tends
to infinity.
• Leakage fluxes are negligible.
• The PM has the B-H curve according to
Fig. 4.
At first, an equivalent circuit of the PM
(length LP and cross-section AP ) has to be
determined. The corresponding constitutive
equation to Fig. 4 is
Br
bp = Br + μ P hp with μP = (6)
Hc
 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

bP The equivalent circuit of the whole HLSM is

depicted in Fig. 6, in which Θ1 , Θ 2 and Θ3 are
Br respectively MMFs of the coil 1, coil 2 and the
⎛ τ⎞ ⎛ τ⎞
PM. G ( x ) , G ⎜ x − ⎟ , G ⎜ x − ⎟ and
⎝ 2⎠ ⎝ 4⎠
μP ⎛ τ ⎞
G ⎜ x + ⎟ are the air gap permeances of Pole 1,
⎝ 4⎠
Pole 1’, Pole 2 and Pole 2’, respectively. They
−Hc hP depend on the mover position x and the tooth
pitch τ . G ( x ) can be expressed as
Figure 4. B-H curve of the PM.
G ( x ) = nt Gt ( x ) (10)
where Br is the remanent flux density and H c is
where nt is the number of teeth per pole and
the coercive force.
After some transformations of Eq. (6) we Gt ( x ) is the air gap permeance corresponding to
obtain the following equation one tooth pitch. Based on finite element
simulations, Gt ( x ) can be approximated with
ϕP
vP = − ΘP (7) the following sinusoidal expression [9] (see Fig.
GP
7)
which describes the equivalent circuit of the PM Gmax + Gmin Gmax − Gmin ⎛ 2π ⎞
shown in Fig. 5 as a magnetomotive force Gt ( x ) = + cos ⎜ x⎟
(MMF) source Θ P in series with an internal  
2 2
 
⎝ τ ⎠
GA GB
permeance GP :
⎛ 2π ⎞
Θ P = H c Lp (8) = GA + GB cos ⎜ x⎟ (11)
⎝ τ ⎠
μ P AP The maximum permeance Gmax and the
GP = (9)
LP minimum permeance Gmin are defined as
ΘP
⎡ c 2 ⎛ π b ⎞⎤
ϕP ϕP
GP Gmax = μ0 L ⎢ + ln ⎜1 + ⎟⎥ (12)
⎣ δ π ⎝ 2δ ⎠ ⎦

vP vP
⎡ b−c 8 δ + 0, 25π b ⎤
Gmin = μ0 L ⎢ + ln
Figure 5. Equivalent circuit of the PM. ⎣ δ + 0, 25π (b − c ) π δ + 0, 25π (b − c) ⎥⎦
(13)

Θ1 Θ3 = Θ P Θ2 where b , c , δ and L are respectively the slot

ϕ1 GP width, tooth width, air gap width and the air gap
ϕ3 = ϕ P ϕ2
depth.
Using the virtual work method, the tangential
force is
⎛ τ⎞ ⎛ τ⎞ ⎛ τ⎞
G ( x) G⎜ x − ⎟ G⎜ x − ⎟ G⎜ x + ⎟ ∂W * ( x, i )
⎝ 2⎠ ⎝ 4⎠ ⎝ 4⎠
Fx = (14)
∂x

where W * is the stored co-energy.

Figure 6. Equivalent circuit of the HLSM.

 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Gt Table 1: Design data of the HLSM

Gmax Specification data Value

GB Tooth pitch, τ 4 mm
Tooth width , c 2 mm
GA Slot width , b 2 mm
Number of teeth per pole, nt 3
Gmin Air gap width, δ 0,1mm
0 0, 25 0,5 0, 75 1 x PM length, LP 10 mm
τ PM face area, AP 200 mm 2
PM remanent flux density, Br 1,3T
Figure 7. Approximated air gap permeance over one
tooth pitch. PM coercive force, H c 106 A m

