Lectures in Finite Element Method

Ph. D. Students Course/ Structures Engineering

2hours/week
15 weeks course

Asst. Dr. Kifah M. Khudair

Basrah Univ.- College of Engineering- Dept of Civil Engineerig

1
Syllabus
Ch.1: Finite element method: Description and application

Ch.2: Finite element formulation

Ch.3: FEM for trusses

Ch.4: FEM for beams

Ch.5: FEM for frames

Ch.6: FEM for plates

Ch.7: FEM for flow in coastal water bodies

Ch.8: FEM for species transport

Ch.9: FEM for heat transfer

References
1- Davies, A. J. (2011) "The Finite Element Method: An Introduction with
Partial Differential Equations", 2nd Ed., Oxford University Press.
2- Fish, J. and Belytschko, T. (2007) " A First Course in Finite Elements",
JohnWiley & Sons, Ltd.
3- Hutton, D. V. (2004) "Fundamentals of Finite Element Analysis",
McGraw−Hill Companies.
4- Kwon, Y. W. and Bang, H. (1997) "The Finite Element Programming
Using MATLAB", CRC Press LLC.
5- Liu, G. R. and Quek, S. S. (2003) "The Finite Element Method: A
Practical Course", Elsevier Science Ltd.
6- Rao, S. S. (2011) "The Finite Element Method in Engineering", 5th Ed.,
Elsevier Inc.
7- Pinder, G. F. and Gray W. G. (1977) "Finite Element Simulation in
Surface and Subsurface Hydrology", ACADEMIC PRESS, INC.
8- Zienkiewicz, O. C. and Taylor, R. L. (2000)" The Finite Element Method:
Fluid Dynamics", 5th , published by Butterworth-Heinemann.

2
 Chapter 1

Finite Element Method: Description and Application

1.1 Introduction
The behavior of a phenomenon in a system depends upon the geometry or
domain of the system, the property of the material or medium, and the
boundary, initial and loading conditions. For an engineering system, the
geometry or domain can be very complex. Further, the boundary and initial
conditions can also be complicated. It is therefore, in general, very difficult to
solve the governing differential equation via analytical means. In practice, most
of the problems are solved using numerical methods. These methods include;
finite element method (FEM), finite difference method, finite volume method,
and boundary element method.

1.2 General Description of FEM

It is one of numerical methods that can be used to solve ordinary and partial
differential equations. In FEM, the actual continuous body of matter like solid,
liquid or gas is represented as an assemblage of subdivisions called finite
elements. These elements are considered to be interconnected at specified joints
which are called nodes or nodal points. The nodes usually lay on the element
boundaries where elements are connected. Since the actual variation of the field
variable (like stress, displacement, temperature, pressure, or velocity) inside the
continuum is not known, it is assumed that the variation of the field variable
inside a finite element can be approximated by simple function. These
approximating functions (also called interpolation models) are defined in terms
of the values of the field variables at the nodes. When field equations for the
whole body are written, the unknowns will be the nodal values of the field

3
variable. By solving the field equations, which are generally in the form of
matrix equations, the nodal values of the field variable will be known.

1.3 Steps of FEM Application

The steps of FEM application include:

1. Discretization of the domain.

2. Selection of interpolation function.
3. Finite element formulation (derivation of finite element equations).
4. Assemblage of element equations to obtain the overall equilibrium
equations (or final global equations).
5. Solution of finite matrix equations.
6. Computation of field functions and other variables.
1.3.1 Domain discretization

The discretization of the domain or solution region into sub regions (finite
element) is the first step in the FEM. By this step, the domain having an infinite
number of degrees of freedom is replaced by a system having finite number of
degrees of freedom. Here, the finite element mesh (or finite element grid map)
is formed. In discretization, the followings must be considered:

a- Type and order of the elements.

b- Size of the elements.
c- Number of the elements.
d- Location of nodes.
e- Nodes numbering.
Why do we discretize???

A- Type and order of elements

The type of the elements depends on the number of the independent variables of
the problem. It may be one-, two- or three-dimensional element.

4
 One- dimensional element

For example if we have a beam, the beam is usually discretized into a number of
one-dimensional elements.

1 2 3 4 5
The order of one-dimensional elements may be;

 1-D linear element (number of nodes=2)

 1-D quadratic element (number of nodes=3)
 1-D cubic element (number of nodes=4)
Two-dimensional elements

2-D elements can be triangular or quadrilateral (e.g., rectangular or

parallelogram) elements.

