IEEE TRANSACTIONS ON EDUCATION, VOL. 36, NO.

2, MAY 1993 233

A Graduate Course on Finite Element Analysis

for Electromagnetic Applications
A. A. Arkadan, Senior Member. IEEE

Abstract-The purpose of this paper is to present a graduate TABLE I

course on finite element analysis for electromagnetic applications COURSEOUTLINE
BY TOPICS
which is offered at Marquette University. This course covers
the theoretical background which is essential in learning how - Basic field formulations
to model electromagnetic devices. In addition, the course stresses - Vector field formulations
the practical aspects for method implementation, and involves - Types of analysis
- The partial differential equations: static and time-varying cases
the application of the finite element method to engineering and - Formulation of the finite element discretization
design problems. - Consideration of applications and method implementation
- Representation of ferrous materials and permanent magnets
- Implementation of nonlinear analysis
I. INTRODUCTION - Preprocessing, postprocessing, and practical aspects

T HE finite element (FE) method for electromagnetic field

analysis, which was introduced over two decades ago
[l], [2] has become a well-established tool for the design
- Three dimensional analysis and other modeling considerations

and analysis of electromagnetic devices. It is heavily used software packages [NI, [19]. As for the final project, each
in both university and industry research. Several publications student submits a proposal for a problem, preferably relevant
with introductory material for integrating numerical methods, to his or her research area, that demonstrates a thorough
including the finite element method, into electromagnetic understanding of the FE method and which requires the
(fields) courses at the undergraduate and graduate levels have approval of the instructor.
been published [3]-[15]. This paper presents the material
covered in a course at the graduate level. This material should 11. BACKGROUND
AND COURSE OUTLINE
enable the graduate student to use finite element analysis As prerequisites, the student should be familiar with meth-
in graduate research as well as when needed later in the ods for solving systems of algebraic equations, should have an
workplace. introductory electromagnetic field theory course, and should be
This course has two main objectives. The first is to help proficient in computer programming. The course outlined by
the student understand the theory behind the FE method, its topics is given in Table I. The course is divided into two parts.
limitations, and the approximations used in applying it to solve The first part covers the theoretical background which is es-
the field problem in practical devices. The second objective is sential in learning how to model electromagnetic devices. This
to enable the student to be a user of the FE method to model includes a review of the basic definitions of vector and scalar
electromagnetic devices by using software (code) developed algebra, as well as the properties of vector fields. In addition,
for the course as well as commercially available software this part includes an overview of the constitutive relations
packages [16], [17]. In doing so, the student can better use and definitions of magnetic quantities, material properties,
the commercially available software packages, modify existing boundary conditions, and Maxwell’s equation in point and
codes, or even develop a code that best suits a class of integral forms. The second part stresses the practical aspects
problems. for method implementation, and involves the application of
In order to fulfill these objectives, the finite element formu- the finite element method to engineering and design problems.
lations for several types of analysis which include static and The course material includes class notes developed specifically
sinusoidally time-varying cases are presented in the course for this course [20], as well as material developed for a short
and are implemented in several class projects. In addition, the course which is presented on an annual basis at Marquette
source code for a two dimensional (2D) linear magnetostatic University [21]. In addition, the book of reference [22] is
finite element analysis program which was developed for this adopted for the course. In the following sections of this paper,
course is discussed in class and is given to the students for use a summary of the theory covered, a list of projects assigned
in the first class project. The remainder of the projects which to students, and a list of sample final (personal) projects
cover other types of analysis require the modification of the developed by the students are presented.
code as well as the verification of the results by comparison to
closed form solutions and results from commercially available 111. FINITEELEMENTANALYSIS
Manuscript received February 1993. The partial differential equation (p.d.e) governing the elec-
The author is with the Department of Electrical and Computer Engineering,
Marquette University, Milwaukee, WI 53233. tromagnetic field problem is derived from Maxwell’s equations
IEEE Log Number 9208302. and constitutive relationships. The closed form solution of
0162-8828/93$03.00 0 1993 IEEE

