The Finite Element

Method in
Engineering
Assoc. Prof. Dr. Ibrahim Abu-Alshaikh

Chapter 1. Introduction
What is finite element Method, FEM?
The FEM is a numerical procedure that can be used to obtain
solutions to a large class of engineering problems involving
stress analysis, heat transfer, electromagnetism, and fluid
flow. That is to obtain solutions for complex geometry and
loading-condition class problems
FEM is a convenient method numerical technique to find an
approximate solution of most of the practical problems.
However, in the absence of any other convenient method to
find even the approximate solution of a given problem, we
have to prefer the FEM.
Continued

 In the FEM, it will often be possible to improve or refine

the approximate solution by spending more
computational effort.
In the FEM, the solution region is considered as built up
of many small, interconnected sub-regions called finite
elements.
Although the FEM has been extensively used in the field
of structural mechanics, it has been successfully applied

to solve several other types of engineering problems,

such as heat conduction, fluid dynamics, seepage flow,
and electric and magnetic fields

Fixed boundary

uniform loading

Finite element
Cantilever
plate
model

in plane strain

Approximate method
Geometric model
Element Node
Element
Mesh
Discretization
Node

Problem: Obtain the

stresses/strains in the
plate

Example 1

The circumference of a circle (S) is

approximated by the perimeters of
inscribed and circumscribed n-sided
polygons as shown in Figure
Thus, the perimeters of the inscribed and
circumscribed polygons are given by

which can be rewritten as

ENGINEERING PROBLEMS
In general , engineering problems are mathematical models of
physical situations. Mathematical models of many engineering
problems are differential equations with a set of corresponding
boundary and/or initial conditions. When possible, the exact
solution of these equations renders detailed behavior of a
system under a given set of conditions. The analytical solutions
are composed of two parts:
1. a homogenous part and
2. a particular part.
The following Table contains some engineering problems that
have analytical solutions

Continued

Continued

In any given engineering

problem, there are two
sets of design parameters
that influence the way in
which a system behaves.
These parameters include
1. material and
2. geometric properties
such as modulus of
elasticity,
thermal
conductivity, viscosity, and
area, and second moment
of area.

Continued

Heat Transfer, Fluid flow and

electrical Examples

Continued

On the other hand, there are parameters that produce

disturbances in a system. These types of parameters are
summarized in the following Table. Examples of these
parameters include external forces, moments, temperature
difference across a medium, and pressure difference in fluid
flow.

NUMERICAL METHODS
There are many practical engineering problems for which we
cannot obtain exact solutions. In contrast to analytical
solutions, which show the exact behavior of a system at any
point within the system, numerical solutions approximate exact
solutions only at discrete (discretization the domain) points,
called nodes. There are two common classes of numerical
methods:
1. finite difference methods and
2. finite element methods.
Although finite difference methods are easy to understand and
employ in simple problems, they become difficult to apply to
problems with complex geometries or complex boundary
conditions. This situation is also true for problems with nonisotropic material properties. I n contrast, the finite element
method uses integral formulations rather than difference
equations to create a system of algebraic equations.

A BRIEF HISTORY OF THE FEM & ANSYS

Hrenikoff, 1941 frame work method
Courant, 1943 piecewise polynomial interpolation

Turner, 1956 derived stiffness matrices for truss,

beam, etc.
Clough, 1960 coined the term finite element

The rapid progress FEM can be seen by noting that: as

estimated in 1995 annually about 3800 papers, about 56,000
papers and 380 books and 400 conference proceedings
published [1.42].
In 1971, ANSYS was released for the first time.
ANSYS is a comprehensive general-purpose FEM computer
program that contains more than 1000,000 lines of code

BASIC STEPS IN THE FEM

Preprocessing Phase
1. Create and discretize the solution domain into FEs; that is, subdivide the
problem into nodes and elements.
2. Assume a shape function to represent the physical behavior of an
element; that is, a continuous function is assumed to represent the
approximate behavior (solution) of an element.
3. Develop equations for an element.
4. Assemble the elements to present the entire problem. Construct the
global stiffness matrix.
5. Apply boundary conditions, initial conditions, and loading. Solution Phase
6. Solve a set of linear or nonlinear algebraic equations simultaneously to
obtain nodal results, such as displacement values at different nodes or
temperature values at different nodes in a heat transfer problem.
Post-processing Phase
7. Obtain other important information. At this point, you may be interested in
values of principal stresses, heat fluxes, and so on.
Continued

In general, there are several approaches to formulating FE

problems:
1. direct formulation,
2. the minimum local potential energy formulation, and
3. weighted residual formulations.