Order of triangular elements:

Linear element Quadratic element Cubic element

(3 nodes) (6 nodes) (10 nodes)
Order of quadrilateral elements

Linear element Quadratic element Cubic element

(4-nodes) (8-nodes) (12-nodes)

5
Three- dimensional elements

3-D elements include tetrahedron, hexahedron and pentahedron elements. They

may be linear, quadratic or cubic.

Note: as the order of elements increases, the accuracy of FEM increases

B- Size of elements

The size of elements influences the convergence of the solution and

hence it has to be chosen with care. If the size of elements is small, the final
solution is expected to be more accurate. However, the use of small elements
means more computational time. Different sizes of elements can be used with
the same problem domain. Generally, small size of elements must be used
where high gradients of field variables are expected.

6
 T2

T1

C- Number of elements

The number of elements is related to;

 Accuracy desired.
 Element size.
 Number of degrees of freedom.
Generally, at first a number of elements is chosen then solution is obtained.
After that, the number of elements is increased. If the results are not the same,
then, there is a need for increasing the elements number. This operation is called
mesh refinement.
Field variable value

No significant improvement
beyond Ne Use Ne

Ne
Number of elements

D- Location of nodes

If the body has no abrupt changes in geometry, material properties and external
conditions (like load, temp., etc..), the body can be divided into equal elements
and hence the spacing of nodes can be uniform. On the other hand if there are

7
any discontinuities in the problem, nodes have to be introduced at these
discontinuities.

A beam with different

cross section

The applied load is varied

abruptly

Soil- solid interaction

solid soil

E- Nodes numbering

Node numbering influences the computer storage requires to solve the

final global matrix. The band width (B) of the overall or global matrix depends
on node numbering scheme and the number of degrees of freedom per node.

B=(R+1)×f

8
Where R is the max largest difference in the node numbers occurring for all
elements of the assemblage and f is the number of degrees of freedom at each
node.

e.g.
2 7
if f=3

B= 7×3=21 1 6

Total storage= B× number of equations

Note: number the nodes in short direction at first

F- Mesh generation

The output of discretization step is mesh generation, which includes:

 Nodes generation
 Elements generation
Nodes generation produces a list contains the number and coordinates of each
node in the finite element mesh. In 2-D (Cartesian coordinates system) FE mesh
it will include;

Node number x- coordinate y- coordinate

Elements generation produces a list contains element numbers of the mesh and
the numbers of nodes associated with each element. For example, if the problem
domain is discretized using linear quadrilateral elements as shown below;

9
 6 9
3

4
2

5 8
2
1

1
4 3
4
4 7
1

Elements generation list shall be;

Element No. Node numbers
1 1 4 5 2
2 2 5 6 3
3 4 7 8 5
4 5 8 9 6
1.3.2 Interpolation functions

After the division of the problem domain into subdomains or elements, the
variation of the variable is defined approximately over the element in terms of
an interpolation function. The most used function type is polynomial
interpolation function. For example if we have a 1-D problem and one variable
say Ø, then;

Variation of Ø over the element =Øe

If linear elements are used;

Øe = α1+ α2 x …(1.1)

Quadratic elements;

Øe = α1+ α2 x+ α3 x2 … (1.2)
10
Cubic elements;

Øe = α1+ α2 x+ α3 x2+ α4 x3 …(1.3)

For 2-D problems;

Linear triangle elements;

Øe = α1+ α2 x+ α3 y …(1.4)

Quadratic triangle elements;

Øe = α1+ α2 x+ α3 y + α4x2+ α5 xy+ α6 y2 …(1.5)

Where; α1, α2 , α3 …and α6 are constants (or the coefficient for the monomials)

Cubic triangular elements (10 nodes)

Fig.(1.1): Pascal triangle of monomials (two- dimensional case)

11
1.3.2.1 Interpolation functions in terms of nodal values

Let Øe = [C] {α} …(1.6)

Where; [C] = [1 x y x2 xy y2] for quadratic triangular elements

{α}=

{ }

In terms of nodal values:

where; Øn is Ø value at node n.

{ } [ ]{ } ; {Ø} = nodal values vector

where; [D]= matrix of nodal points coordinates

[D]= …(1.7)

[ ]

12
We want the values of α1, α2, …….and α6

{ } [ ] { } …(1.8)

Substituting Eq.(1.8) into Eq.(1.6);

[ ][ ] { } …(1.9)

Let [ ] [ ][ ] ; N= shape function in terms of x and y

[ ]{ }

or;

[ ]

{ }

The variation of the field variable Ø over the element e is approximated as;

Where;

Ni= shape function of node i

n= number of nodes in element e

If the case of two field variables is considered;

Let u and v are displacements in x- and y-direction, respectively.

