Вы находитесь на странице: 1из 18

MEE3017 Computer Modeling Techniques in Engineering

Chapter 6 Finite Element Method (FEM)


6.1 Introduction to FEM
History and Development
Basic Finite Element Concepts
Electromagnetic Analysis
Program Techniques and Software Package
6.1.1 History and Development
Turner, Clough, Martin, and Topp introduced the finite element concepts 1956. Almost
simultaneously, Argyris and Kelsey developed similar concepts in a series of publications
on energy theorems. Courant, Hrennikoff and McHencry are also early precursors of the
finite element methods.
Starting in 1967, many books have been written on the finite element method. The three
editions of the book authored by Prof. Zienkiewicz received worlwide diffusion. Gallagher,
Rockey et al., as well as the books written by Absi and Imbert.
The first FEM book (in English) for electrical engineer written by P. P. Silvester is
published in 1983.
1. Who and When
Any computer codes in the early 1960s for solving structural analysis problems in a manner
similar to what is now called the finite element approach existed only within the aircraft
industry, which was one of the earliest motivated to have improved analysis methods.
Turner, Clough, Martin, and Topp introduced the finite element concepts 1956. Almost
simultaneously, Argyris and Kelsey developed similar concepts in a series of publications
on energy theorems. Courant, Hrennikoff and McHencry are also early precursors of the
finite element methods.
Starting in1967, many books have been written on the finite element method. The three
editions of the book authored by Prof. Zienkiewicz received worldwide diffusion.
Gallagher, Rockey et al., as well as the books written by Absi and Imbert.
The first FEM book (in English) for electrical engineer written by P. P. Silvester is
published in 1983.

2. What and Why
The principles used to establish valid equations describing the behavior of the engineering
problem at hand include equilibrium, Newton's idea regarding force acting on amass,
Chapter 6 Page 1
MEE3017 Computer Modeling Techniques in Engineering
potential energy, strain energy, conservation of total energy, virtual work, thermo-
dynamics, conservation of mass, Maxwell's equations, and many more.
The problem seemed to always be that once all the hard work of formulating the problem
was complete, solving the resulting mathematical equations (sometimes linear and often
nonlinear partial differential equations), was almost impossible.
Ritz had extended his idea to use different geometric regions, establish separate
approximating functions in each region, and then hook them together. This is precisely
what finite elements are about.
The idea had to wait until modern digital computers took away the fear of large numbers
of algebraic equations. Soon matrices and matix methods of organizing large numbers of
algebraic equations where brought into the finite element approach; recall that the word
"matrix" was in the title of the important Air Force Conference in 1965.
Now mathematicians reentered the scene to remind the engineers of all the methods of
solving matrix equations from linear algebra.
Restated, the finite element method is one wherein the difficulty of mathematically solving
large complex geometric problems is transformed from a differential equation approach to
an algebraic problem, wherein the building blocks or finite elements have all the complex
equations solved for their simple shape (say a triangle, rod, beam, etc.)
The representation of the relationship of the important variables for the little, but not
infinitesimal, element is determined through a Rayleigh or Ritz approach just for each
element. Once this is done, a matrix of size equal to the number of unknowns for the
element can be produced which represents the element.
Infect, the method can be used to solve almost any problem that can be formulated as a
field problem. The development of additional software products and use within industry
has taken the last 25 or 30 years.
The area still under the early phase of use is probably the electromechanical and
electromagnetic area, probably owing to delayed recognition of need and transfer of
knowledge from structural to electrical engineers. today's increased world competition, the
pressure to design for electromagnetic compatibility, and new electronic and
communication devices.


6.1.2 Basic Finite Element Concepts
1. Field Analysis
Engineering design is aided by engineering analysis, the calculation of performance of a
trial design. To predict the performance it is often necessary to calculate a field, which is
defined as a quantity that varies with position within the device analyzed.
Chapter 6 Page 2
MEE3017 Computer Modeling Techniques in Engineering
There are many kinds of fields,and each field has a different in fluency on the device
performance.
Table 2.1 Various Aspects of Performance
________________________________
Field Potential
_________________________________
Heat flux Temperature
Mechanical stress Displacement
Electric field Voltage
Magnetic field Magnetic
vector potential
Fluid velocity Fluid potential
_________________________________

Various Problems in Electromagnetic
Fields:

Static Fields
Electrostatic and magnetostatic field calculations for both linear and non-linear
problems