Authorized licensed use limited to: University of Grenoble Alpes. Downloaded on December 08,2023 at 10:34:24 UTC from IEEE Xplore. Restrictions apply.
 234 IEEE TRANSACTIONS ON EDUCATION, VOL. 36, NO. 2, MAY 1993

Contour C n

Region R r” = “rl -7
X
Fig. 1. Finite element grid.
Fig. 2. Elemental coefficients.

these problems is often complicated and almost impossible be-

cause of many factors. One important factor is the complicated are derived. The region is discretized into subregions called
geometries and boundary contours in most of the practical “elements”. The behavior of the potential within each element
problems. Another factor is that the p.d.e is nonlinear in many is usually expressed by means of a polynomial interpolation
cases. With the advent of computers and numerical techniques, function. In the 2D magnetostatic case, using the mvp and first-
methods that are based on variational formulations [23] are order triangular elements, a linear interpolation polynomial of
used to solve the field problem. These methods allow one to the form [l], [2], [21], [22]:
replace the problem of integrating the p.d.e. governing the field
problem by the equivalent problem of seeking a functional and
minimizing it [l], [2]. The FE method is based on variational j=l,m,n

formulation of the field problem where the fields are obtained is used. Here, A, is the area of the eth element, and the
in a solution region through the minimization of an energy indexes, Z,m, and n are the nodal numbers of the element
functional. The minimization of this functional is equivalent vertexes with respect to the global system as shown in Fig. 1.
to solving the original partial differential equation governing The geometric coefficients, p j , q j , and rj are functions of the
the field. coordinates of the elemental vertexes, as shown in Fig. 2, and
Several types of analysis are obtained using the finite are given as follows:
element method. Those can be categorized as follows [24]:
Pl =XmYn - XnXm 41 = Ym - Yn r1 = X n - X m
1) static; electric or magnetic with no time variations,
2) sinusoidally time varying; fields and excitations are Pm = XnY1 - X l Y n Q m = ?/n - y l rm = 21 - X n
Pn = X L Y m - X m Y 1 Q n = y1 - Ym rn= X m - 21.
sinusoidal functions of time,
3) transient; the solution requires stepping in time, (3)
4) eigenvalue problem; used in high frequency applications
to determine natural and resonant frequencies of a system. The set of equations that approximates the minimization
As can be appreciated, these analysis types cannot be of the functional with respect to the set of magnetic vector
thoroughly covered in a one semester course. Accordingly, the potentials and their variations in the field region R can be
different types are discussed briefly, including the formulations written as follows:
for three dimensional problems [21], [25]. However, the theory
and derivations for only the two dimensional (2D) static and Snnxnn . Ann = Inn (4)
sinusoidally time-varying analysis types are covered in detail.
where is the global coefficient matrix, nn is the total number
The cases covered in this course, their corresponding dif-
of nodes, A is the vector of nodal mvps in a given grid
ferential equations, and the resulting FE equations are given
covering the field region R, and Inn consists of the nodal
below.
currents, which are the forcing functions of the system of
1) Magnetostatic; the governing p.d.e. is given as follows:
equations given in (4). The kth entry of 1is defined as:

Ik = C Ae 7
e
Je
(5)
where A is the z-component of the magnetic vector potential
(mvp), U is the magnetic reluctivity, and J is the z-component where, e ranges over all elements sharing the kth node and
of the excitation current density vector. Je is the current density in an element, e. The matrix S
As was mentioned above, the FE method is based on vari- is sparse, banded, symmetric, and singular. Hence, in order
ational formulation of the field problem where the fields are to remove this singularity and to obtain a unique solution,
obtained in a solution region through the minimization of an the boundary conditions (BCs) which depend on the physical
energy functional. The algebraic equations which approximate nature of a given problem and the chosen grid, must be applied.
the minimization of the energy functional with respect to the Three types of BCs are discussed in this course. Those are the
potential and its variations over a given region, R, Fig. 1, Neumann, Dirichlet, and periodicity BCs.

