308 JOURNAL OF COMPUTERS, VOL. 8, NO.

2, FEBRUARY 2013

The Simulation of Electromagnetic Suspension

System Based on the Finite Element Analysis
Zhengfeng Ming
School of Electronic & Machanical Engineering, Xidian University, Xi’an, China
Email: mingzf@xidian.edu.cn

Tao Wen and Tuo Chen

National Key Laboratory of Antennas and Microwave Technology, Xidian University, Xi’an, China
Email: wt02051327@126.com, ckvty@163.com

Abstract—In this article, the large air gap electromagnetic analysis and the Maxwell's equations, the space magnetic
suspension system model is accurately divided into the mesh field intensity is analyzed and calculated. The finite
by the method of the finite element analysis, then the element analysis is applied to the electro-magnetic
electromagnetic force of the suspended matter in the system suspension technology, and we get the space magnetic
is calculated. Finally, the fitting curves of the experimental
data of a new kind of distributed electromagnetic array
field intensity and calculate several change trends of
system can be obtained through the least squares method electromagnetic force under the large air gap. Finally, the
and the fitting expressions of different model system in the method of least squares is used to get the relevant fitting
large air gap are achieved. mathematical expressions.

Index Terms—electromagnetic suspension system; finite II. THE FINITE ELEMENT ANALYSIS THEORY AND
element analysis; large air gap ELECTROMAGNETIC SUSPENSION SYSTEM

A. The Finite Element Analysis Theory

I. INTRODUCTION
The finite element analysis method is a numerical
The finite element analysis method is a numerical method based on variation principle approximate
method solving the numerical and physical boundary interpolation and discretization. Firstly, the boundary
value approximately. Its development experience is value problem to be solved is converted into the
consisted of Node Finite Element Method (N-FEM) and corresponding variation problem and the variation
Edge-Based Finite Element Method (EB-FEM). The so- problem is converted into the extremal problem of
called Node Finite Element Method uses scalar general multivariate function. Finally, an algebraic
interpolation function as a base function, but the great equation is obtained, the solution of which is a numerical
inconvenience of dealing with conductor and medium solution.
edge horn is the biggest weakness of this method. Thus The most important step of the finite element analysis
the vector finite element method is proposed, which gives is the regional discretization, which will influence the
the freedom to the edge instead of node. computing time, the requirements of computer memory
The core idea of finite element is the discretization of and the numerical results precision. For the three
the structure. Firstly, a real structure is imaginarily dimensional （3D） problem, the most common unit is
discretized into a finite number of inerratic combination the tetrahedron which is the simplest and most suitable
units. Secondly, the physical parameters of the actual for discrete arbitrarily volume area, as shown in Fig. 1.
structure can be calculated by the analysis of the discrete After discretization of the volume area, a suitable basis
body. Finally, the approximate result that meets the function is needed to approximately express the unknown
engineering accuracy is obtained and replaces the functions in each unit which mainly includes Node Finite
analysis of the actual structure. As one of the popular Element Method (N-FEM) and Edge-Based Finite
advanced technology in recent years, the technology Element Method (EB-FEM) (which is commonly used
research is receiving more and more attention [1] [2] [3]. now). For the vector function wi , j , we have
In this paper, several kinds of mathematical models of
the large air gap electromagnetic levitation system have wi , j = λi ∇ λ j − λ j ∇ λi (1)
been given. On the basis of the theory of finite element

© 2013 ACADEMY PUBLISHER

doi:10.4304/jcp.8.2.308-312
JOURNAL OF COMPUTERS, VOL. 8, NO. 2, FEBRUARY 2013 309

where i, j are the vertexes. Since a tetrahedron has four

vertexes, the volume coordinates of the tetrahedron can TABLE I.
EDGE DEFINITION OF TETRAHEDRON UNIT
be described as
λi = ai + bi x + ci y + di z (2) Edge i Start node i1 End node i2
where 1 1 2
−1
⎡ b1 c1 d1 a1 ⎤ ⎡ x1 x2 x3 x4 ⎤
⎢b c d a ⎥ ⎢ y y y y ⎥ 2 1 3
⎢ 2 2 2⎥ (3)
2
=⎢ 1 2 3 4⎥
3 1 4
⎢b3 c3 d 3 a3 ⎥ ⎢ z1 z2 z3 z4 ⎥
⎢ ⎥ ⎢ ⎥ 4 2 3
⎣b4 c4 d 4 a4 ⎦ ⎣ 1 1 1 1 ⎦
5 2 4