Quasi-Static Fields
Time-dependent fields, including the transient and steady state behavior of
electromagnetic devices, eddy currents and skin effect

Wave Propagation
Wave propagation problems including microwaves and antennas, Scattering and
radiation

Optimization
Optimization using deterministic and stochastic methods, inverse problems, AI
applications, neural networks

Material Modeling
Modelling of material properties covering superconducting, composite, and microwave
absorbing materials and the numerical treatment of anisotropy, semi-conductor,
hysteresis, permanent magnets
Chapter 6 Page 3
MEE3017 Computer Modeling Techniques in Engineering

Coupled Problems:
Moving boundary problems, as well as electromagnetic fields coupled to mechanical,
electronic, thermal and/or flow systems

2. Finite Element Modelling
Calculation of all the above fields and potentials can be performed using finite element
analysis. The analysis begins by making a finite elements model of the device. The model
is an assemblage of finite elements, which are pieces of various sizes and shapes.
The finite element model contains the following information about the device to be
analyzed:
geometry, subdivided into finite elements
materials
excitations
constraints
Materials properties, excitations, and constraints can often be expressed quickly and easily,
but geometry is usually difficult to describe.
Figure 1 shows a typical engineering problem that happens to be a static thermal or heat
transfer problem.



Copper wire
Rubber sheath



Fig. 1 Cross-section of current-carrying copper wire with a rubber cover sheath.
The wire and its current extend into and out of the page.

To perform finite element thermal analysis, the finite element model shown in Fig.2 was
constructed by dividing the device into finite elements.


Chapter 6 Page 4
MEE3017 Computer Modeling Techniques in Engineering







Fig.2 Two-dimensional finite element model of Fig.1.

3. Energy Functional Minimization and Galerkin's Method
All the desired unknown parameters in the finite element model is by minimizing an
energy functional. An energy functional consists of all the energies associated with the
particular finite element model.
The law of conservation of energy is that the total energy of a device or system must be
zero. Thus, the finite element energy functional must equal zero.
The finite element method obtains the correct solution for any finite element model by
minimizing the energy functional. Thus, the solution obtained satisfies the law of
conservation of energy.
The minimum of the functional is found by setting the derivative of the functional with
respect to the unknown grid point potential to zero.
It is known from calculus that the minimum of any function has a slope or derivative equal
to zero. The basic equation for finite element analysis is

0 =
dp
dF
(2.1)
where F is the functional (energy) and p is the unknown grid point potential to be
calculated. The above simple equation is the basis for finite element analysis. The
functional F and unknown p vary with the type of problem.
In variational calculus the functional is shown to obey a relationship called Euler's
equation. Substitution in the appropriate Euler's equation yields the differential equation of
the physical system. Thus, the finite element solution obeys the appropriate differential
equation.
Chapter 6 Page 5
MEE3017 Computer Modeling Techniques in Engineering
The Euler's equation can not be always found in some differential equations. In this case,
we have to use Galerkin's method to find the discretized equation.
G P N dxdy
i i
S
*
= = 0 (2.2)
where N
i is interpolation function.
(a) one-dimensional elements
Linear(2) Quadratic(3)
Cubic(4)
(b) Two-dimensional elements
Triangular elements
Linear(3) Quadratic(6)
Cubic(9)
Quadrilateral elements
Linear(4) Quadratic(8)
Cubic(12)
(c) There-dimensional elements
Tetrahedronal elements
Linear(4) Quadratic(10) Cubic(16)
Chapter 6 Page 6
MEE3017 Computer Modeling Techniques in Engineering
Hexahedron elements
Linear(8) Quadratic(20)
Cubic(32)
Prismatical elements

Linear(6) Quadratic(15)
Cubic(24)