Authorized licensed use limited to: University of Grenoble Alpes. Downloaded on December 08,2023 at 10:34:24 UTC from IEEE Xplore. Restrictions apply.
 IEEE TRANSACTIONS ON EDUCATION, VOL. 36, NO. 2, MAY 1993 235

The jkth element (entry) of the 3 matrix is defined as

follows:

where, e ranges over all the elements sharing the j t h and kth
nodes.
2) Electrostatic; the governing p.d.e. is given as follows:

where 11, is the electrostatic scalar potential, E is the permit- Fig. 3. Elemental contribution of PM material.
tivity, and p is the electrostatic potential dependent charge
density, which represents the forcing function Of the system. The jkth element (entry) of the matrix for the electrostatic
The set of equations for the electrostatic case can be written
can be written as:
as follows:

where 9 is the vector of nodal scalar electric potential (volt- where r,, is the radius at the centroid of an element. Mean-
age) in a given grid covering the field region R. The kth entry while, for magnetostatic cases, the jkth element (entry) of the
Q which is the forcing function of the system is defined as:
of - matrix can be written as [20], [26]:

(9)

where p is the nodal electrostatic potential dependent charge where q; = qj + (2A/3rce).

density. In some types of analysis, the system of algebraic (FE)
The jkth element (entry) of the S matrix is defined as equations can be nonlinear. For instance, the set of equations
follows: in the magnetostatic case is nonlinear if one accounts for
magnetic material nonlinearities. Accordingly, methods to
model the B-H curve, such as the cubic spline [21], as well
as techniques to solve systems of nonlinear equations, such as
where E , is the permittivity. Equation (8) can be used to solve the Newton-Raphson [22], are discussed and implemented in
for the electrostatic Poissonian equation and the electrostatic this course. In addition, models to represent permanent magnet
Laplacian equation. For the Laplacian case, p e is set to zero, in (PM) materials are presented. In one of the models discussed,
which case the known potentials are used to rearrange (8) and the contribution of the elements representing PM materials to
solve for the unknown potentials in the solution domain R. the 2 matrix and the forcing function term are determined at
3) Sinusoidally Time Varying; the governing p.d.e. is given the elemental level. In other words, if an element in region
as follows: R is representing a PM material, in which the direction of
magnetization is given by a unit magnetization vector m as
shown in Fig. 3, the corresponding elemental contribution to
S is as follows [21]:
-
For this case, (1) is modified to include the induced eddy
current term, iwaA, where, i is the imaginary unit for complex
numbers, and u is the electrical conductivity. Also, the mvp,
A , and the excitation current density vector J are expressed where U, = H,/B,, and H,, and B, represent the PM
as complex phasors and the material is assumed to be linear. coercivity and residual flux density, respectively. In addition,
In this case, the jkth element (entry) of the S matrix is the values of the nodal currents, 4, I,, and I,, at the three
defined as follows: vertexes of the element, 1,m, and n, Fig. 3, which represent
sjk =
e
"(qjqk
4Ae
+ rjrk) + iwaAm/l2 (12) the elemental contribution to the forcing function are as
follows:
1
where m = 1 if j # k and m = 2 if j = k .
4) Axisymmetric; in cases 1)-3) above, the Cartesian coordi-
a
4 = Hc[mzrl - myal
nate system is used, and the fields were assumed to be constant 1
Im = - Hc[mxrm - myqm]
along the z-axis. In another class of applications, the field can 2
1
be assumed constant about an axis. In such cases, the p.d.e. In = - &[m,r, - my%] (16)
given above are expanded using the cylindrical coordinate 2
system. where m, = cos 8 , and my = sin Om, see Fig. 3.