The variations of u and v over the element (e) are approximated as;

13
 ∑ & ∑

Hint: u and v may be defined using different types of shape functions.

u and v are connected into a vector de which is called displacement vector;

{ } [ ]

{ }

Linear interpolation polynomials in terms of global coordinates

1-D linear elements

Ø
For 1-D linear elements;
…(1.13)

j
i
x
𝑙

The solution of the above two equations gives;

Substituting the values of α1 and α2 into Eq.(1.13) gives;

Where; Ni and Nj are shape functions of nodes i and j, respectively.

[ ]{ }
14
Where;

[N] = matrix of shape functions (1×2)

{Ø}=vector of nodal unknowns of element e

Hints:

 Ni =1 at node i and Ni =0 at the other nodes.


2-D Triangular element

For linear elements;

…(1.16)

+
Ø
+
+
Øk

Øi k x
Øj
i
y j

Solving the above equations for α gives;

[( ) ( ) ]

[( ) ( ) ]

[( ) ( ) ]

Where; A=area of the triangle i j k given by;

| |

15
By substituting the values of α into Eq.(1.16) and arranging the results;

Where;

( )

Hints;
 Ni =1 at node i and Ni =0 at the other nodes and so on.


Where; Øi, Øj, and Øk are the nodal values of Ø and they are independent on x
or y. A constant value of and over an element means that many small
elements have to be used in locations where rapid changes of Ø are expected.
Linear interpolation polynomials in terms of local coordinates
The derivation of element matrices and vectors involves the integration of the
shape functions or their derivatives or both over the element. These integrals

16
can be evaluated easily if the interpolation functions are written in terms of a
local coordinates system. A local coordinates system is one which is located on
or within the boundaries of the element and it is different for different elements.

r-s local coordinates system

x-y global coordinates system s s
xc i

Where xc and yc denotes the j r

centroidal coordinates given by; k yc

( )
( )
H.W.1 Show that;

(̅ ̅ ̅ )

(̅ ̅ ̅ )

(̅ ̅ ̅ )
where;
̅ ̅ ̅ ⁄
̅ ̅ ̅
̅ ̅ ̅
H.W.2 Show that the integral of shape function in global coordinates system is
the same as the integral of shape function in local coordinates system.

Natural coordinates system

It is a local coordinates system that permits the specification of any point inside
the element by a set of non-dimensional numbers, which have magnitude lies
between 0 and 1.

17
1-D linear element

Any point in the element (p) is identified by two natural coordinates L 1 and L2
which are defined as;

and l
l2 l1
or; and 01 x 1 p 2
where; L1+L2=1 (L1,L2)
(1,0) (0,1)

From above it can be shown that;

and
Integration of polynomial terms in natural coordinates system can be done
using; ∫

2-D Linear triangular element

A natural coordinate system for a triangular element (also known as the

triangular coordinate system) is shown in the figure below. Although three
coordinates L1, L2, and L3 are used to define a point P, only two of them are
independent. The natural coordinates are defined as;

; ;

3
y
𝐴 p𝐴
𝐴 2

1
x

where A1 is the area of the triangle formed by the points P, 2 and 3; A 2 is the
area of the triangle formed by the points P, 1 and 3; A3 is the area of the triangle
formed by the points P, 1 and 2; and A is the area of the triangle 123. Because
Li are defined in terms of areas, they are also known as area coordinates. Since
18
A1 + A2 +A3 = A, then;
L1 + L2 +L3 = 1 …(1.17)
A study of the properties of L1, L2, and L3 shows that they are also the shape
functions for the 2-D triangular element:

; ;

To every set of natural coordinates (L1, L2, L3), there corresponds a unique set
of Cartesian coordinates (x, y). The relation between the natural and Cartesian
coordinates is given by;
x = x1L1 +x2L2 + x3L3 …(1.18)
y = y1L1 + y2L2 + y3L3 …(1.19)
At node 1, L1 = l and L2 = L3 = 0, and so on. The linear relationship between Li
(i = 1, 2, 3) and (x, y) implies that the contours of L1 are equally placed straight
lines parallel to the side 2, 3 of the triangle (on which L1 = 0), and so on.