4. Two Dimensional Finite Elements
Two-dimensional finite elements connect three or more grid points lying in a two-
dimensional plane as shown in Fig.2. We will briefly derive the equations for a triangular
finite element modeling a physical system that obeys Poisson's differential equation.
Poisson's equation in two dimensions is
d
dx
k
dT
dx
d
dy
k
dT
dy
P + = (2.3)
where x and y are two dimensions and k is a material property.
Equation (2.3) governs static tempera-turns T, in which case P is power input per unit
volume. It also governs static electric or magnetic fields, in which case T is potential ( or
A) and P is charge density (Q) or current density (J), respectively.
The energy functional for all these physical problems is
F k del T ds PT dS
s s
=
|
\

|
.
|

(

|
\

|
.
|

1
2
1
2
2
( ) (2.4)
Chapter 6 Page 7
MEE3017 Computer Modeling Techniques in Engineering
where S is the surface area in the two dimensions. The first term of Eq. (2.4) is the energy
stored in the cases of electric or magnetic fields and is related to power dissipated in the
case of thermal fields.
The second term is the input energy in the cases of electric or magnetic fields and is related
to power input in the case of thermal fields.
The first term involves the gradient
G d (6.x 2.5) el T T = =
G x y
T
x
u
T
y
u
x
( , ) = +
y

(6.x 2.6)
where u
x
and u
y
are unit vectors.
The simplest type of two-dimensional finite element assumes a linear, or firs-order,
variation of the unknown potential T over the element.
y

x
Fig.3 Triangular finite element in the xy plane.
Within this first-order element T is related to the three unknown T values at the three
triangular grid points according to
T T a b x c
k k k k
k m n
= + +
=
y (
, , l
) (2.7)
Evaluating Eq. (2.7) at the three vertices gives the solution for the a, b, c coefficients:

a a a
b b b
c c c
x y
x y
x y
l m n
l m n
l m n
l l
m m
n n

(
(
(
=

(
(
(

1
1
1
1
(2.8)
Substituting Eq.(2.7) in Eq. (2.5) gives
Chapter 6 Page 8
MEE3017 Computer Modeling Techniques in Engineering
G x (2.9) y b T u c T u
k k k k k k
k m n
( , ) [ ]
, ,
= +
=

l
Thus, the temperature gradient is constant within a particular triangular finite element.

The grid point potentials Tk can be found by minimizing the functional (2.4), where
ds=dxdy. Substituting Eq. (2.9) and (2.4) in Eq. (2.1) and considering one triangular finite
element yields

T
kG PT
ds j l m n
j s
2
2 2
0

(
= = , , (2.10) ,
)
)
)
Carrying out the integration over the triangle can be shown to yield the 3-by-3 matrix
equation
[S][T]=[P] (2.11)
where the "stiffness" matrix is
[ ] (2.12)
( ) ( ) (
( ) ( ) (
( ) ( ) (
S k
bb c c bb c c bb c c
b b c c b b c c b b c c
b b c c b b c c b b c c
l l l l l m l m l n l n
m l m l m m m m m n m n
n l n l n m n m n n n n
=
+ + +
+ + +
+ + +

(
(
(

where is the area of the triangle,


= (2.13)

(
(
(
1
2
1
1
1
x y
x y
x y
l l
m m
n n
and the right-hand side is the "load vector":
[ ] (2.14) P
P
P
P
=

(
(
(

3
Equation (2.11) solves for the potential T in a region containing the one triangle with l, m,
n in Figure 3.
For practical problems with N nodes (grid point), the above process is repeated for each
finite element, obtaining a stiffness matrix or element coefficient matrix [S] with N rows
and N columns. [P] and [T] are then column vectors containing N rows.
5. Axisymmetric Finite Elements
Chapter 6 Page 9
MEE3017 Computer Modeling Techniques in Engineering
All devices designed by engineers are in reality three-dimensional. A special case of a
three-dimensional device is one that has axial symmetry. The basic equation can be
written as

k
r r
r
T
r
k
T
z
P

( ) +
2
2
= (2.15)







Fig. 4 Axisymmetric finite element.

6. Finite Elements in Three Dimension
Three dimensional Poisson's equation:


x
k
T
x y
k
T
y z
k
T
z
P ( ) ( ) ( ) + + = (2.16)

A three-dimensional finite element has at least four grid points, and they do not all lie in
one plane. The element forms a solid shape containing a volume of material.
The Two-dimensional and axisymmetric finite elements described above are used
whenever possible, because they are simpler to describe and uses than three-dimensional
finite elements.
Fig.4 shows the three most common three dimensional solid finite elements.




Chapter 6 Page 10
MEE3017 Computer Modeling Techniques in Engineering
Fig.4 Three-D finite elements:
The derivation of the matrix equation for the tetrahedral finite element is rather similar to
that of the triangular element. Thus in the case of quantities obeying Poisson's equation the
tetrahedral element is similar to the two-D triangular element.
Extending Eq. (2.7) gives
T T (2.17) a b x c y d
k k k k k
k m n o
= + + +
=
(
, , , l
z)
The derivation of the matrix equations proceeds in a fashion similar to that described in 2-
D.