1. DIRECT FORMULATION
The following problem illustrates the steps and the
procedure involved in direct formulation.

Example 2 Using direct formulation

Consider a bar with a variable cross
section supporting a load P, as shown
in the Figure. The bar is fixed at one
end and carries the load P at the other
end. Let us designate the width of the
bar at the top by w1 at the bottom by
w2, its thickness by t, and its length by
L. The bar's modulus of elasticity will
be denoted by E. We are interested in
approximating how much the bar will
deflect at various points along its
length when it is subjected to the load
P. We will neglect the weight of the bar
in the following analysis, assuming
that the applied load is considerably
larger than the weight of the bar:

Continued

1. Discretize the solution domain into finite elements. We

begin by subdividing the problem into nodes and elements

However, note that we can increase the accuracy of our

results by generating a model with additional nodes and
elements. This task is left as an exercise for you to complete.
Continued

2. Assume a solution that approximates the behavior of an

element. In order to study the behavior of a typical element,
let's consider the deflection of a solid member with a
uniform cross section A that has a length L when subjected
to a force F, thus, the average stress in the member is given
by
F

A
The average normal strain of the member is defined as the
change in length per unit original length of the member that
is:
L

L
Over the elastic region , the stress and strain are related by
Hooke's law, according to the equation
E
where E is the modulus of elasticity of the material.
Continued

Combining the above three equations and simplifying, we have

F
L
EA
E
E

F
L

A
L
L
Therefore, a centrally loaded member of uniform cross section
may be modeled as a spring with an equivalent stiffness of

EA
keq.

L
Thus, the bar is
represented by a
model consisting of
four elastic springs
(elements) in series;
as shown above
Continued

3. Develop equations for an element.

the elastic behavior of the ith element with nodes i and i + 1 is modeled by
an equivalent linear spring according to the equation

E Ai 1 Ai
EAavr
f keq ui 1 ui
ui 1 ui
ui 1 ui
2L
L

Ai and Ai+1 are the cross-sectional areas of the member at nodes i and i + 1
respectively, and L is the length of the element.

4. Assemble the elements: Static equilibrium

requires that the sum of the forces acting on
each node be zero. This requirement creates
the following five equations:
node 1: R1 k1 (u2 u1 ) 0
node 2: k1 (u2 u1 ) k2 (u3 u2 ) 0
node 3: k2 (u3 u2 ) k3 (u4 u3 ) 0
node4: k3 (u4 u3 ) k4 (u5 u4 ) 0
node5: k4 (u5 u4 ) P 0

Continued

or in matrix form as:

k1
k
1
0

0
0

k1
k1 k2
k2

0
k2
k2 k3

0
0

k3
0

or as
K u F

0
0
k3
k3 k4
k4

0 u1 R1
0 u2 0

0 u3 0

k4 u4 0
k4 u5 P

(stiffness matrix) (disp. Matrix) = (load matrix)

5. Apply boundary conditions

u1 0
Continued

6. Solve a system of algebraic equations simultaneously.

In order to obtain numerical values of the nodal
displacements, let us assume that
6
2
. t 0.125 in P 1000 lb L 10 in E 10.4 10 lb/in

w1 2 in, w2 1 in
Let us find the area

A( )

w
w1 w2
tan( )
0.05
2L

A( ) t (2 w w2 ) 0.125 (0.1 1)
A(7.5) A(10)
0.234375 in 2
2
A(5) A(7.5)
A2
0.203125 in 2
2
A(2.5) A(5)
A3
0.171875 in 2
2
A(0) A(2.5)
A4
0.140625 in 2
2
A1

Continued

 EA1 10.4 106 (0.234375)