The co-energy can be written in matrix form

as follows
W * = W1* + W2* + W3*
1 1 1
= Θ1ϕ1 + Θ 2ϕ 2 + Θ3ϕ3 FEM RNA
2 2 2 2

1
= Θ1 ( G11Θ1 + G12 Θ 2 + G13 Θ3 ) 1
2 F 0 Θ1= 0
1
+ Θ 2 ( G21Θ1 + G22 Θ 2 + G23 Θ3 ) (15) N
2 -1
1
+ Θ3 ( G31Θ1 + G32 Θ 2 + G33 Θ3 ) -2
2 0 0.2 0.4 x 0.6 0.8 1
τ
⎛ G11 G12 G13 ⎞ ⎛ Θ1 ⎞
1 ⎜ ⎟⎜ ⎟ 100
= ( Θ1 Θ 2 Θ3 ) ⎜ G21 G22 G23 ⎟ ⎜ Θ 2 ⎟ Θ 1 = 400 A
2   

⎜G G32 G33 ⎟⎠ ⎜⎝ Θ3 ⎟⎠
50
ΘΤ ⎝31
 
N F
Θ 0
G N
1 T
= Θ GΘ -50
2 Θ 1= - 400 A
-100
0 0.2 0.4 x 0.6 0.8 1
The elements of G can be calculated by using τ
the superposition method. Eq. (14) becomes 400
Θ 1 = 800 A
1 ∂G 200
Fx = ΘT Θ (16)
2 ∂x F
0
N
5. Simulation -200
Θ 1= - 800 A
Table 1 shows some design data of the -400
0 0.2 0.4 0.6 0.8 1
x
HLSM used in the following simulations. The
τ
force calculation using the RNA has been
performed in MATLAB. The results have been
compared with those from the FEM simulation
in COMSOL Multiphysics (see Fig. 8). Thereby Figure 8. Tangential force versus displacement when
a constant high permeability has been used for phase 1 is excited ( Θ 2 = 0 ).
the steel parts.
 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

It can be seen that the RNA does not take

into account the cogging force
Θ2 = 0
F ( Θ1 = 0, Θ 2 = 0 ) . This is due to the 100

assumption made in Eq. (11) which does not 50

consider higher even harmonics of the air gap F
permeance [10]. N
0

In the following simulation the nonlinear -50

B-H curve in Fig. 9 has been used for all steel
-100
parts. This curve can be expressed as 2000
1000 1
B = a ⋅ asinh ( b ⋅ H ) (17) Θ1 0
0.5
x
A -1000
τ
-2000 0
m
with a = 0, 25 T and b = 0, 06 . Accordingly, Θ1 = 0
A
100
the relative permeability is
50
b⋅B
μr = (18) F 0
⎛B⎞ N
μ0 ⋅ sinh ⎜ ⎟
⎝a⎠ -50

-100
The use of this analytical expression instead of 2000
a lookup table leads to a faster convergence of 1000
Θ2 0 x
1

the simulation in COMSOL. A -1000

0.5
τ
The tangential force has been calculated over -2000 0

one tooth pitch for −2000 A ≤ Θ1 ≤ 2000 A and τ

x=
−2000 A ≤ Θ 2 ≤ 2000 A using the virtual work 100 8
method. The results are presented in Fig. 10. The 50
maximum force reached is about 80 N . F
0
Remember that the required maximum force is N
495 N . However, the used motor geometry is -50

not optimized. Furthermore the force can be -100

2000
increased by using a double-sided design for the 1000 2000
HLSM. Θ2 0
0 Θ1
1000
-1000 -1000
A -2000 -2000 A

3
Figure 10. Results of the FEM simulation.
2.5
Magnetic Flux Density B in T

5. Simulink Model for Dynamics

2

The derivation of the Simulink model is

1.5
based on the electrical equations
1 d Ψ1 ( i1 ,x )
u1 = R ⋅ i1 + (19)
dt
0.5

d Ψ 2 ( i 2 ,x )
0 u2 = R ⋅ i2 + (20)
0 0.5 1 1.5 2 2.5
5
dt
Magnetic Field Strength H in A/m x 10
and on the equation of motion
Figure 9. Used B-H curve.
x = F ( x, i1 , i2 )
m  (21)
 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