19
Equations (1.17), (1.18) and (1.19) can be written in a matrix form as;

…(1.20)
Then;

…(1.21)
Where A is the area of the triangle 1, 2, 3 given by;

…(1.22)

Or;

{ } { }

Thus at any point (p) in the element there is a value of N1, N2 and N3. If f is a
function of L1, L2 and L3, the differentiation with respect to x and y can be
performed as;

Derive as an example

By using natural coordinates system, there is an exact expressions for integrals

in terms of natural coordinates.
20
  Integration over an edge
∫ Γ …(1.23)

Where; Γ= length of the edge where the integration is carried out.

Note: any point locates on the edge between points 1 and 3, as an example, has
coordinates in terms of L1 and L3. While a point inside the triangle has three
coordinates L1, L2 and L3.

 Integration over an area

∫ …(1.24)

e.g.; if we have an integration

∫ ∫

3- D Tetrahedron linear element

The natural coordinates for a tetrahedron element can be defined analogous to
those of a triangular element. Thus, four coordinates L1, L2, L3, and L4 will be
used to define a point P, although only three of them are independent. These
natural coordinates are defined as;

; ;
where Vi is the volume of the tetrahedron formed by the points P and the
vertices other than the vertex i (i = 1, 2, 3, 4), and V is the volume of the
tetrahedron element defined by the vertices 1, 2, 3, and 4 (Figure 3.15). Because
the natural coordinates are defined in terms of volumes, they are also known as
volume or tetrahedral coordinates. Since;

21
V1 +V2 + V3 +V4 = V
Then;
L1 +L2 + L3 +L4 = 1
The volume coordinates L1, L2, L3, and L4 are also the shape functions for a
three-dimensional simplex element:
; ; ;

The Cartesian and natural coordinates are related as;

22
Then;

The inverse relation can be expressed as;

Where;

23
Also;

∫ 5

Higher Order Elements in Terms of Natural Coordinates

H.W
1- Show that for 1-D quadratic elements;
N1 = L1 (2L1−1) ; N2 = 4L1L2 ; N3 = L2 (2L2 −1)

2- Show that for 2-D quadratic triangular elements;

; ;

5
6
2

4
1

2-D Quadrilateral elements

A different type of natural coordinate system can be established for a

quadrilateral element in two dimensions as shown in the figure below. For the
local r, s (natural) coordinate system, the origin is taken as the intersection of
lines joining the midpoints of opposite sides and the sides are defined by r = ±1

and s = ±1.

24
1- Linear quadrilateral elements

; i=1, 2, 3, 4 …(1.26)

2- Quadratic quadrilateral element

For nodes at r=±1 , s=±1

For nodes at r=0 , s=±1

…(1.28)

For nodes at r =±1 , s=0

Nodes Mapping

The nodes in r-s plane may be mapped into corresponding nodes in x- y plain
using the following relations;

∑ [ ]{ } …(a)

∑ [ ]{ } …(b)

25
Where {xi} and {yi} are list the nodal coordinates x and y and Ni (i=1, 2, … n)
are shape functions in terms of natural coordinates.

Based on the order of shape function used to describe the geometry, the
elements are classified into three types:

1. Isoparametric element: In this type, the geometry and field variable

are defined using the same order of shape function.
2. Subparametric element: In this type, the geometry is described by a
lower order shape function than that used for describing the field
variable.
3. Superparametric element: In this type, the geometry is described by
a higher order shape function.
Matrix transformation

The finite element formulations usually contain terms of N, and . Since N

is expressed as a function of local coordinates r and s, it is necessary to express

, and dxdy in terms of r and s, also. This can be done as;

Or;

{ } [ ]
{ }

26
Let [J] =[ ]

Then;

{ } [ ]
{ }

Where [J] is defined as a Jacobian matrix and evaluated by using

equations (a) and (b) for each element. The desired derivatives are
obtained by inverting the above equation and this involves the finding
of Jacobian matrix inverse;

{ } [ ] { } …(1.30)

Similarly;

| |

Where | | [ ]

27
 Chapter two

Finite Element Formulation

2.1 Methods of Finite Element Formulation

Consider a domain Ω described by a set of independent variables x 1, x2, x3,….,

xn and typically say xj. Let the dependent variables be Ø1, Ø2, Ø3, …., Øm and
typically say Øi. In the domain, the relationships between Øi and xj are
expressed by one or more of field equations of the form;

There will be also one or more boundary equations;

The set of the equations (2.1) and (2.2) are the governing equations of the
system.