3
k1

975

10
lb/in

2.5

L1
EA2 10.4 106 (0.203125)
3
k2

845

10
lb/in

2.5

L2
EA3 10.4 106 (0.171875)
3
k3

715

10
lb/in

2.5

L3
EA4 10.4 106 (0.140625)
3
k4

585

10
lb/in

2.5

L4
Continued

 975
975

0
0

975
0
0
975 845
845
0
845
845 715
715
0
715
715 585
585
0
0
585
585

0 0
R1
0
0 u2

3
0 u3 10 0

u
0
4

u5
P

0
u1

u 0.001025641026
2

u3 0.002209072978

u4 0.003607674377
u5 0.005317076086

R1 975 10 u2 1000
3

Continued

7. Obtain other important information; such as

the average normal stresses in each element.
E Ai 1 Ai
EAavr
f keq ui 1 ui
ui 1 ui
ui 1 ui
2L
L

f
E

ui 1 ui
Aav L
0.001025641026 0
1

4266.666666
0.002209072978 0.001025641026
4923.076920
6
10.4

10
2

lb/in 2

3 0.003607674377 0.002209072978
5818.181820
2.5

0.005317076086

0.003607674377
7111.111108

Example 3 Using direct formulation

A typical exterior frame wall (made up of 2 X 4 studs) of a
house contains the materials shown in the table below. Let us
assume an inside room temperature of 70F and an outside
air temperature of 20F, with an exposed area of 150 ft 2 . We
are interested in determining the temperature distribution
through the wall.

Continued

 BTU (British Thermal Units)

1 Btu / hour =0.00029307107 kilowatts

How to convert Fahrenheit to Celsius

0 degrees Fahrenheit is equal to -17.77778

degrees Celsius:

0 F = -17.77778 C
The temperature T in degrees Celsius (C) is equal

to the temperature T in degrees Fahrenheit (F)

minus 32, times 5/9:

T(C) = (T(F) - 32) 5/9

1. Discretize the solution domain into finite elements.

We will represent this problem by a model that has seven
nodes and six elements,

2. Assume a solution that approximates the behavior of an

element. In this example, two modes of heat transfer
(conduction and convection). When there exists a temperature
gradient in a medium, conduction heat transfer occurs. The
energy is transported from the high-temperature region to the
low-temperature region by molecular activities. The steady-state
thermal behavior of the elements (2), (3), (4), and (5) may be
modeled using Fourier's law. That is

Continued

T
qx kA
X
qx is the X-component of the heat transfer rate, k is the
thermal conductivity of the medium, A is the area normal to
T
heat flow and X is the temperature gradient.
For the ith element we can write

kA
q Ti 1 Ti
L
Or as

q UA Ti 1 Ti
k
U
L

where

Continued

The steady-state thermal behavior of elements (1) and (6) may be

modeled using Newton's law of cooling. Convection heat transfer
occurs when a fluid in motion comes into contact with a surface
whose temperature differs from the moving fluid. The overall heat
transfer rate between the fluid and the surface is governed by
Newton's law of cooling, according to the equation

q hA Ts T f

where h is the heat transfer coefficient, Ts is the surface temperature,

and Tf represents the temperature of the moving fluid. Newton's law of
cooling can also be written in terms U as

q UA Ts T f

where U = h, represents the reciprocal of thermal resistance due to

convection boundary conditions. Under steady-state conduction, the
application of energy balance to a surface, with a convective heat
transfer, requires that the energy transferred to this surface via
conduction must be equal to the energy transfer by convection. This
principle,

T
kA
hA Ts T f
X

Continued

3. Develop equations for an element.

In general, for conduction problems, the heat transfer
rates qi and qi+I and the nodal temperatures Ti and Ti+1
for an element are related according to the equations

kA
qi Ti Ti 1 ,
L

kA
qi 1 Ti 1 Ti
L

Continued

Or in matrix form as

qi K11
Q K T q K
i 1 12
(e)

( e)

K11
K
K12
(e)

K (e)

( e)

K12
1
(e) (e) 1
U A

K22
1 1

is the conductance matrix where

k (e)

For conduction heat transfer

U (e ) h( e)

For Convection heat transfer

U (e)

