CHAPTER

Discretization of the Domain

CHAPTER OUTLINE
2.1 Introduction 53
2.2 Basic Element Shapes 53
2.3 Discretization Process 56
2.3.1
2.3.2
2.3.3
2.3.4

Type of Elements 56
Size of Elements 59
Location of Nodes 60
Number of Elements 60

2.3.5 Simplifications Afforded by the

Physical Configuration of
the Body 61
2.3.6 Finite Representation of
Infinite Bodies 61

2.4 Node Numbering Scheme 63

2.5 Automatic Mesh Generation 65

2.1 INTRODUCTION
In most engineering problems, we need to find the values of a field variable such as
displacement, stress, temperature, pressure, and velocity as a function of spatial coordinates
(x, y, z). In the case of transient or unsteady-state problems, the field variable has to be
found as a function of not only the spatial coordinates (x, y, z) but also time (t). The
geometry (domain or solution region) of the problem is often irregular. The first step of
the finite element analysis involves the discretization of the irregular domain into smaller
and regular subdomains, known as finite elements. This is equivalent to replacing the
domain having an infinite number of degrees of freedom (dof) by a system having a finite
number of dof.
A variety of methods can be used to model a domain with finite elements. Different
methods of dividing the domain into finite elements involve varying amounts of
computational time and often lead to different approximations to the solution of the
physical problem. The process of discretization is essentially an exercise of engineering
judgment. Efficient methods of finite element idealization require some experience and
knowledge of simple guidelines. For large problems involving complex geometries, finite
element idealization based on manual procedures requires considerable effort and time
on the part of the analyst. Some automatic mesh generation programs have been
developed for the efficient idealization of complex domains requiring minimal interface
with the analyst.

2.2 BASIC ELEMENT SHAPES

The shapes, sizes, number, and configurations of the elements have to be chosen carefully
such that the original body or domain is simulated as closely as possible without
increasing the computational effort needed for the solution. Mostly the choice of the

The Finite Element Method in Engineering. DOI: 10.1016/B978-1-85617-661-3.00002-7

2011 Elsevier Inc. All rights reserved.

53

PART 2
Basic Procedure

type of element is dictated by the geometry of the body and the number of independent
coordinates necessary to describe the system. If the geometry, material properties, and the
field variable of the problem can be described in terms of a single spatial coordinate, we
can use the one-dimensional or line elements shown in Figure 2.1(a). The temperature
distribution in a rod (or fin), the pressure distribution in a pipe flow, and the
deformation of a bar under axial load, for example, can be determined using these
elements. Although these elements have a cross-sectional area, they are generally shown
schematically as a line element (Figure 2.1(b)). In some cases, the cross-sectional area of
the element may be nonuniform.

Node
1

Node
2

(a)
Node
1

Node
2

(b)

54

Node 1

Node 2
(c)

FIGURE 2.1
One-Dimensional Elements.

For a simple analysis, one-dimensional elements are assumed to have

two nodes, one at each end, with the corresponding value of the field
variable chosen as the unknown (degree of freedom). However, for
the analysis of beams, the values of the field variable (transverse
displacement) and its derivative (slope) are chosen as the unknowns
(dof) at each node as shown in Figure 2.1(c).
When the configuration and other details of the problem can be
described in terms of two independent spatial coordinates, we can use
the two-dimensional elements shown in Figure 2.2. The basic element
useful for two-dimensional analysis is the triangular element. Although
a quadrilateral element (or its special forms, the rectangle and
parallelogram) can be obtained by assembling two or four triangular
elements, as shown in Figure 2.3, in some cases the use of quadrilateral
(or rectangle or parallelogram) elements proves to be advantageous.
For the bending analysis of plates, multiple dof (transverse
displacement and its derivatives) are used at each node.
If the geometry, material properties, and other parameters of the body
can be described by three independent spatial coordinates, we can
idealize the body by using the three-dimensional elements shown in
Figure 2.4. The basic three-dimensional element, analogous to the

Node 4

3
1

Node 1

3
1

1
Triangle

Node 3

Rectangle
Node 2

2
4
3
2

i Element number

1
3

4
Quadrilateral

FIGURE 2.2
Two-Dimensional Elements.

4
Parallelogram

FIGURE 2.3
A Quadrilateral Element as an Assemblage of Two or Four
Triangular Elements.