6 3 4

The final results of finite element formula is a group of

linear equations, which can be described as
a11 x1 + a12 x2 + a13 x3 + ... + a1n xn = b1
a21 x1 + a22 x2 + a23 x3 + ... + a2 n xn = b2
(10)
a31 x1 + a32 x2 + a33 x3 + ... + a3n xn = b3
...
an1 x1 + an 2 x2 + an 3 x3 + ... + ann xn = bn
which can be reduced as
Figure 1. Tetrahedron unit of the finite element division Ax = b (11)
where A is a coefficient matrix of size n × n , x is an
The following result can be obtained unknown quantity and b is a known vector. For 3D
∇ ⋅ w12 = 0, ∇ × w12 = 2∇λ1 ×∇λ2 (4) electromagnetic field problem, the dimension of the
Assume that e1 represents a unit vector from node 1 to coefficient matrix A increases rapidly with the increase
node 2, λ1 is the linear function that changes from 1 of solving region and A is a sparse matrix.
(node 1) to 0 (node 2) and λ2 is the linear function that B. The Basic Principle of Electromagnetic Suspension
System
changes from 1 (node 2) to 0 (node 1), therefore
1 1 The edge method is used to solve the static magnetic
e1 ⋅ ∇λ1 = − , e1 ⋅∇λ2 = (5)
l1 l1 field, which is using the freedom of division on the unit.
The basic Maxwell's equations of 3D static magnetic field
where l1 represents the length of the edge connecting are shown as follow
nodes 1 and 2, so ∇ × H ( x, y , z ) = J ( x, y , z )
λ1 + λ2
1 ∇ B ( x, y , z ) = 0
(12)
e1 ⋅ w12 = = (6)
l1 l1 B = μH
It means that w12 has a constant tangential component where B ( x, y, z ) is the magnetic flux density, H ( x, y , z ) is
along the edge (1, 2). As a vector base function of an the magnetic field intensity and J ( x, y , z ) is the current
edge field related to the edge (1, 2), w12 has all the density. The third equation describes the constitutive
necessary features. If the edge is defined as edge 1, we relationship between magnetic field intensity and the
have magnetic flux density, where μ is the dielectric magnetic
N1 = w12l1 = (λ1∇λ2 − λ2∇λ1 )l1 (7) permeability. The cylindrical coordinate system and the
Similarly, the vector base function of the edge i is magnetic vector potential A are brought in for the
described as below axisymmetric problems [4] [5].
N i = wi1i2 li = (λi1 ∇λi2 − λi2 ∇λi1 )li (8) The magnetic flux density of the permanent magnet is
described as
where the definitions of edge number, node i1 and node
B = μ0 μ H + μ0 M P (13)
i2 are shown in Fig. 1 and Table I. where μ0 = 4π × 10−7 H/m is the magnetic permeability in
So the vector field of the unit can be divided into
6
absolute vacuum, μ r is the relative magnetic permeability
E e = ∑ Nie Eie (9)
and H P (Mp) is the intensity of polarization of the
i =1

where E e represents the field of the unit, N ie represents permanent magnetic material.
The number of the degrees of freedom to be solved on
the vector base function of the unit and E e represents the the tetrahedral partition adopted is up to 10, four are the
tangential fields along the edge i . magnetic scalar potential on the four vertices of

© 2013 ACADEMY PUBLISHER

310 JOURNAL OF COMPUTERS, VOL. 8, NO. 2, FEBRUARY 2013

tetrahedron, the other six are the magnetic flux densities

of the six sides of tetrahedron. The method of quadratic
interpolation is used to approximate the field quantity in a
single division.
The electromagnetic force equation is
Φ (i , δ )
( )
−1
Bc (i, δ ) = c = 0.5μ 0CΦ Ni (t ) S c−1 (δ ) ⋅ ∫ S c−1 (δ )d δ
Sc (δ ) δ