K12 Ti
K22 Ti 1

Continued

4. Assemble the elements to present the entire problem.

U1
q1
U
q
1
2
0
q3

A
0
4
0
q5

0
q6
0
q
7

U1
U1 U 2
U 2
0
0
0
0

0
U 2
U 2 U3
U 3
0
0
0

0
0
U 3
U3 U 4
U4
0
0

0
0
0
0
0
0
U 4
0
U 4 U5
U 5
U 5
U5 U6
0
U 6

0 T1
0 T2
0 T3

0 T4
0 T5

U 6 T6
U 6 T7

5. Apply boundary conditions and thermal loads.

Also note that, the heat transfer rate through each
element was caused by temperature differences
across the nodes of a given element. Thus, the
external nodal heat flow values are zero in the heat
flow matrix, that is q2 q3 q4 q5 q6 0
Continued

For the given problem, the exterior of the wall is exposed

to a known air temperature T1 and the room temperature,
T7, is also known. Thus, we want the first row to read T1 =
20F and the last row to read T7 = 70F. So, we have

0
20
1/150
0
5.88 7.11

0
0
1.23

150
0
0

0
0
0
0

0
0
0
70
0
0

0
1.23
1.99
0.76
0
0
0

0
0
0
0 T1
0
0
0
0 T2
0.76
0
0
0 T3

0.851
0.091
0
0 T4
0.091
2.311 2.22
0 T5

0
2.22
3.69 1.47 T6
0
0
0
1/150 T7

Or as
Continued



(5.88)20 7.11

0 1.23

0
0 0

1.47(70)

1.23
1.99
0.76
0
0

0
0
0.76
0
0.851
0.091
0.091 2.311
0
2.22

0 T2
0 T3

0 T4
2.22 T5

3.69 T6

6. Solve a system of algebraic equations simultaneously.

20
T1

T 20.58975226
2

T3 23.40905570

27.97187578
4

T5 66.07894447

T6 67.64099099
T

70

Continued

7. Obtain other information.

For this example, we may be interested in obtaining other information,
such as heat loss through the wall. Such information is important in
computing the heat load for a building. Because we have assumed
steady-state conditions, the heat loss through the wall should be equal
to the heat transfer through each element. This value can be
determined from the equation

q UA Ti 1 Ti

The heat transfer through each element is

20.58975226 20 (5.88)(150)

520.1614933
23.40905570

20.58975226
(1.23)(150)

520.1614847

27.97187578-23.40905570 (0.76)(150)

520.1614891
q 66.07894447 27.97187578 (0.091)(150)
Btu/hr
67.6409909 66.07894447 (2.22)(150) 520.1614877

520.1614912

70 67.64099099 (1.47)(150)
520.1614867

This problem is just another example of how we can

generate finite element models using the direct method.

EXAMPLE 4: Using direct formulation

In this example, let us consider a shaft that is made of two
parts, as shown in the Figure. The shaft is fixed at both
ends. A torque of T= 200 lb ft is applied at D.
Shear modulus of elasticity is given as
GAB 3.9 106 lb/in 2 , GBC 4 106 lb/in 2
Polar moment of inertia

J AB

(1.5)

in , J BC
4

(1) 4 in 4

32
32
Using FEM with three elements (direct formulation),
determine the angle of twist at D and B, and the torsional
reactions at the boundaries.

Continued

Recall from your previous study of the mechanics of

materials that the angle of twist for a shaft with a constant
cross-sectional area with a polar moment of inertia J and
length L, made of homogenous material with a shear
modulus of elasticity G, subject to a torque T is denoted
and given by

TL

GJ

we can show that for an element

comprising two nodes, the
stiffness matrix, the angle of
twists, and the torques are
related
according
to
the
equation
(e) (e)
T

G
J
1
(e)
(e)
( e)
T K T L(e)
2

1 1 1
1 1

2
Continued

So, for element (1), the stiffness matrix is

(1) (1)
G
J
(1)
K
L(1)

1 1 3.9 106 ( )(1.5)4

1 1 32(2.5)(12)