CHAPTER 2
Discretization of the Domain

triangular element in the case of two-dimensional

problems, is the tetrahedron element. In some cases
the hexahedron element, which can be obtained by
assembling five tetrahedrons as indicated in Figure 2.5,
can be used advantageously. Some problems, which are
actually three-dimensional, can be described by only one
or two independent coordinates. Such problems can be
idealized by using an axisymmetric or ring type of
elements shown in Figure 2.6. The problems that possess
axial symmetry, such as pistons, storage tanks, valves,
rocket nozzles, and reentry vehicle heat shields, fall into
this category.

1
2

1
4
3

6
5

2
Tetrahedron

8
Rectangular prism

2
3

For the discretization of problems involving curved

geometries, finite elements with curved sides are useful.
Typical elements having curved boundaries are shown in
Figure 2.7. The ability to model curved boundaries has
been made possible by the addition of mid-side nodes.
Finite elements with straight sides are known as linear
elements, whereas those with curved sides are called
higher-order elements.

6
7

8
Hexahedron

FIGURE 2.4
Three-Dimensional Finite Elements.
8

55
4

1
1
6

3
5

8
8

6
8

3
3

6
1

3
2

FIGURE 2.5
A Hexahedron Element as an Assemblage of Five Tetrahedron Elements.

PART 2
Basic Procedure

1
z
(b) Two-dimensional axisymmetric
(toroidal) element

(a) One-dimensional axisymmetric

(shell) element

FIGURE 2.6
Axisymmetric Elements.
2
4

5
2

2
Curved line element

9
1

9
10

3
6
Tetrahedron with
curved sides

4
1
Annular element

6
Plane triangle with
curved sides

10
14

11

3
12 4
18 15
13 17 6
16
7
5
19
20
8
Hexahedron with curved
sides

2
6

4
1 5 3
Axisymmetric ring element with
curved triangular section
2

56

Rectangular shell
element

2
1
Truncated conical
shell element

Doubly curved
triangle

FIGURE 2.7
Finite Elements with Curved Boundaries.

2.3 DISCRETIZATION PROCESS

Various considerations to be taken in the discretization process [2.1] are discussed in the
following sections.

2.3.1 Type of Elements

Often, the type of elements to be used will be evident from the physical problem. For
example, if the problem involves the analysis of a truss structure under a given set of load
conditions (Figure 2.8(a)), the type of elements to be used for idealization is obviously the
bar or line elements as shown in Figure 2.8(b). Similarly, in the case of stress analysis of
the short beam shown in Figure 2.9(a), the finite element idealization can be done using
three-dimensional solid elements as shown in Figure 2.9(b). However, the type of elements
to be used for idealization may not be apparent, and in such cases one has to choose the
type of elements judicially. As an example, consider the problem of analysis of the thinwalled shell shown in Figure 2.10(a). In this case, the shell can be idealized by several types
of elements as shown in Figure 2.10(b). Here, the number of dof needed, the expected
accuracy, the ease with which the necessary equations can be derived, and the degree to
which the physical structure can be modeled without approximation will dictate the choice

CHAPTER 2
Discretization of the Domain

P1

P1

P2

P2

P3

P3

FIGURE 2.8
(a) Original structure

(b) Idealization using bar elements

A Truss Structure.

A
B

C
i

P2
E

D
H
F

P1

A
C

D
G

E
(a) Original beam

F
Element i

(b) Idealization using three-dimensional elements

Using conical ring elements

Using axisymmetric ring

elements

57

FIGURE 2.9
A Short Beam.

(a) Original shell

i
Using flat triangular plate
elements

i
Using curved triangular plate
elements

(b) Idealization using different types of elements

FIGURE 2.10
A Thin-Walled Shell under
Pressure.

PART 2
Basic Procedure

Rib elements

Spar elements

FIGURE 2.11
Idealization of an Aircraft
Wing Using Different Types
of Elements.

58

Cover plate elements

Flange areas

of the element type to be used for idealization. In certain problems, the given body cannot
be represented as an assemblage of only one type of elements. In such cases, we may have
to use two or more types of elements for idealization. An example of this would be the
analysis of an aircraft wing. Since the wing consists of top and bottom covers, stiffening
webs, and flanges, three types of elementsnamely, triangular plate elements (for covers),
rectangular shear panels (for webs), and frame elements (for flanges)have been used in
the idealization shown in Figure 2.11.

EXAMPLE 2.1
A helical spring is subjected to a compressive load as shown in Figure 2.12(a). Suggest different
methods of modeling the spring using one-dimensional elements.