Authorized licensed use limited to: University of Grenoble Alpes. Downloaded on December 08,2023 at 10:34:24 UTC from IEEE Xplore. Restrictions apply.
 236 IEEE TRANSACTIONS ON EDUCATION, VOL. 36, NO. 2, MAY 1993

TABLE I1 encouraged to develop their own programs as can be seen

SAMPLE
CLASS PROJECTS from the list of final projects Table 111.
- Analysis of the fields in a transformer
One gratifying result during the period this course has been
- Fields and inductance calculations for a cable system taught at Marquette University is the number of students who
- Calculation of the electric fields and the capacitance for a coaxial cable are using finite element analysis and the areas of research
system
- Determination of eddy currents power loss in a damper bar
where it is applied. A list of some of the final projects per-
- Determination of inductances in a PM device formed by students is given in Table 111. As can be appreciated
from the list of individual projects, the students are discovering
the advantages of this powerful technique and are applying it
TABLE 111 in different areas of research.
SAMPLE
FINALPROJECTS

- Finite element eddy current analysis for nondestructive testing

- Fields and action potentials in the cell membranes of nerves V. CONCLUSION
- Nonlinear analysis of the fields in a transformer
- Modeling a brushless DC motor A graduate course on finite element analysis for electromag-
- 3-D modeling of an electrostatic device
- Modeling of an axisymmetric PM device
netic applications which is offered at Marquette University
- Calculation of even and odd mode impedances for coupled geometries was presented. The course stresses the practical aspects for
- Development of a preprocessor for FE analysis method implementation, and involves the application of the
- Development of a PC based postprocessor for FE analysis
finite element method to engineering and design problems.
A summary of the formulations covered in this course was
discussed. In addition, a list of sample projects which require
One of the main objectives of this course is to help the the modification of a computer code by the student were
student understand the theory behind the FE method so he presented. Based on the experience gained from teaching this
or she can understand the limitations of this approach, as course, it is concluded that the use of the finite element method
well as the approximations used in applying this method. raises the student interest in traditional areas of research such
Accordingly, the details of implementation of this method as electromagnetics and power.
are covered in the course through the assignment of several
projects. The majority of these projects requires modification
of a computer code to determine the fields for the analysis
cases given above.
REFERENCES
[I] P. Silvester and M.V.K. Chari, “Finite element solution of saturable
magnetic field problems,” IEEE Trans. Power Appar. Sysr., vol. PAS
IV. CLASSPROJECTS 89, pp. 1642-1652, 1970.
[2] M. V. K. Chari, “Finite element analysis of nonlinear magnetic fields,”
As was stated earlier, one of the main objectives of this Ph.D. dissertation, McGill Univ., Montreal, P.Q., Canada 1970.
course is to enable the student to use commercially available [ 3 ] P. Silvester, Modern Electromagnetic FieZds. Englewood Cliffs, NJ:
Prentice-Hall, 1968.
software packages [16], [17], modify existing codes and even
. _R. Harrington, Field Comtutation by the Moment Method. Florida:
141
develop a code most suitable for a class of applications. c i e g e r , 1968.
Accordingly, several projects are assigned to the students in [5] E. Della Torre and W. Kinsner, “Teaching undergraduates to solve field
problems numerically,” IEEE Trans. Educ., vol. 20, no. 1, .. pp. 52-54,
addition to a final project which is picked individually. The Feb. 1977.
projects cover 1) the magnetostatic analysis, 2) the electrostatic (61 P. Silvester and N. Rumin, “In defense of canned programs,” IEEE
analysis, 3) axisymmetric analysis as applied to electrostatic Trans. Educ., vol. 20, no. 1, pp. 68-70, Feb. 1977.
[7] A.M. Lohse, P.B. Johns, and A Wexler, “Computer graphics for
and magnetostatic cases, and 4) sinusoidally time-varying transient fields,” IEEE Trans. Educ., vol. 20, no. I, pp. 64-68, Feb.
cases (eddy current problem), as well as the problems which 1977.
[XI J.F. Hoburg and J.L. Davis, “A student-oriented finite element pro-
include permanent magnet materials. The students are given gram for electrostatic potential problems,” IEEE Trans. Educ., vol. 26,
the source code for a simple 2D linear magnetostatic case, pp. 138-142, NOV. 1983.
which is based on (1)-(6) for use in the first project. The [9] S. R. H. Hoole and P. R. P. Hoole, “Finite element programs for teaching
electromagnetics,” IEEE Trans. Educ., vol. 29, no. 1, pp. 21-26, Feb.
other projects require the modification of the code, based on 1986.
(7)-(16), to cover different types of analysis. Table 11 includes 101 S. R. H. Hoole, “Teaching electromagnetics through finite elements; Part
a list of sample projects that cover different types of analysis. I: The rationale,” Int. J . Elect. Eng. Educ., vol. 25, pp. 33-49, Jan. 1988.
111 S. R. H. Hoole, P. R. P. Hoole, S. Jayakumaran, and N. R. G . Hoole,
Many of these projects require the calculation of parameters “Teaching electromagnetics through finite elements; Part 11: The pro-
such as energy and equivalent inductance (project #2), capac- gram,” Int. J . Elect. Eng. Educ., vol. 25, pp. 151-161, Apr. 1988.
itance (project #3), equivalent impedances and losses (project 121 S. R. H. Hoole, “A novel proof of natural boundary conditions for the
Poisson equation,” IEEE Trans. Educ., vol. 31, no. 1, pp. 4-8, Feb.
#4). The class projects require the verification of the results by 1988.
comparison to closed form solutions and corresponding results [I31 M.F. lskander et al. “A new course on computational methods in
electromagnetics,” IEEE Trans. Educ., vol. 31, pp. 101 - 115, May 1988.
from commercially available software packages [ 181, [ 191. In [14j M. N. 0. Sadiku, “A simple introduction to finite element analysis of
addition, as shown in the course outline, Table I, preprocessing electromagnetic problems,”IEEE Trans. Educ., vol. 32, pp. 85-93, May
and postprocessing are covered in the course. In doing so, 1989.
[I51 M. N. 0. Sadiku, A. Z. Makki, and L. C. Agba, “A further introduction
a graphics program is given to the students to plot the FE to finite element analysis of electromagnetic problems,” IEEE Trans.
grids and the fields for a given device and the students are EdUC., vol. 34, pp. 322-328, NOV. 1991.