In the finite element procedure, once the mesh for the solution domain is
decided, the behavior of the field variables over the domain is approximated as;

̂ ∑

Where the circumflex distinguishes the approximation ̂ from the true solution
Øi and are the nodal values of Øie.

or;

̂ [ ]{ } k=1,l …(2.4)

Over the domain, ̂ values are given by;

28
{̂ } ∑ [ ]{ } …(2.5)

The finite element method is subdivided according to the procedure by

which the equations in the nodal values are formulated into;

1- Weighted residual method.

2- Variational principle method.
3- Direct finite element method.
2.2 Weighted Residual Method

The method of weighted residuals is a technique for obtaining

approximate solutions to linear and nonlinear partial differential equations.

When ̂ is substituted into Eq.2.1, the equation will not be satisfied, that is;

(̂ )

where R is the residual or error that results from approximating by ̂ The

weighted residual method seeks to determine the nodal unknowns in such a way
that the error R over the entire solution domain is small.

In order to make R identically zero, a set of weighting functions W are

employed such that;

For the domain;

Where W is any function. If the number of unknowns is nn and if nn

independent weighting functions Wk are chosen, Eq.(2.7) can be written as;

29
∫ ∫ (̂ )

There are a variety of weighted residual techniques because of the board

of wk or error distribution principle. These techniques include:

 Collection method.
 Subdomain collection method.
 Galerkin method.
 Least squares method.
Galerkin method is generally the most used one and gives the best approximate
solution.

H.W:
Write a brief report about collection, subdomain collection and least squares
methods.
2.2.1 Galerkin WRM

According to this method the weighting functions are chosen to be the same as
the shape functions used to represent Øi, that is Wk=Nk for k=1, 2, 3,…,nn. Thus
Eq.(2.8) becomes;

∫ (̂ ) . …(2.9)

In the finite element method, for each subdomain or element, the

contribution would be;

∫ (̂ )

such that for the whole domain;

30
∑ ∫ (̂ ) …(2.11)

Where is element domain and ne is the total number of elements.

The choice of shape functions ensures continuity of Ø at the element

boundaries as well as continuity of its partial derivatives up to one order less
than the highest order derivatives appearing in the expression to be integrated.

For example when the highest order derivative is second order ( , then, C1

and C0 continuity are required, i.e., Ø and must be continuous. To escape C1

continuity requirement, integration by parts is applied to the second order

derivative terms. Integration by part offers a convenient way to introduce the
natural boundary conditions that must be satisfied on some portion of the
boundary.

Then, if second order derivative is existing in the governing equations, the

following procedure is adopted;

1- Use the product rule for differentiation first which produces

the following expression:

…(2.12)

2- Use Gauss’s divergence theorem which can be stated

mathematically as;

…(2.13)
where θ is the angle of the outwards normal on the boundary Γe of the
element with respect to the x-axis.

31
2.3 Variational Principle Method

In this method the variation of the dependent variable over the whole
domain is represented in terms of an integral functional (I). The term
"functional" means function of group of functions and I usually has a clear
physical meaning in most of the applications.

∫ ( ) ∫ ( ) …(2.14)

For example, in structural and solid mechanics, the potential energy (п) plays
the role of functional (п is a function of the displacement vector Ø, whose
components are u, v and w, which is a function of coordinates x, y and z).

Here, we are looking for an exact solution to satisfy the boundary conditions
and make the integral functional minimum. From mathematical point of view,
the solution that satisfies the boundary conditions and minimizes the integral
functional will satisfy the partial differential equation (i.e., solution to the
P.D.E.).

The steps of solution are:

1- Write the integral functional for each element;

∫ ( ) ∫ ( ) …(2.15)

2- Approximate Øe ( ∑ and substitute it into Eq.(2.15)

3- Write I for the whole domain;

where I is in term of nodal values for the whole domain.

32
 4- The solution is that makes I minimum;
…(2.16)

where i=1, 2, ……, n and n=total number of degrees of freedom for the
whole domain.

That means;

∑
∑ …(1)

∑ …(2)

∑ …(3)

∑ …(n)

The unknowns are the nodal values so we get the matrix equation for the whole
domain.

2.3.1Variational Principle in Elasticity Problems

Theorem of Potential Energy

Of all displacements satisfying the given boundary conditions, those

satisfy the equations of equilibrium are distinguished by a stationary (extreme)
value of potential energy.

33
 Total potential energy (п)

A component related to potential

A component resulting from energy of the internal and applied
strain energy (пs) loads (пp)

where wp= work done by internal and applied loads.