Solution
Approach: Use various one-dimensional or line elements.
The helical spring (in the form of curved wire) can be divided into several line or one-dimensional
segments. These segments can be straight or curved. Each of the straight line segments (or elements) can
be assumed to be a spatial truss element with each of its endpoints (or nodes) having three displacement
dof (parallel to the x, y, and z axes) as shown in Figure 2.12(b). Since this element has only translational
degrees of freedom (with no rotational degrees of freedom), it will not be able to carry any moment. As
such, the element may not be able to represent the behavior of the helical spring accurately.
Alternately, each of the straight line segments (or elements) can he assumed to be a spatial frame element
with each of its endpoints (or nodes) having three displacement dof (parallel to the x, y, and z axes) and
three rotational dof (about the x, y, and z axes) as shown in Figure 2.12(c). In the case of the curved line
segments (elements), each element can be treated as a curved frame element with three displacement
dof (parallel to the x, y, and z axes) and three rotational dof (about the x, y, and z axes) at each end as
shown in Figure 2.12(d). Because of the inclusion of rotational degrees of freedom, the models shown in
Figures 2.12(c) and (d) will be able to simulate the behavior of the helical spring more accurately.

CHAPTER 2
Discretization of the Domain

i
i

j
F
Helical spring in compression
(a)

Element e

Element e

Element e

Spatial truss element

(b)

Spatial frame element

(c)

Curved frame element

(d)

FIGURE 2.12
Modeling of a Helical Spring.

2.3.2 Size of Elements

The size of elements influences the convergence of the solution directly, and hence it has to be
chosen with care. If the size of the elements is small, the final solution is expected to be more
accurate. However, we have to remember that the use of smaller-sized elements will also
mean more computation time. Sometimes, we may have to use elements of different sizes in
the same body. For example, in the case of stress analysis of the box beam shown in Figure
2.13(a), the size of all the elements can be approximately the same, as shown in Figure 2.13(b).
However, in the case of stress analysis of a plate with a hole shown in Figure 2.14(a),

P2
P1

(a) Original structure

FIGURE 2.13
A Box Beam.

(b) Finite element idealization

59

PART 2
Basic Procedure

(a) Original structure

(b) Idealization using elements of

different sizes

FIGURE 2.14
A Plate with a Hole.

elements of different sizes have to be used, as shown in Figure 2.14(b). The size of elements
has to be very small near the hole (where stress concentration is expected) compared to
distant places. In general, whenever steep gradients of the field variable are expected, we have
to use a finer mesh in those regions. Another characteristic related to the size of elements that
affects the finite element solution is the aspect ratio of the elements. The aspect ratio describes
the shape of the element in the assemblage of elements. For two-dimensional elements, the
aspect ratio is taken as the ratio of the largest dimension of the element to the smallest
dimension. Elements with an aspect ratio of nearly unity generally yield best results [2.2].

2.3.3 Location of Nodes

If the body has no abrupt changes in geometry, material properties, and external conditions
(e.g., load and temperature), the body can be divided into equal subdivisions and hence the
spacing of the nodes can be uniform. On the other hand, if there are any discontinuities in
the problem, nodes have to be introduced at these discontinuities, as shown in Figure 2.15.
60

2.3.4 Number of Elements

The number of elements to be chosen for idealization is related to the accuracy desired,
size of elements, and the number of dof involved. Although an increase in the number of
elements generally means more accurate results, for any given problem, there will be a
certain number of elements beyond which the accuracy cannot be significantly improved.
This behavior is shown graphically in Figure 2.16. Moreover, since the use of a large
number of elements involves a large number of dof, we may not be able to store the
resulting matrices in the available computer memory.
P
Node
Concentrated load
on a beam

Steel
Aluminum

Nodal line

A bimetallic beam
(c) Discontinuity in material properties
Node
Abrupt change in the
distributed load

Nodes

(a) Discontinuity in loading

Node
Abrupt change in crosssection of beam
(b) Discontinuity in geometry

FIGURE 2.15
Location of Nodes at Discontinuities.

Cracked plate under loading

(d) Discontinuity in material

CHAPTER 2
Discretization of the Domain

Exact
solution
Solution given by
finite element method

N0
No significant improvement beyond N0

Number of
elements

FIGURE 2.16
Effect of Varying the Number of Elements.