with 6. Conclusions
uμ : Input voltage of phase μ
iμ : Current of phase μ The HLSM has been analyzed by using RNA
and FEM in COMSOL Multiphysics. The
Ψμ : Flux linkage of phase μ comparison of the obtained results has showed
F ( x, i1 , i2 ) : Electromagnetic force that RNA is not as precise as FEM. However, by
tuning the equivalent circuit of the HLSM
Thereby damping, friction, eddy currents and the
according to the FEM results an improved RNA
mutual inductance have been neglected. The
model can be obtained. Moreover nonlinearity
state equation describing the model can be
effects can be also implemented in this model.
written as
So the RNA could be an alternative to FEM,
u1 − R ⋅ i1 ( Ψ1 , x ) especially in optimisation analysis.
⎛Ψ  ⎞ ⎛ ⎞
Furthermore it has been shown that the
⎜ ⎟
u2 − R ⋅ i2 ( Ψ 2 , x )
1
⎜ ⎟ ⎜ ⎟ required maximum force was not reached. This
Ψ
⎜ 2⎟=⎜
1 ⎟ (22) can be traced back to the fact that the used
⎜ v ⎟ ⎜ ⋅ F ( x, i ( Ψ , x ) , i ( Ψ , x ) ) ⎟
⎜⎜ ⎟⎟ ⎜ m 1 1 2 2
⎟ geometry is not optimised and that a single-sided
⎝ 
x ⎠ ⎜ ⎟ design was used.
⎝ v ⎠
Finally, a Simulink model has been
F ( x, i1 ( Ψ1 , x ) , i2 ( Ψ 2 , x ) ) is directly provided by presented to simulate the dynamics of the HLSM
based on FEM results.
the FEM simulation as a 3D matrix and can be
implemented in Simulink as a 3D lookup table. 7. References
However, the matrices Ψ1 ( i1 ,x ) and Ψ 2 ( i 2 ,x )
have to be transformed in i1 ( Ψ1 , x ) and 1. Butzmann, S., Sensorlose Regelung
elektromagnetischer Aktuatoren für die
i2 ( Ψ 2 , x ) . The Simulink model is presented in Betätigung von Gasventilen im Otto-Motor,
Fig. 11 with the input variables u1 and u2 and Universität Bochum, Bochum Germany (2000)
the output variables x , v and a . In [11] a model 2. Warburton, A.; Fleming L.; Scott, J.; Butler,
N., Intelligent Valve Actuation (IVA) System for
is presented considering friction, damping and
Gasoline and Diesel Engines, SAE, 2005-01-
eddy currents.
0772, (2005)
3. Montanari, M.; Ronchi, F.; Rossi, C.,
Trajectory Generation for Camless Internal
2D lookup table
Combustion Engine Valve Control, Proceedings
of IEEE International Symposium on Industrial
Electronics, (2003).
u1 i1 ( Ψ1 , x ) 4. Chang, W.S.; Parlikar, T.A.; Seeman, M.D.;
∫ Perreault, D.J.; Kassakian, J.G.; Keim, T.A., A
New Electromagnetic Valve Actuator, IEEE
R Power Electronics in Transportation, (2002)
5. Wang, Y.; Megli, T.; Haghgooie, M.;
Peterson, K.S.; Stefanopoulou, A.G., Modelling
F ( x, i1 , i2 ) 1m x and Control of Electromechanical Valve
v Actuator, SAE, 2002-01-1106, (2002)
3D lookup table 6. Daud, A.K.; Hanitsch, R., Design of Hybrid
R
a Linear Stepping Motor (HLSM) for Long Stroke
2D lookup table
Operation, Proceedings of the 16th International
Conference on Electrical Machines (ICEM’04),
u2 i2 ( Ψ 2 , x ) (2004)
7. Kallenbach, E.; Eick, R.; Quendt, P.; Ströhla,
T.; Feindt, K; Kallenbach, M., Elektromagnete,
Figure 11. Simulink model for dynamics. Teubner, (2003)
 Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

8. Draeger, J.; Moczala, H., Elektrische Linear-

Kleinmotoren, Franzis-Verlag, (1987)
9. Gieras, J.F.; Piech, Z.J.: Linear Synchronous
Motors, CRC Press Inc, (1999)
10. Balkovoi, A.; Kallenbach, E., Linear
Stepping Motor Model for Thrust Analysis and
Control, 45th International Scientific Colloquium,
Ilmenau Technical University, (2000)
11. Räumschüssel, E.; Lipfert, R., Nichtlineares
Modell eines Linearschrittmotors auf der Basis
von Daten aus der Magnetfeldberechnung, 45th
International Scientific Colloquium, Ilmenau
Technical University, (2000)