Strain energy

Strain energy (пs) for an elastic body of volume dv under an initial strain of
is;

{ } { }- { } { }

where; for 2-D elasticity problems

{ } { }

{ } { }

∫ [{ } { } { } { }]

or;

34
 ∫ [{ } { } { } { }] …(2.17)

From Hook's law;

{ } [ ]{ } [ ]{ } …(2.18)

Where [D] is material properties matrix and it depends on the considered

problem, whether it is plane strain or plain stress problem. For 2-D problems,
[D] is 3×3 matrix. For plane stress problem;

[ ] ; E=modulus of elasticity and = Poisson's ratio

For plane strain problems;

[ ]

For 2-D problems, we have displacement in x-dir. (u) and displacement in y-dir.
(v). Approximate ue and ve using an interpolation function;

[ ]{ } & [ ]{ }

[[ ]{ }] [ ]{ }

35
 [[ ]{ }] [ ]{ }

[[ ]{ }] [[ ]{ }]

Generally,

[ ]{ }

Where; [B]= matrix contains derivative of N. if N is linear, then, [B]=constant

∫ [{ } [ ] [ ][ ]{ } { } [ ] [ ]{ }

{ } [ ]{ }]

Properties

(1) { } [ ]{ } (2) { } { } { } { }
{ } { } [ ]
Term-2 Wp

There are three types of loads;

1- Self-weight (internal or body force)

2- Concentrated external load.
3- Distributed external load.
1-Self weight

Let X, Y, and Z are body forces in x-, y-, and z-directions, respectively, then;

∫ { } [ ] { }

36
Since the body forces are variable for each element;

∫ { } [ ] { }

2- Concentrated external load

Usually a node is located at the location of concentrated load;

{ } { } { } { }

Note: if we have a distributed load p acting on the surface of

the element (S), analyze it into Px, Py, and Pz and

Wx=Px u; Wy=Py v; Wz=Pz w

For the whole domain;

By minimizing п (for the whole domain);

∑
{ } { }

Where {U} is displacement nodal values vector. For 3-D problems;

{ } { }

By substituting пs and Wp;

37
 ∑ [∫[ ] [ ][ ] { } ∫ [ ] [ ]{ }
{ }

∫ [ ] { } ∫ [ ] { } ] { }

The above formulation gives;

[ ]{ } { } { }

Where;

[ ] ∑ ∫[ ] [ ][ ]

But

[ ] ∑ [ ]

[ ] ∫[ ] [ ][ ]

[ ] element stiffness matrix

{ } ∑ { }

{ } ∫ [ ] [ ]{ } ∫ [ ] { } ∫ [ ] { }

{ } { }

Notes: For non-linear elements, use numerical integration.

38
 [{ } [ ]{ }] [ ]{ }
{ }

Equations For Two-Dimensional Solids

Stress and Strain
Three-dimensional problems can be simplified if they can be treated as 2-D
solid. For representation as a 2D solid, one coordinate (usually the z-axis) is
usually removed, and hence assume that all the dependent variables are
independent of the z-axis, and all the external loads are independent of the z
coordinate, and applied only in the x–y plane. Therefore, a 2-D system has only
the x and y coordinates. There are primarily two types of 2D solids; plane stress
solid and plane strain solid.
Plane stress solids are solids whose thickness in the z direction is very
small compared with dimensions in the x and y directions. All the external
forces are applied within the x–y plane, and hence the displacements are
functions of x and y only. Herein, the stresses in the z direction (zz, xz and yz)
are all zero, as shown in Fig. (2.1)
Plane strain solids are those solids whose thickness in the z direction is
very large compared with the dimensions in the x and y directions and the cross-
section and the external forces do not vary in the z direction. External forces are
applied evenly along the z axis, and the movement in the z direction at any point
is constrained. The strain components in the z direction (εzz, εxz and εyz) are,
therefore, all zero, as shown in Fig.(2.2). Note that for the plane stress
problems, the strains εxz and εyz are zero, but εzz will not be zero.