2.3.5 Simplifications Afforded by the

Physical Configuration of the Body

If the configuration of the body as well as the external

conditions are symmetric, we may consider only half of
the body for finite element idealization. The symmetry
conditions, however, have to be incorporated in the
solution procedure. This is illustrated in Figure 2.17,
where only half of the plate with a hole, having
symmetry in both geometry and loading, is considered
for analysis.1 Since there cannot be a horizontal
displacement along the line of symmetry AA, the
condition that u = 0 has to be incorporated while
finding the solution.

Symmetry condition
u = 0 along this nodal
line

2.3.6 Finite Representation of

Infinite Bodies
In most of the problems, like in the analysis of beams,
y (v)
plates, and shells, the boundaries of the body or
A
A
continuum are clearly defined. Hence, the entire body
can be considered for element idealization. However,
x (u)
in some cases, as in the analysis of dams, foundations,
(a) Plate with hole
(b) Only half of plate
and semi-infinite bodies, the boundaries are not clearly
can be considered for
analysis
defined. In the case of dams (Figure 2.18), since the
geometry is uniform and the loading does not change
FIGURE 2.17
in the length direction, a unit slice of the dam can be considered for idealization and
A Plate with a Hole with
Symmetric Geometry
analyzed as a plane strain problem. However, in the case of the foundation problem
and Loading.
shown in Figure 2.19(a), we cannot idealize the complete semi-infinite soil by finite
elements. Fortunately, it is not really necessary to idealize the infinite body. Since the effect
of loading decreases gradually with increasing distance from the point of loading, we can
consider only that much of the continuum in which the loading is expected to have a
significant effect as shown in Figure 2.19(b). Once the significant extent of the infinite body
is identified as shown in Figure 2.19(b), the boundary conditions for this finite body have to
be incorporated in the solution. For example, if the horizontal movement only has to be
restrained for sides AB and CD (i.e., u = 0), these sides are supposed to be on rollers as
1

In this example, even one-fourth of the plate can be considered for analysis due to symmetry about both
horizontal and vertical center lines.

61

PART 2
Basic Procedure

Unit slice considered for

plane strain analysis

FIGURE 2.18
A Dam with Uniform Geometry and Loading.

62

Footing
Semi-infinite soil

FIGURE 2.19
A Foundation under
Concentrated Load.

B
(a) Original foundation

C
(b) Idealization of semi-infinite medium

shown in Figure 2.19(b). In this case, the bottom boundary can be either completely fixed
(u = v = 0) or constrained only against vertical movement (v = 0). The fixed conditions
(u = v = 0 along BC) are often used if the lower boundary is taken at the known location of
a bedrock surface.
In Figure 2.19, the semi-infinite soil has been simulated by considering only a finite
portion of the soil. In some applications, the determination of the size of the finite domain
may pose a problem. In such cases, one can use infinite elements for modeling [2.32.5].
As an example, Figure 2.20 shows a four-node element that is infinitely long in the
x direction. The coordinates of the nodes of this infinite element can be transformed
to the natural coordinate system (s, t) as follows:

m
1 y3 yx1 + y y1 x4
; m1
s = 12
x
y3 y1


y3 y
t = 12
y3 y1
See Section 4.3.3 for the definition of the natural coordinate system.

CHAPTER 2
Discretization of the Domain

(x4, y4 = y3)
4

(x3 = , y3)
3

y
2
(x2 = , y2 = y1)

1
(x1, y1)

FIGURE 2.20
A Four-Node Infinite
Element.

2.4 NODE NUMBERING SCHEME

As seen in Chapter 1, the finite element analysis of practical
problems often leads to matrix equations in which the matrices
involved will be banded. The advances in the finite element
analysis of large practical systems have been made possible
largely due to the banded nature of the matrices. Furthermore,
since most of the matrices involved (e.g., stiffness matrices) are
symmetric, the demands on the computer storage can be
substantially reduced by storing only the elements involved in
half bandwidth instead of storing the entire matrix.
The bandwidth of the overall or global characteristic matrix
depends on the node numbering scheme and the number of dof
considered per node [2.6]. If we can minimize the bandwidth, the
storage requirements as well as solution time can also be
minimized. Since the number of dof per node is generally fixed for
any given type of problem, the bandwidth can be minimized by
using a proper node numbering scheme. As an example, consider
a three-bay frame with rigid joints, 20 stories high, shown in
Figure 2.21. Assuming that there are 3 dof per node, there are 252
unknowns in the final equations (including the dof corresponding
to the fixed nodes), and if the entire stiffness matrix is stored in the
computer, it will require 2522 = 63,504 locations. The bandwidth
(strictly speaking, half-bandwidth) of the overall stiffness matrix
can be shown to be 15, and thus the storage required for the upper
half-band is only 15 252 = 3780 locations.