1 1
1 1

and its position in the global stiffness matrix is

1 1
6
4
3.9 10 ( )(1.5) 1 1
(G1)
K
32(2.5)(12) 0 0

0 0

0
0
0
0

0 1
0 2
0 3

0 4
Continued

So, for element (2), the stiffness matrix is

(2) (2)
G
J
(2)
K
L(2)

1 1 3.9 106 ( )(1.5)4

1 1
32(1)(12)

1 1
1 1

and its position in the global stiffness matrix is

0 0 0
6
4
3.9 10 ( )(1.5) 0 1 1
(G 2)
K
0 1 1
32(1)(12)

0 0 0

0 1

0 2
0 3

0 4
Continued

So, for element (3), the stiffness matrix is

(3) (3)
G
J
(3)
K
L(3)

1 1 4 106 ( )(1)4
1 1 32(2)(12)

1 1
1 1

and its position in the global stiffness matrix is

0
6
4
4 10 ( )(1) 0
(G3)
K
32(2)(12) 0

0 0 0 1

0 0 0 2
0 1 1 3

0 1 1 4
Continued

The final global matrix is obtained simply by assembling,

or adding, elemental descriptions:
(G1)
(G 2)
( G 3)

K


G

7.8975
0
0
7.8975

1106 ( ) 7.8975 7.8975 19.74375 19.74375 0

G
K 32(12) 0
19.74375
19.74375 2 2

0
2
2
0
7.8975
0
0 1
7.8975
7.8975 27.64125 19.74375 0
6
110 ( )
G
2

19.74375 21.74375 2 3
32(12) 0

0
0

2
2

T K
(G)

(G )

(G)

Continued

7.8975
0
0 1 T1
7.8975
7.8975 27.64125 19.74375 0 T
6
110 ( )
2 2
19.74375 21.74375 2 3 T3
32(12) 0

0
0

2
2

4 T4
Applying the fixed boundary conditions at points A and
C and applying the external torque, we have

7.8975
0
0 0 T1

7.8975
7.8975 27.64125 19.74375 0 200 12
6
110 ( )

0
19.74375 21.74375 2 3
32(12) 0

0
2
2 0 T4
0

Which is reduced to
Continued

32(12)

27.64125 19.74375 2 2400 6

10
(

19.74375 21.74375

3
0

1 0

-0.03020056731
2
rad
3 -0.02742270540

4 0
1106 ( )
T1
(7.8975)2 1951.297032
32(12)
1106 ( )
T4
(2)3 448.7029680
32(12)

Example 5 Using direct formulation

A steel plate is subjected to an axial load, as shown in the
Figure. Approximate the deflections and average stresses
along the plate. The plate is 1/16 in thick and has a
modulus of elasticity E 29 106 lb/in 2
We may model this problem using four nodes and four
elements, as shown.

Continued

Next, we compute the equivalent stiffness coefficient for

each element as:
(e) (e)
E
A
K ( e )
L( e )

1 1
1 1

where

F1 E (e) A(e)
F L(e)
2

1 1 u1
1 1 u

So, for element (1), the stiffness matrix is

(1) (1)
E
A
K (1)
L(1)

1 1 29 106 (5) 1 1
1 1 16(1) 1 1

and its position in the global stiffness matrix is

1 1
1 1
6
29 10 (5)
K (G1)
16(1) 0 0

0 0

0
0
0
0

0 u1
0 u2
0 u3

0 u4

Continued

So, for element (2,3), the stiffness matrix is

(2,3) (2,3)
E
A
(2,3)
K
L(2,3)

1 1 29 106 (2) 1 1
1 1 16(4) 1 1

and its position in the global stiffness matrix is

0 0 0
0 1 1
6
29 10 (2)
(G 2,G 3)
K

16(4) 0 1 1

0 0 0

0 u1

0 u2
0 u3

0 u4
Continued

So, for element (4), the stiffness matrix is

(4) (4)
E
A
(4)
K
L(4)

1 1 29 106 (5) 1 1
1 1 16(2) 1 1

and its position in the global stiffness matrix is

0
0
6
29 10 (5)
(G 4)
K
16(2) 0

0 0 0 u1

0 0 0 u2
0 1 1 u3

0 1 1 u4
Continued

The final global matrix is obtained simply by assembling,

or adding, elemental descriptions:
(G1)
(G 2)
( G3)
( G 4)