Authorized licensed use limited to: University of Grenoble Alpes. Downloaded on December 08,2023 at 10:34:24 UTC from IEEE Xplore. Restrictions apply.
 IEEE TRANSACTIONS ON EDUCATION, VOL. 36, NO. 2, MAY 1993 237

2. J. Cendes, “Unlocking the magic of Maxwell’s equations,” IEEE transformers,” Conference Paper A77-122-5, IEEE-PES Winter Meeting,
Spectrum, pp. 29-33, Apr. 1989. New York, Jan. 1977.
D. G. Swanson, “Simulating EM fields,” IEEE Spectrum, pp. 34-37,
Nov. 1991.
The MacNeal Schwendler Corporation, ‘‘MSCEMAS,” Finite Element
Software Package. Abd A. Arkadan (S’79-M’88-SM’91) received
Ansoft Corporation, “Maxwell,” Finite Element Software Package.
the B.S. degree from the University of Mississippi,
A. A. Arkadan, “EECE 282 Class Notes,” Marquette Univ., Milwaukee,
in 1980, the M.S. degree from Virginia Polytechnic
WI.
A.A. Arkadan, Ed., Short Course Notes, “Finite Elements for Electro- Institute, Blacksburg, in 1981, and the Ph.D. degree
magnetic Applications,” May 1989-1992, Marquette Univ., Milwaukee, from Clarkson University Potsdam, NY, in 1988, all
Wl. in electrical engineering.
S. R. H. Hook, Computer-AidedAnalysis and Design of Electromagnetic From 1981 to 1984, he worked in industry. In
Devices. New York Elsevier, 1989. 1988, he joined the Department of Electrical and
S. G. Mikhlin, Variational Methods in Mathematical Physics. New Computer Engineering at Marquette University as
York: Macmillan, 1964. an Assistant Professor. His interests include de-
J. R. Brauer, Ed., What Every Engineer Should Know About Finite sign, analysis, and development of electronically-
Element Analysis. New York: Marcel-Dekker, 1988. operated machine systems and drives, and computer-aided solution of elec-
N.A. Demerdash and R. Wang, “Theoretical and numerical difficul- tromagnetic field problems in electromagnetic devices.
ties in 3-D vector potential methods in finite element magnetostatic Dr. Arkadan is a member of ASEE, Sigma Xi, Phi Kappa Phi, Pi Mu
computations,” IEEE Trans. Mag., vol. 26, pp. 1656-1658, Sept. 1990. Epsilon, Eta Kappa Nu, and Tau Beta Pi. Arkadan is a Senior Member of the
J. R. Brauer, “Finite element analysis of electromagnetic induction in IEEE where he is active in several committees and subcommittees.