(1,2,3)
1
A

5
(13,14,15)

(4,5,6)
2
B

(7,8,9)
3

D 6

(10,11,12)
4

(16,17,18) (19,20,21) (22,23,24)

63
20
Stories

77

81

78

79

80

82

83

84

(241,242,243)

Before we attempt to minimize the bandwidth, we discuss the method of calculating the
bandwidth. For this, we consider again the rigid jointed frame shown in Figure 2.21. By
applying constraints to all the nodal dof except number 1 at node 1 (joint A), it is clear
that an imposed unit displacement in the direction of 1 will require constraining forces at
the nodes directly connected to node Athat is, B and C. These constraining forces are
nothing but the cross-stiffnesses appearing in the stiffness matrix, and these forces are
confined to the nodes B and C. Thus, the nonzero terms in the first row of the global
stiffness matrix (Figure 2.22) will be confined to the first 15 positions. This defines the
bandwidth (B) as
Bandwidth (B) = (maximum difference between the numbered dof
at the ends of any member + 1)

(250,251,252)

FIGURE 2.21
A Three-bay Frame.

PART 2
Basic Procedure

This definition can be generalized so as to be applicable for any type of finite element as
Bandwidth B = D + 1 f

(2.1)

where D is the maximum largest difference in the node numbers occurring for all elements
of the assemblage, and f is the number of dof at each node.
The previous equation indicates that D has to be
minimized in order to minimize the bandwidth. Thus, a
shorter bandwidth can be obtained simply by numbering
the nodes across the shortest dimension of the body. This
is clear from Figure 2.23 also, where the numbering of
nodes along the shorter dimension produces a
bandwidth of B = 15 (D = 4), whereas the numbering
along the longer dimension produces a bandwidth of
B = 66 (D = 21).

B = Bandwidth = 15

As observed previously, the bandwidth of the overall

system matrix depends on the manner in which the nodes
are numbered. For simple systems or regions, it is easy to
label the nodes so as to minimize the bandwidth. But for
large systems, the procedure becomes nearly impossible.
Hence, automatic mesh generation algorithms, capable of
discretizing any geometry into an efficient finite element
mesh without user intervention, have been developed [2.7,
2.8]. Most commercial finite element software has built-in
automatic mesh generation codes. An automatic mesh
generation program generates the locations of the node
points and elements, labels the nodes and elements, and
provides the elementnode connectivity relationships.

240
Equations

64

FIGURE 2.22
Banded Nature of the
Stiffness Matrix for the
Frame of Figure 2.21.

22

43

64

23

44

65

24

45

66

77

78

79

80

20

41

62

83

81

82

83

84

21

42

63

84

(a) Node numbering along the

shorter dimension

FIGURE 2.23
Different Node Numbering Schemes.

(b) Node numbering scheme

along the longer dimension

CHAPTER 2
Discretization of the Domain

EXAMPLE 2.2
A drilling machine is modeled using one-dimensional beam elements as shown in Figure 2.24(a). If
two dof are associated with each node, label the node numbers for minimizing the bandwidth of the
stiffness matrix of the system.

Solution
Approach: Number the nodes along the shorter side of the machine first.
Because the column (vertical member) of the machine has 5 nodes and the arm (horizontal member)
has only 4 nodes, we number the nodes along the shorter side as shown in Figure 2.24(b). Noting that
the maximum difference between the numbers of the end nodes among all the elements is 2, the
bandwidth of the resulting stiffness matrix of the system is given by
B = D + 1f = 2 + 12 = 6
Note that the nodes can also be numbered as shown in Figure 2.24(c), which also yields the same
bandwidth of B = 6.

2
3

65
(a)

(b)

(c)

FIGURE 2.24
A Drilling Machine.