(14)
where C is the magnetic flux leakage coefficient, which
Φ

has relations with the leakage magnetic flux Φ ; l

Sc (δ ) is
the magnetic flux equivalent area function, which has
relations with δ (t ) . Figure 2. Finite mesh dissection map of single electromagnetic
Suspension suffered electromagnetic attraction is suspension model
B 2 A μ0 AN 2 i 2 (t ) (15)
Fm ( z , i ) = = * 2
μ0 4 z (t )
where F is the levitation force (N) of each
electromagnet; A is the area of permanent magnet
（cm2）; B is the magnetic field strength (T).
C. Model
Electromagnetic suspension system makes use of the
repulsion between the electromagnetic coils and the
permanent magnets in the space, which can provide
enough electromagnetic levitation force for offsetting the
gravity of the suspended.
Figure 3. Space magnetic field distribution map of single
III. THE ANALYSIS OF THE SINGLE ELECTROMAGNETIC electromagnetic mode
SUSPENSION SYSTEM
According to the analysis above, the model of the
electromagnetic levitation system is established. As one
of the most important factors for the electromagnetic
levitation force, the air gap is analyzed and studied,
model specific parameters are shown in the Table II.

TABLE II.
SINGLE ELECTROMAGNETIC MODEL GEOMETRY
PARAMETERS

Permanent Permanent
Parameter Current
magnet radius magnet thickness
Variable R1/mm T1/mm I/A
Figure 4. Relation between electromagnetic force and height
Value 350 20 16000

Parameter
Electromagnet Electromagnet In order to explore the rules of the electromagnetic
core radius core thickness levitation system under the large air gap, the least square
Variable R2/mm T2/mm method is used in data fitting. Compared to other data
fitting method such as interpolation method and
Value 275 400
approaching function method, the least square method
works better [6] [7]. The base functions that can be
The finite mesh dissection of the single electro- selected in the least square method are various, such as
magnetic suspension model can be made through the the common polynomial, Chebyshev polynomial,
finite element method, and the results are shown in Fig. 2. Bemstein polynomial, etc. According to the electro-
Other conditions remain unchanged, the working magnetic field theory formula, the least square method is
environment is air and the air gap size changes from 0m used in data fitting based on the rational fitting function,
to 4 m. Fig. 3 shows the magnetic field distribution of the The fitting formula is
electromagnetic suspension system and the date of f ( x) = 0.2259 x −2 + 653x −1 − 308.5 (16)
electromagnetic force is shown in Fig. 4. Fig. 4 shows the fitting graphics. From the fitting
formula we can see that the second item coefficient of the

© 2013 ACADEMY PUBLISHER

JOURNAL OF COMPUTERS, VOL. 8, NO. 2, FEBRUARY 2013 311

formula is small while the first item coefficient is large, the fitting curve of the axial electromagnetic forces that
they are different from the theoretical formula. Because changed with the horizontal displacement. The fitting
during 2-4 meters, the electromagnet magnetic fields formula is
intensity produced by electromagnet in space is abate, the f ( x) = −41.05x2 − 6.244 x − 99.77 (17)
characteristics of magnetic field has changed. Fig.8 shows the fitting curve of the lateral electro-
magnetic forces which responses with the horizontal
IV. THE ANALYSIS OF THE DISTRIBUTED displacement. The fitting formula is
ELECTROMAGNETIC SUSPENSION SYSTEM f ( x) = −24.25x2 − 71.39 x + 0.1023 (18)
In order to be more close to requirements of real
experiment, we increase the number of the research
object.
Taking four-distribution array as an example, the
original single electromagnet is divided into four small
single electromagnets. Model geometry parameters are
shown in the Table III.

TABLE III.
DISTRIBUTED ELECTROMAGNETIC MODEL GEOMETRY
PARAMETERS

Permanent Permanent
Parameter Current
magnet radius magnet thickness
Variable R1/mm T1/mm I/A Figure 5. Space magnetic field distribution map of distributed
electromagnetic model.
Value 350 30 18000
Electromagnet Electromagnet
Parameter
core radius core thickness
Variable R2/mm T2/mm

Value 500 400

The finite mesh dissection of the distributed

electromagnetic suspension model can be made and the
result is shown in Fig.5. The magnetic field distribution
of the electromagnetic suspension system is shown in
Fig.6. The working environment is air and the suspension
is 2m in height, moving the suspended matter along
horizontal direction, we can observe the changing law of
the electromagnetic force, the result is shown as below.
Figure 7. Relation between axial force and horizontal displacement.