7. Example of 2-D FEM

y =1.0 V

x =0 V

Boundary condition:
1=2=0.0
5=6=1.0

For the triangular element, the equations that must be satisfied for this linear representation
to agree with the functional values at the vertices are
Chapter 6 Page 11
MEE3017 Computer Modeling Techniques in Engineering
{ }

( ) e
e e e
e
e
e
x y x y = + + =

1 2 3
1
2
3
1 (2.18)
In matrix notation,
1
1
1
1 1
2 2
3 3
1
2
3
1
2
3
x y
x y
x y
e e
e e
e e
e
e
e
e
e
e

(
(
(

(2.19)
then

1
2
3
1 1
2 2
3 3
1
1
2
3
1
1
1
e
e
e
e e
e e
e e
e
e
e
x y
x y
x y

(
(
(

(2.20)

{ }

e
e e
e e
e e
e
e
e
x y
x y
x y
x y
=

(
(
(

1
1
1
1
1 1
2 2
3 3
1
1
2
3
(2.21)
The expression for (x,y) is equivalent to
{ }

e
e
e e e
e e e
e e e
e
e
e
x y
b b b
c c c
d d d
=

(
(
(

1
2
1
1 2 3
1 2 3
1 2 3
1
1
2
3

( )

(2.22)
where
b x y x y
ie je ke ke je
=

c y y
ie je ke
=
} (2.23)
d x x
ie ke je
=

and

b b b
c c c
d d d
x y
x y
x y
e e e
e e e
e e e
e
e e
e e
e e
1 2 3
1 2 3
1 2 3
1 1
2 2
3 3
1
2
1
1
1

(
(
(
=

(
(
(

( )
(2.24)
Chapter 6 Page 12
MEE3017 Computer Modeling Techniques in Engineering

( ) e
e e
e
e e
x y
x y
x y
=

(
(
(
1
2
1
1
1
1 1
2 2
3 3
e

(2.25)
Substitute interpolation functions
(2.26)
( ) e
ie ie
i
N =
=

1
3
where Nie are interpolation or shape functions.
N b c x
ie e ie ie ie
= + +
1
2
( )
( d y) (2.27)
Substitute Eq.(2.26) to the energy functional corresponding to Laplace's equation,
2
=0,
W dS
x y
dxdy = =
|
\

|
.
|
+
|
\

|
.
|

(
1
2 2
2
2
2

(2.28)
and then Eq.(2.28) can be written as
W
N
x
N
x
N
y
N
y
dxdy
i j
j i
i
j
i
j
S
e
= +
|
\

|
.
|
= =

1
2
1
3
1
3


( )
(2.29)
where
S
N
x
N
x
N
y
N
y
dxdy
ij
i
j
i
j
S
e
= +
|
\

|
.
|


( )
(2.30)
Nie/x and Nie/y are given as

N
x
c
ie ie
e
=
2
( )
(2.31)

N
y
d
ie ie
e
=
2
( )
(2.32)
dxdy
e
S
e
=
( )
( )
(2.33)
The energy functional W corresponding to i,W/i=0, we derive
(i=1,2,...,n) (2.34) S
ij j
j
n
=
=
0
1
The matrix form is shown as follows:
Chapter 6 Page 13
MEE3017 Computer Modeling Techniques in Engineering
[S]{}={0} (2.35)
where
(2.36) [ ] S
S S
S S
S
nn
=

(
(
(
(
11 12
21 22
L L
M O
M
{ }

1
2
M
n
(2.37)
where [S] is called element coefficient matrices.