2.5 AUTOMATIC MESH GENERATION2

Mesh generation is the process of dividing a physical domain into smaller subdomains
(called elements) to facilitate an approximate solution of the governing ordinary or partial
differential equation. For this, one-dimensional domains (straight or curved lines) are
subdivided into smaller line segments, two-dimensional domains (planes or surfaces) are
subdivided into triangle or quadrilateral shapes, and three-dimensional domains (volumes)
are subdivided into tetrahedron and hexahedron shapes. If the physical domain is simple
and the number of elements used is small, mesh generation can be done manually.
However, most practical problems, such as those encountered in aerospace, automobile,
and construction industries have complex geometries that require the use of thousands and
sometimes millions of elements. In such cases, the manual process of mesh generation is
impossible and we have to use automatic mesh generation schemes based on the use of a
CAD or solid modeling package.
Automatic mesh generation involves the subdivision of a given domain, which may be in
the form of a curve, surface, or solid (described by a CAD or solid modeling package)
into a set of nodes (or vertices) and elements (subdomains) to represent the domain as
2

This section may be omitted without loss of continuity in the text material.

PART 2
Basic Procedure

closely as possible subject to the specified element shape and size restrictions. Many
automatic mesh generation schemes use a bottom-up approach in that nodes (or
vertices or corners of the domain) are meshed first, followed by curves (boundaries), then
surfaces, and finally solids. Thus, for a given geometric domain of the problem, nodes are
first placed at the corner points of the domain, and then nodes are distributed along the
geometric curves that define the boundaries. Next, the boundary nodes are used to
develop nodes in the surface(s), and finally the nodes on the various surfaces are used to
develop nodes within the given volume (or domain). The nodes or mesh points are used
to define line elements if the domain is one-dimensional, triangular, or quadrilateral
elements if the domain is two-dimensional, and tetrahedral or hexahedral elements if the
domain is three-dimensional.
The automatic mesh generation schemes are usually tied to solid modeling and computeraided design schemes. When the user supplies information on the surfaces and volumes
of the material domains that make up the object or system, an automatic mesh generator
generates the nodes and elements in the object. The user can also specify minimum
permissible element sizes for different regions of the object. Many mesh generation schemes
first create all the nodes and then produce a mesh of triangles by connecting the nodes to
form triangles (in a plane region). In a particular scheme, known as Delaunay triangulation,
the triangular elements are generated by maximizing the sum of the smallest angles of the
triangles; thus the procedure avoids generation of thin elements.

66

The most common methods used in the development of automatic mesh generators
are the tesselation and octree methods [2.9, 2.10]. In the tesselation method, the user
gives a collection of node points and also an arbitrary starting node. The method then
creates the first simplex element using the neighboring nodes. Then a subsequent or
neighboring element is generated by selecting the node point that gives the least distorted
element shape. The procedure is continued until all the elements are generated. The stepby-step procedure involved in this method is illustrated in Figure 2.25 for a twodimensional example. Alternately, the user can define the boundary of the object by a
series of nodes. Then the tesselation method connects selected boundary nodes to
generate simplex elements. The stepwise procedure used3 in this approach is shown in
Figure 2.26.
The octree methods belong to a class of mesh generation schemes known as tree structure
methods, which are extensively used in solid modeling and computer graphics display
methods. In the octree method, the object is first considered enclosed in a threedimensional cube. If the object does not completely (uniformly) cover the cube, the cube is
subdivided into eight equal parts. In the two-dimensional analog of the octree method,
known as the quadtree method, the object is first considered enclosed in a square region. If
the object does not completely cover the square, the square is subdivided into four equal
quadrants. If any one of the resulting quadrants is full (completely occupied by the object)
or empty (not occupied by the object), then it is not subdivided further. On the other
hand, if any one of the resulting quadrants is partially full (partially occupied by the
object), it is subdivided into four quadrants. This procedure of subdividing partially full
quadrants is continued until all the resulting regions are either full or empty, or until some
predetermined level of resolution is achieved. At the final stage, the partially full quadrants
are assumed to be either full or empty arbitrarily based on a prespecified criterion.
The approaches indicated in this section can be extended naturally to three and higher
dimensional spaces.

3
A simplex in an n-dimensional space is defined as a geometric figure having n + 1 nodes or corners. Thus, the
simplex will be a triangle in a two-dimensional space and a tetrahedron in three-dimensional space.

CHAPTER 2
Discretization of the Domain

9
8
8
7
5

7
6

1
(a) Nodes in the object or region

3
2

1
(b) Generation of simplex elements

(c) Complete set of nodes and elements

FIGURE 2.25
Mesh Generation Using Tesselation Method.