K



G

5
0
0
5

5
5

0.5

0.5

0.5

0.5
0
29 10

0.5 0.5 0.5 0.5 2.5 2.5

16 0

0
2.5
2.5
0

0 u1
5 5 0

1
0 u2
29 106 5 6

16 0 1 3.5 2.5 u3

0 0 2.5 2.5 u4

F K u
(G)

(G )

(G)

Continued

0
0 u1 F1
5 5
u F
6

5
6

1
0
29 10
2 2
16 0 1 3.5 2.5 u3 F3

0 0 2.5 2.5 u4 F4
Applying the fixed boundary conditions at the fixed
support and applying the external force, we have

0
0 0 F1
5 5
u 0
6

5
6

1
0
29 10
2
16 0 1 3.5 2.5 u3 0

0 0 2.5 2.5 u4 800
Which is reduced to
Continued



u 0
6
6

1
0
29 10
2
16 1 3.5 2.5 u3 0

0 2.5 2.5 u4 800

0
u1

u 0.08827586200
2
103 in
u3 0.5296551721

u4 0.7062068961

E
and the stresses in each element are ui 1 ui
L

2560
1
0.08827586200 0 /1
29 106 0.5296551721-0.08827586200 / 4 3200 lb/in 2

0.7062068961 0.5296551721 / 2 2560

3

2. MINIMUM TOTAL
POTENTIAL ENERGY
FORMULATION
The minimum total potential
energy formulation is a common
approach in generating FE
models in solid mechanics.
External loads applied to a body
will cause the body to deform.
During the deformation, the work
done by the external forces is
stored in the material in the form
of strain energy. Let us consider
the strain energy in a solid
member when it is subjected an
axial force F, as shown in the
Figure

Continued

When the member is stretched by a differential amount dy',

the stored energy in the material is
y

1
2
Fdy k ydy k ( y)
2
0
0
We can this Equation -for a piece of material from the
member in the form of differential volume-in terms of the
normal stress and strain as:

1
1
1
d k ( y) d y ( y dxdz)( y dy) ( y y )dV
2
2
2
Therefore, for a member or an element under axial loading,
the strain energy is obtained by adding up the stored
energy in all pieces (differential volumes) making up the
member:
2
2
2

(e)

E
dV
dV=
dV=
dV
2
2
2EA
2E
2
V
V
V

Continued

where V is the volume of the member. The total potential

energy for a body consisting of n elements and m
nodes is the difference between the total strain energy
and the work done by the external forces:
n

Fu
i i
e 1

(e)

i 1

The minimum total potential energy principle simply

states that for a stable system, the displacement at the
equilibrium position occurs such that the value of the
system's total potential energy is a minimum.

ui ui

(e)

ui
e 1
n

Fu
i 1

i i

0 for all i 1, 2,..., n

The following examples offer insight into MINIMUM

TOTAL POTENTIAL ENERGY FORMULATION

EXAMPLE 5
Consider the following situations: We have applied a force
F = 12 N to a linear spring as shown in the Figure. Assume
that the slender rod weighs 8 N and is supported by a
spring with a stiffness k = 20 N/cm. We are interested in
determining the deflection of the spring.

Approach: First, we solve this problem by

applying the static equilibrium conditions and
then apply the minimum total potential energy
concept.
Continued

Static
equilibrium
requires that sum of the
moments of the forces
acting on the rod about
point A be zero.

5W 10P 8Fs Fs 20 N
We note that elastic energy stored in the system is
predominantly due to elastic energy of the spring which is
given by
1 2
2

kx 10 x
2

The work done by the external forces is calculated by

multiplying the external forces (at C and G) by their
displacements. To do this we first relate the
displacements of points G and C to the displacement of
the spring (point B) according to

Continued

xc x xG

xG 5 x / 8
10 8 5

xc 10 x / 8

Thus, the work done by the external forces is given by

10 x 5 x
Fu
8

i i 12

8 8
i 1
m

Thus, the total potential energy of the system is

1 2 10 x 5 x
2
kx 12

10
x
20 x

2
8 8
and

Fs 20 N

2 10 x 5 x
10 x 12
8 10 x 20 0

x x
8 8
x 1

Continued

It is clear from examining the Figure that the minimum total

potential energy occurs at x = 1 cm.