For the single element, the equation can be written as

W
W
W
S S S
S S S
S S S
e
e
e
e
e
e
e e e
e e e
e e e
e
e
e
( )
( )
( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
1
2
3
11 12 13
21 22 23
31 32 33
1
2
3

(
(
(

(2.38)
where
S
c c d d
ij
e
ie je ie je
e
( )
( )
=
+
4
(2.39)
Substituting all of Sij and in Eq. (2.35), and then system matrix equation can be solved
by computer program.
For the four-element example, N=6, so the matrix equati0n s written as

(
(
(
(
(
(
(
(







0
0
0
0
0
0
2 1 1 0 0 0
1 2 0 1 0 0
1 0 4 2 1 0
0 1 2 4 0 1
0 0 1 0 2 1
0 0 0 1 1 2
2
6
5
4
3
2
1

Chapter 6 Page 14
MEE3017 Computer Modeling Techniques in Engineering
Transposing all known voltage values to the right-hand side of equation, then the unknown
nodal voltages are obtained as the solution of this equation. The results are
V 2 / 1
4 3
= =
If =1
) (
) (
) (
) (
4
4
e
je ie x je ie y d
ij
e
je ie je ie e
ij
d d c c
S
d d c c
S

+
=

+
=


Supermatrix and band matrix
The matrix can be stored in a rectangular array of dimension n(b+1) as shown below:
(
(
(
(

(
(
(
(
(
(
(
(
(
(
(
(

0 0 0
0 0
0
0 0 0 0 0
0 0 0 0
0 0 0
0 0
0 0
0 0
0 0 0
0 0 0 0
0 0 0 0 0
12
99 88 77 66 55 44 33 22 11
99 98 97 96
89 88 87 86 85
79 78 77 76 75 74
69 68 67 66 65 64 63
58 57 56 55 54 53 52
47 46 45 44 43 42 41
36 35 34 33 32 31
25 24 23 22 21
14 13 12 11
a
a a a a a a a a a
a a a a
a a a a a
a a a a a a
a a a a a a a
a a a a a a a
a a a a a a a
a a a a a a
a a a a a
a a a a

(
(
(
(

=
(
(
(
(

(
(
(
(

4
3
2
1
4
3
2
1
44 43 42 41
34 33 32 31
24 23 22 21
14 13 12 11
y
y
y
y
x
x
x
x
a a a a
a a a a
a a a a
a a a a

Chapter 6 Page 15
MEE3017 Computer Modeling Techniques in Engineering
(
(
(
(

=
(
(
(
(

(
(
(
(

1 41 4
1 31 3
1 21 2
1
4
3
2
1
44 43 42
34 33 32
24 23 22

0
0
0
0 0 0 1
x a y
x a y
x a y
x
x
x
x
x
a a a
a a a
a a a

Program Techniques
The procedure of numerical techniques:
1. Pre-processor
Mesh generation and material description by visual techniques
Dimensions and symmetrical models
Linearity and nonlinearity
2. Numerical analysis
FEM program flowchart. See Fig.1 FEM program flowchart.
Matrix solver.
FINITE ELEMENTS













Chapter 6 Page 16
S
RESULT
OUTPUT
EVALUATION OF FLUXES PER
ELEMENT
SOLUTION OF THE SYSTEM OF
EQUATIONS FOR ALL
ASSEMBLING OF THE TOTAL SYSTEM
OF EQUATION
EVALUATION OF ELEMENT MATRICES
INTRODUCTION OF THE ESSENTIAL
BOUNDARY CONDITIONS
DATA
MEE3017 Computer Modeling Techniques in Engineering








Electromagnetic Analysis
1. Laplace's Equation

2. Poisson's Equation

A special case of a three-dimensional device is one that has axial symmetry. The basic
equation in a cylindrical coordinate system can be written as

1
2 0
r r
r
A
r z
A
z
J
r z

( ) ( ) + =

System matrix equation can be expressed as

[S]{A}={K}

For a single element, K has the form of

K
J
i
e
e
=
0
3




where J
0
is the current density.


The coefficient matrix for a singal element is

Chapter 6 Page 17
MEE3017 Computer Modeling Techniques in Engineering


W u
W u
W u
S S S
S S S
S S S
A
A
A
K
K
K
e
e
e
e
e
e
e e e
e e e
e e e
e
e
e
e
e
e
( )
( )
( )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
/
/
/
1
2
3
11 12 13
21 22 23
31 32 33
1
2
3
1
2
3

(
(
(



where Sij can be expressed as

S
r
c c d d
ij
e
e
e
ie je ie je
( )
( )
( )
=
+

(
2
4
0



where


c z z
d r r
ie je ke
ie ke je
=
=



( )
r
r r r
e e e
0
1 2 3
3
=
+ +
( ) ( ) ( )



3. Helments' Equation

Program Techniques and Software Packages

Supermatrix and band matrix

1. Mesh generation and confirmation
2. Developing your own program
3. Post-process

Chapter 6 Page 18