A Unique Instructional Tool for

Visualizing Equipotentials and its
Use in an Introductory Fields Course
R. F. Lyvers and B. R. Horowitz, Member, IEEE

Abstract-This paper describes Charge Master, an IBM PC- provide an added dimension to capture student interest. The
compatible software package that aids in visualizing equipoten- idea behind an early computer game [5] based on electron tra-
tials produced by systems of point charges. It accomplishes this jectories in a space containing fixed charges could serve such
via an educational game and an option that plots equipotentials.
An assignment designed to accompany the software and its use a purpose.’ While some efforts focused on providing software,
in an introductory fields course are also described. The software others have produced textbooks (e.g., [6], [7]) and courses [8]
has been well received by students. that focus on the underlying computational methods.
The present paper describes Charge Master [9], [lo], an IBM
I. INTRODUCTION
PC-compatible, menu-driven, software package, designed for
use in a first course in electromagnetics, that aids in visualizing

R ECENT work in the teaching of electromagnetics [ 11 has equipotentials produced by coplanar point charges? Charge
underscored the value of using software for solving prob- Master consists of an educational game which requires careful
lems and visualizing fields. Examples include an NSFDEEE thought if it is to be played effectively and an option that
software dissemination project [2] and the availability of produces labeled equipotential plots. The latter can be used
software to complement textbooks [3]. There is also a wide to explore “what-if‘ scenarios and to create materials for
variety of commercial software for electromagnetic simulation classroom use.
and design [4] that is often available for educational use at
steep discounts. The use of software permits the presentation
of a larger number of examples in greater depth and richness 11. T H E GAME
of visual detail than is possible in traditional classrooms, texts The object of the game, whose inspiration came from
or exercises. It also facilitates experimentation and permits SUPERCHARGE [ll],a board game developed to present the
the simulation of dynamic events. Educational games can concepts of equipotentials in an enjoyable way, is to find the
location of the zero-potential contour in the common plane of
Manuscript received February 1993.
R. F. Lyvers was with the Department of Electrical Engineering, University
of Louisville, Louisville, KY; he is now with the Application Business ‘Whereas educational games must use accurate modeling, this game was
Systems Division, IBM Corporation, Rochester, MN 55901. apparently not intended for that use and employed altered physical laws to
B. R. Horowitz is with the Department of Electrical Engineering, University obtain desired game dynamics.
of Louisville, Louisville, KY 40292. Please address all correspondence regarding availability of the software to
IEEE Log Number 9208294. B. R. Horowitz.

0162-8828/93$03.00 0 1993 IEEE

Authorized licensed use limited to: University of Grenoble Alpes. Downloaded on December 08,2023 at 10:34:24 UTC from IEEE Xplore. Restrictions apply.