8
2

3
6

7
5
8

5
1

3
4

4
9

(a) Nodes on the boundary of the object or region

(b) Geometry of the object or region

(c) Complete set of nodes and elements

FIGURE 2.26
Tesselation Method with Nodes Defined on the Boundary.

EXAMPLE 2.3
Generate the finite element mesh for the two-dimensional object (region) shown by the crossed lines in
Figure 2.27(a) using the quadtree method.

Solution
Approach: Use the quadtree method.
First, the object is enclosed in a square region as shown by the dotted lines in Figure 2.27(a). Since the
object does not occupy the complete square, the square is divided into four parts as shown in Figure
2.27(b). Since none of these parts are fully occupied by the object, each part is subdivided into four
parts as shown in Figure 2.27(c). It can be seen that parts 1, 3, and 4 of A, part 3 of B, parts 2 to 4 of C,
and parts 1 to 3 of D are completely occupied by the object, whereas parts 1, 2, and 4 of B and part 1 of C
are empty (not occupied by the object). In addition, part 2 of A and part 4 of D are partially occupied by
the object; hence, they are further subdivided into four parts each as shown in Figure 2.27(d). It can be
noted that parts and of part 2 (of A) and parts and of part 4 (of D) are completely occupied while
the remaining parts, namely and of part 2 (of A) and and of part 4 (of D), are empty. Since all the
parts at this stage are either completely occupied or completely empty, no further subdivision is necessary.
The corresponding quadtree representation is shown in Figure 2.27(e). Note that the shape of the finite
elements is assumed to be square in this example.
(Continued )

67

PART 2
Basic Procedure

EXAMPLE 2.3

(Continued )

(a) Given object enclosed by a square

68

(c) Division of each partially occupied part

into four parts

1 2

(b) Division into four parts

3 4

1 2

Partially occupied

(d) Division of each partially occupied part

into four parts

3 4

Fully occupied

Not occupied

(e) Quadtree representation

FIGURE 2.27
Mesh Generation Using Quadtree Method.

REFERENCES
2.1 O.C. Zienkiewicz: The finite element method: From intuition to generality, Applied Mechanics Reviews, 23,
249256, 1970.
2.2 R.W. Clough: Comparison of three dimensional finite elements, Proceedings of the Symposium on Application of
Finite Element Methods in Civil Engineering. Vanderbilt University, Nashville, 126, November 1969.

CHAPTER 2
Discretization of the Domain

2.3 P. Bettess: Infinite elements, International Journal for Numerical Methods in Engineering, 11, 5364, 1977.
2.4 F. Medina and R.L. Taylor: Finite element techniques for problems of unbounded domains, International
Journal for Numerical Methods in Engineering, 19, 12091226, 1983.
2.5 S. Pissanetzky: An infinite element and a formula for numerical quadrature over an infinite interval,
International Journal for Numerical Methods in Engineering, 19, 913927, 1983.
2.6 R.J. Collins: Bandwidth reduction by automatic renumbering, International Journal for Numerical Methods in
Engineering, 6, 345356, 1973.
2.7 J.E. Akin: Finite Elements for Analysis and Design, Academic Press, London, 1994.
2.8 K. Baldwin, Ed.: Modern Methods for Automatic Finite Element Mesh Generation, American Society of Civil
Engineers, New York, 1986.
2.9 P.L. George: Automatic Generation of Meshes, Wiley, New York, 1991.
2.10 C.G. Armstrong: Special issue: Automatic mesh generation, Advances in Engineering Software, 13, 217337, 1991.
2.11 K. Ho-Le, Finite element mesh generation methods: A review and classification, Computer Aided Design,
Vol. 20, No. 1, JanuaryFebruary 1988, pp. 2738.
2.12 J.A. Talbert and A.R. Parkinson: Development of an automatic, two-dimensional finite element mesh generator
using quadrilateral elements and Bezier curve boundary definition, International Journal for Numerical Methods
in Engineering, Vol. 29, No. 7, 1990, pp. 15511567.
2.13 M.S. Shephard and M.K. Georges: Automatic three-dimensional mesh generation by the finite octree
technique, International Journal for Numerical Methods in Engineering, Vol. 32, No. 4, 1991, pp. 709749.