10 x 20 x
2

Verification EXAMPLE
Now, let us turn our attention back
to Example 2. The strain energy
for an arbitrary element (e) can be
determined as

(e)

E 2
E 2

dV=
Aavr L
2
2
V

E Aavr L ui 1 ui 2 E Aavr 2

(
)
(ui 1 ui2 2ui 1ui )
2
L
2L

(e)

Minimizing the strain energy with respect to ui and ui+1

leads to
E Aavr
( e ) E Aavr 2
2

(ui 1 ui 2ui 1ui )

(ui ui 1 )
ui
2L ui
L

E Aavr
( e ) E Aavr
2
2

(ui 1 ui 2ui 1ui )

(ui 1 ui )
ui 1
2L ui 1
L

Continued

Or in matrix form as

( e )
u
i E Aavr
( e )
L

ui 1

1 1 ui
1 1 ui
1 1 u keqv 1 1 u

i 1

i 1

Minimizing the work done by the external forces at nodes

i and i + 1 of an arbitrary element (e), we get

Fi ui Fi
ui

Fi 1 ui 1 Fi 1
ui 1
Thus, the minimum total potential energy formulation
leads to a global stiffness matrix that is identical to the
one obtained from direct formulation.

Continued

In this section, le us proceed to find the exact solution of

the same problem and compare the finite element
formulation displacement results for this problem to the
exact displacement solutions.
The statics equilibrium requires the sum of the forces in
the y-direction to be zero. That is

P avg A( y) 0

P E A( y) 0

du
P E A( y)
0
dy
However , A(y) can be derived as

w2 w1

A( y ) w1
L

y t

Continued

w2 w1

P E w1
L

du
y t
0
dy

Which is a separable DE to be solved by direct integration

L

P
u
w2 w1
0 Ew
1
L

y t

dy

w2 w1
PL
u( y)
ln w1
L ln w1

E t w2 w1
L

It is clear from examination the following Table that all of the

results are in agreement with each other.
Continued

HW: Verify the results for stress

3. WEIGHTED RESIDUAL FORMULATIONS

The weighted residual methods are based on assuming an
approximate solution for the governing differential
equation. The assumed solution must satisfy the initial and
boundary conditions of the given problem. Because the
assumed solution is not exact, substitution of the solution
into the differential equation will lead to some residuals or
errors. Simply stated, each residual method requires the
error to vanish over some selected intervals or at some
points. To demonstrate this concept, let's turn our attention
to the previous Example

du
E A( y)
P 0 With boundary condition u(0)=0
dy
With

w2 w1

A( y ) w1
L

y t

In this section, let us proceed to find the exact solution of

the same problem and compare the finite element
formulation displacement results for this problem to the
exact displacement solutions.
The statics equilibrium requires the sum of the forces in
the y-direction to be zero. That is

P avg A( y) 0

P E A( y) 0

du
P E A( y)
0
dy
However , A(y) can be derived as

w2 w1

A( y ) w1
L

y t

Continued

w2 w1

P E w1
L

du
y t
0
dy

Which is a separable DE to be solved by direct integration

L

P
u
w2 w1
0 Ew
1
L

y t

dy

w2 w1
PL
u( y)
ln w1
L ln w1

E t w2 w1
L

Continued

Next, we need to assume an approximate solution. Again,

keep in mind that the assumed solution must satisfy the
boundary condition. For instant we choose

u( y) C1 y C2 y 2 C3 y 3
Which is certainly satisfies the fixed boundary condition
represented by u(0)=0. C1, C2, and C3 are unknown
coefficients. Now, substituting of the assumed solution into
the governing differential equation yields to the error
function :

w2 w1
P

2
y t C1 2C2 y 3C3 y
w1
L
E E

1000
2

0.25 0.0125 y C1 2C2 y 3C3 y

E
E

A. Collocation Method
In the collocation method the error, or residual, function is
forced to be zero at as many points as there are unknown
coefficients. Because the assumed solution in this
example has three unknown coefficients, we will force the
error function to be equal zero at three points. We choose
the error function to vanish at y = L/3=10/3, y = 2L/3=20/3 ,
and y = L=10:
2