Figure 8. Finite mesh dissection map of distributed electromagnetic

model. Figure 6. Relation between lateral force and horizontal displacement.

In the distributed system model, the axial force change

According to the electromagnetic field theory formulas, formula mainly presents the decline of the parabolic
the data is fitted by the least square method. Fig. 7 shows tendency, the second item coefficient is far bigger than

© 2013 ACADEMY PUBLISHER

312 JOURNAL OF COMPUTERS, VOL. 8, NO. 2, FEBRUARY 2013

the first one; On the contrary, the second item coefficient Zhengfeng Ming (M’08) was born in
of the lateral force change formula is less than the first Xi’an County, Shaanxi Province, China.
one, which is closer to linear characteristics. The rule that He received the B.S. degree in automatic
electromagnetic force changes with the distance is control from Xi’an University of
Technology, Xi’an, China, in 1988, the
inversely proportional to the distance, and the x here is M.S. degree in the industrial electrical
just a part of the direct distance. automation from Xi’an Science
University, Xi’an, in 1993, and the Ph.D.
V. CONCLUSION degree in electrical engineering from
Xi’an University of Technology, in 2002.
This paper analyzes the electromagnetic field of mixed From March 2003 to April 2005, he was a Postdoctoral
suspension electromagnet with finite element analysis Researcher of electrical engineering in Zhejiang University,
method. We model and analyze the suspension Hangzhou, China. He joined Department of Electrical
electromagnet to study the relationship between the Engineering, Xidian University, thereafter and is currently an
electromagnetic force, current, and suspension gap. Then, Associate Professor and Chair of the Department. He was a
the change law of the electromagnetic force with visiting Associate Professor in the Department of Electrical and
horizontal offset electromagnetic field is also studied. Computer Engineering at New Jersey Institute of Technology
Then the calculated dates are fitted by the least square (NJIT), Newark, in 2009. He has authored more than 30
published technical conference and journal papers in the area of
method. Finally, we compare the fitting formula with the
power electronics. His research interests include digital control
theory one. These results lay the foundation for the for power electronics applications, high-power converters,
further experimental study. three-phase inverters and rectifiers, motor drives, soft-switching
techniques, and integrated power electronics systems.
REFERENCES
[1] M. Hano, “Finite-element solution of three-dimensional
resonator problems: Novel rectangular parallelepiped
elements,” Electron. Commun. Jpn. Pt. 2, vol. 71, no. 7,
pp.27-34, 1998.
[2] G. Mur and A. T. de Hoop, “A finite-element method for
computing three- dimensional electromagnetic fileds in
inhomogeneous media,” IEEE Trans. Magn. vol. MAG-21,
pp. 2188-2191, Nov. 1985.
[3] Mimpei Morishita, Teruo Azukizawa, Shuji Kanda. “A
New Maglev System for Magnetically Levitated Carrier
System,” IEEE Transactions on Vehicular technology.
November 1989, vo1. 38, No 4.
[4] Coulomb. J, “A methodology for the determination of
global electromechanical quantities from a finite element
analysis and its application to the evaluation of magnetic
forces, torques and stiffness,” IEEE Trans. on Mag, 1983,
19(6):2514-2518.
[5] Golob M, Tovornik B. Modeling and control of the
magnetic suspension system [J]. ISA Transactions, 2003,
42 (1): 89-100.
[6] Kiasi F, “An interpolative fuzzy inference using least
square principle by means of β-function and high order
polynomials,” IEEE International Conference on
Mechationics & Automation, Vol. 1, pp. 545 – 550, 2005.
[7] Roychowdhury S, “Fuzzy curve fitting using least square
principles,”1998 IEEE International Conference on
Systems, Man and Cybemetics, 4:4022-4027

© 2013 ACADEMY PUBLISHER