PROBLEMS
2.1 A thick-walled pressure vessel is subjected to an internal pressure as shown in Figure 2.28. Model the
cross section of the pressure vessel by taking advantage of the symmetry of the geometry and load condition.
2.2 A rectangular plate with a V-notch is shown in Figure 2.29. Model the plate using triangular elements by
taking advantage of the symmetry of the system.
2.3 The plate shown in Figure 2.30 is modeled using 13 triangular and 2 quadrilateral elements. Label the nodes
such that the bandwidth of the system matrix is minimal. Compute the resulting bandwidth assuming
1 degree of freedom at each node.

FIGURE 2.28
A Thick-walled Pressure Vessel.

69

PART 2
Basic Procedure

FIGURE 2.29
A Rectangular Plate with a Notch.

3 dof per node

70

FIGURE 2.30

FIGURE 2.31

A Plate Modeled with Triangular and Quadrilateral Elements.

A Planar Frame.

2 dof per node

FIGURE 2.32
A Planar Truss.

2.42.8 Label the elements and nodes for each of the systems shown in Figures 2.31 through 2.35 to produce a
minimum bandwidth. In addition, find the resulting bandwidth in each case.
2.9 Consider the collection of node points shown in Figure 2.36 for a two-dimensional object. Generate the finite
element mesh using the tesselation method.
2.10 Generate the finite element mesh for the two-dimensional object shown in Figure 2.37 using the quadtree
method.
2.11 State the reasons for the desirability of elements with an aspect ratio close to 1 in finite element modeling.
2.12 Give two practical examples each of which one can be modeled using one-, two-, and three-dimensional finite
elements to achieve different levels of accuracy.
2.13 The transmission lines (wires) carrying electricity are supported by electric transmission towers and are
subjected to axial tension and gravity, wind and snow loads. Discuss possible types of finite elements/models
that can be used for the stress analysis of transmission lines.

CHAPTER 2
Discretization of the Domain

2 dof per node

3 dof per node

FIGURE 2.34
A Planar Truss.

FIGURE 2.33
An Aircraft Wing.

71
2 dof per node

FIGURE 2.35
A Planar Truss.

FIGURE 2.36

FIGURE 2.37

Node Points for a Two-Dimensional Object.

A Two-Dimensional Object.

PART 2
Basic Procedure

2.14 Water at pressure pi flows through an underground cast iron pipe. The outer surface of the pipe is subjected to
a uniform axisymmetric pressure along its length by the surrounding soil (Figure 2.38). The pipe has an inner
diameter of 0.7 m, outer diameter of 1.0 m, and a length of 1000 m. Indicate a suitable finite element
idealization for the deformation and stress analysis of the water pipe.
2.15 Label the nodes of the planar truss shown in Figure 2.39 to minimize the bandwidth of the resulting stiffness
matrix. Assume that each node has 2 dof (components of the displacement of the node parallel to the x and
y axes). Also determine the resulting bandwidth of the stiffness matrix.
2.16 Label the nodes of the truncated conical shell shown in Figure 2.40. Assume that each node has three dof
(components of the displacement of the node parallel to the x, y, and z axes). Also determine the resulting
bandwidth of the stiffness matrix.
2.17 A semicircular plate is subjected to loads P1 and P2 as shown in Figure 2.41. Using straight-sided triangular
elements to model the plate is proposed. Show a sequence of three finite element meshes with increasing
number of elements ensuring that each finer mesh includes the previous coarse mesh(es).

p0

pi

FIGURE 2.38
An Underground Pipe.

72
Bar elements

z
y
x

FIGURE 2.39

FIGURE 2.40

A Planar Truss.

A Truncated Conical Shell.

P2
P1

FIGURE 2.41
A Semi-circular Plate.

CHAPTER 2
Discretization of the Domain

P
E, I

L
2

L
2

FIGURE 2.42
A Uniform Fixed-fixed Beam.

FIGURE 2.43
A Mechanical Link.

73

FIGURE 2.44
A Two-Dimensional Plate.

2.18 The data for the uniform beam shown in Figure 2.42 are given by L = 100 cm, P = 1000 N, E = 70 109 Pa,
and I = 2 cm4. Determine the deflection of the beam at the center using a single beam element.
Hint: Use the symmetry of the geometry and load of the beam.
2.19 A mechanical link is subjected to a symmetric distributed load p and concentrated loads P as shown in
Figure 2.43. Indicate the boundary conditions to be incorporated if only a quarter of the link is to be
modeled using suitable finite elements to determine the stresses induced in the link.
2.20 Generate the finite element mesh for the two-dimensional plate shown in Figure 2.44 using the quadtree
method.