10
10

10 1000
0 0.25 0.0125 C1 2C2 3C3
0

E y 10
3
3
E

3
3

E y 20
3

20
20

20 1000
0 0.25 0.0125 C1 2C2
3C3
0

3
3
E

0
E y 10

0.25 0.125 C1 20C2 300C3

1000
0
E

Solving the above equations yields

C1 423.0776 10

C2 21.65 1015
C3 1.153848 10

Substitution of the c-coefficients into the assumed

solution yields the approximate displacement profile:

u( y) 423.0776 106 y 21.65 1015 y2 1.153848 106 y3

B. Subdomain Method
In the subdomain method, the integral of the error
function over some selected subintervals is forced to be
zero. The number of subintervals chosen must equal the
number of unknown coefficients. Thus, for our assumed
solution, we will have three integrals:
L /3

L /3

1000

2
dy 0 0.25 0.0125 y C1 2C2 y 3C3 y
dy 0

E
E
0

2 L /3

L /3
L

2 L /3

1000

2
dy 0 0.25 0.0125 y C1 2C2 y 3C3 y
dy 0

E
E
L /3
L

1000

2
dy 0 0.25 0.0125 y C1 2C2 y 3C3 y
dy 0

E
E
2 L /3
2 L /3
Which leads to three equations to be solved for C1, C2,
and C3 as:

Solving the above equations yields

C1 391.35088 106
C2 6.075 106
C3 809.61092 109
Substitution of the c-coefficients into the assumed
solution yields the approximate displacement profile:

u( y) 391.35088 106 y 6.075 106 y2 809.61092 109 y3

C. Galerkin Method
The Galerkin method requires the error to be orthogonal to
some weighting functions i , according to the integral
b

dy 0 for i=1,2,3,..., N

The weighting functions are chosen to be members of the

approximate solution. Because there are three unknowns
in the assumed approximate solution we need to generate
three equations. Recall that the assumed solution is

u( y) C1 y C2 y 2 C3 y 3
Thus, the weighting functions are selected to be

1 y, 2 y 2 , 3 y 3

This selection leads to the following equations:

L

L /3

2 L /3

1000

dy

0.25

0.0125
y
C

2
C
y

3
C
y

1
E ydy 0
2
3
0 E 1
0

1000 2

2
0 E 2dy 0 L/3 0.25 0.0125 y C1 2C2 y 3C3 y E y dy 0

1000 3

2
0 E 3dy 0 2 L/3 0.25 0.0125 y C1 2C2 y 3C3 y E y dy 0

Solving the above equations yields

C1 400.642 106 , C2 4.006 106 , C3 0.9352 106
Substitution of the c-coefficients into the assumed
solution yields the approximate displacement profile:

u( y) 400.642 106 y 4.006 106 y2 0.9352 106 y3

D. Least-Squares Method
The least-squares method requires the error to be
minimized with respect to the unknown coefficients in the
assumed solution, according to the relationship

b 2
minimize dy
a

This leads to
b

a Ci dy 0, for i 1, 2,3,..., N
Because there are three unknowns (N=3) in the
approximate solution, generates three equations with
three unknowns ( C1, C2 and C3) that can be solved as
follows

Solving the above generated three equations

yields
C1 389.773 106 , C2 6.442 106 , C3 0.789 106
Substitution of the c-coefficients into the assumed
solution yields the approximate displacement profile:

u( y) 389.773106 y 6.442 106 y2 0.789 106 y3

HW: Verify the results of the above 4

WEIGHTED RESIDUAL Methods

In the previous Section weighted residual methods are introduced. It is

clear from the above Table that the results are in agreement with each
other.
The Galerkin method (GM)
1. Is one of the most commonly used procedures in FE formulations.
2. will be offered later in Chapters 6 and 9.
3. Is used to formulate one- and two-dimensional problems.
4. will assume linear or nonlinear solutions that are valid only over
each element and then assemble, the elemental solutions.

Recently, ANSYS have become a

common tool in hands of design
engineers to verify Eng. Problems

Chapter Summary