39
 Fig.(2.1) Plane stress problem. Fig.(2.2): Plane strain problem.

For 1-D elasticity problems

∫ { } { }

The potential energy of the external forces, being opposite in sign from the
external work expression because the potential energy of external forces is lost
when the work is done by the external forces, is given by;

∫ ∫ ∑

Where; the first, second, and third terms on the right side of the above equation
represent the potential energy of (1) body forces, typically from the self-weight
of the bar (in units of force per unit volume) moving through displacement
function u, (2) surface loading or traction Tx, typically from distributed loading
acting along the surface of the element (in units of force per unit surface area)
moving through displacements us, where us are the displacements occurring
over surface S, and (3) nodal concentrated forces fix moving through nodal
displacements uix. The three forces are considered to act in x- direction of the
bar as shown in Fig.(2.3). V is the volume of the body and S1 is the part of the
surface S on which surface loading acts. For a bar element with two nodes and
one degree of freedom per

40
node, m =2.

Fig.(2.3): General forces acting on a one-dimensional bar

From Hook's law;

{ } [ ]{ }

∑ [ ∫[ ] [ ][ ] { } ∫ [ ] { }
{ }

∫ [ ] { } ] { }

The above formulation gives;

[ ]{ } { } { }

Where;

[ ] ∑ ∫[ ] [ ][ ]

But

[ ] ∑ [ ]

[ ] ∫[ ] [ ][ ]

41
[ ] element stiffness matrix

{ } ∑ { }

{ } ∫ [ ] { } ∫ [ ] { }

{ } { }

H.W.1: Solve the following problem using Galerkin weighted residual method
and variational principle method.

For the bar subjected to the linear varying axial load shown below, determine
the nodal displacements and axial stress distribution using (a) two equal-length
elements and (b) four equal-length elements. Take E=210 GPa and A=4×10-4 m2

10x kN/m

4m
Hint: for the simple two-noded bar element subjected to a linearly varying load
(triangular loading), place one-third of the total load at the node where the
distributed loading begins (zero end of the load) and two-thirds of the total load
at the node where the peak value of the distributed load ends.

42
 Chapter Three

FEM for Trusses

3.1 General
A truss is one of the simplest and most widely used structural members. It is a
straight bar that is designed to take only axial forces, therefore it deforms only
in its axial direction. The cross-section of the bar can be arbitrary, but the
dimensions of the cross-section should be much smaller than that in the axial
direction. Finite element equations for such truss members will be developed in
this chapter. The element developed is commonly known as the truss element or
bar element. Such elements are applicable for analysis of the skeletal type of
truss structural systems both in two-dimensional planes and in three-
dimensional space.
In planar trusses there are two components in the x and y directions for the
displacements as well as for the forces. For space trusses, however, there will be
three components in the x, y and z directions for both displacements and
forces. In skeletal structures consisting of truss members, the truss members are
joined together by pins or hinges (not by welding), so that there are only forces
(not moments) transmitted between the bars.
3.2 FEM Equations
Consider a structure consisting of a number of trusses or bar members. Each of
the members can be considered as a truss/bar element of uniform cross-section
bounded by two nodes. Consider a bar element with nodes 1 and 2 at each end
of the element, as shown in Fig.(3.1). The length of the element is le. The local
x- axis is taken in the axial direction of the element with the origin at node 1. In
the local coordinate system, there is only one DOF at each node of the element,

43
and that is the axial displacement (u). Therefore, there is a total of two DOFs for
the element.

Fig.(3.1): Truss element and the coordinate system.

For 1-D linear element, the matrix of shape functions is;
[N]= [N1 N2]
where the shape functions for a truss element can be written as;

3.2.1 Element Matrices in the Local Coordinate System

The stiffness matrix for truss elements can be written as;
[ ] ∫ [ ] [ ][ ] ∫ [ ] [ ]

[ ] 3.1)

where Ae is the area of the cross-section of the truss element.

The mass matrix for truss elements can be obtained as:

[ ] ∫ [ ] [ ]

∫ [ ] [ ]

44
 To find the nodal force vector for truss elements, suppose the element is
loaded by an evenly distributed force fx along the x-axis, and two concentrated
forces fs1and fs2, respectively, at two nodes 1 and 2, as shown in Fig.1; the total
nodal force vector becomes;

{ } ∫ [ ] { } ∫ [ ] { }

{ }

3.2.2 Element Matrices in the Global Coordinate System

Element matrices in Eqs. (3.1), (3.2) and (3.3) were formulated based on the
local coordinate system, where the x-axis coincides with the mid axis of the bar
1–2, shown in Fig.(3.1). In practical trusses, there are many bars of different
orientations and at different locations. To assemble all the element matrices to
form the global system matrices, a coordinate transformation has to be
performed for each element to formulate its element matrix based on the global
coordinate system for the whole truss structure. The following performs the
transformation for both spatial and planar trusses.
A- Spatial trusses
Assume that the local nodes 1 and 2 of the element correspond to the global
nodes i and j, respectively, as shown in Fig.(3.1). The displacement at a global
node in space should have three components in the X, Y and Z directions, and
numbered sequentially. For example, these three components at the ith node are
denoted by D3i−2, D3i−1 and D3i. The coordinate transformation gives the
relationship between the displacement vector {U} based on the local coordinate
system and the displacement vector {D} for the same element, but based on the
global coordinate system XYZ:

45
 { } [ ]{ } …(3.4)
Where;

{ }

{ }
and [T] is the transformation matrix for the truss element, given by;

In which;

are the direction cosines of the axial axis of the element. The length of the
element, le, can be calculated using the global coordinates of the two nodes of
the element by;

The matrix [T] for a truss element transforms a 6 × 1 vector in the global
coordinate system into a 2 × 1 vector in the local coordinate system. The
transformation matrix is, also, applied to the force vectors between the local and
global coordinate systems:
{ } [ ]{ }

46
 { }

{ }
in which F3i−2, F3i−1 and F3i stand for the three components of the force vector at
node i based on the global coordinate system.
The element equation based on the global coordinate system can be
written as:
[ ][ ]{ } [ ]{ }
Multiply both side by [T]T;
[ ] [ ][ ]{ } [ ] [ ]{ }
Let [Ke] = element stiffness matrix in global coordinates system, then;
[Ke]= [ ] [ ][ ]
Since[ ] [ ] = identity matrix ; Prove?
then;
[ ]{ } { } …(3.5)
Where [Ke]=

Where [Ke]=

47
 and {Fe}=

Note that the element stiffness matrix [Ke] has a dimension of 6 × 6 in the three-
dimensional global coordinate system, and the displacement {D} and the force
vector {Fe} have a dimension of 6 × 1.
B- Planar trusses
For a planar truss, the global coordinates X–Y can be employed to represent the
plane of the truss. All the formulations of coordinate transformation can be
obtained from the counterpart of those for spatial trusses by simply removing
the rows and/or columns corresponding to the z- (or Z-) axis. The displacement
at the global node i should have two components in the X and Y directions only:
D2i−1 and D2i . The coordinate transformation, which gives the relationship
between the displacement vector {U} based on the local coordinate system and
the displacement vector {De}, has the same form as Eq. (3.4), except that {De}
and {Fe} are 4×1 vectors and [T] is 2×4 matrix.
3.3 Boundary Conditions
The stiffness matrix [Ke] in Eq. (3.5) is usually singular, because the whole
structure can perform rigid body movements. There are two DOFs of rigid
movement for planer trusses and three DOFs for space trusses. These rigid body
movements are constrained by supports or displacement constraints. In practice,
truss structures are fixed somehow to the ground or to a fixed main structure at
a number of the nodes. When a node is fixed, the displacement at the node must
be zero. This fixed displacement boundary condition can be imposed on Eq.

48
(3.5). The imposition leads to a cancellation of the corresponding rows and
columns in the stiffness matrix.
3.4 Determination of Stress and Strain
Eq. (3.5) can be solved using standard routines and the displacements at all the
nodes can be obtained after imposing the boundary conditions. The
displacements at any position other than the nodal positions can also be
obtained using interpolation by the shape functions. The stress in a truss
element can also be determined using the following equation:
σx = E[B]{U} = E[B][T]{De}
In deriving the above equation, Hooke’s law in the form of σ = Eε is used.

49
 Chapter Four

FEM for Beams

4.1 General
A beam is another simple but commonly used structural component. It is also
geometrically a straight bar of an arbitrary cross-section, but it deforms only in
directions perpendicular to its axis. Note that the main difference between the
beam and the truss is the type of load they carry. Beams are subjected to
transverse loading, including transverse forces and moments that result in
transverse deformation. Herein, the cross-section of the beam structure is
assumed uniform. If a beam has a varying cross-section, the beam should be
divided into shorter beams, where each can be treated as a beam with a uniform
cross-section.
4.2 FEM Equations
In planar beam elements there are two degrees of freedom (DOFs) at a node in
its local coordinate system. They are deflection in the y direction, v, and rotation
in the x–y plane, θz with respect to the z-axis. Therefore, each beam element has
a total of four DOFs.
4.3 Shape Function Construction
Consider a beam element of length l with nodes 1 and 2 at each end of the
element, as shown in Fig.1. The local x-axis is taken in the axial direction of the
element with its origin at the beginning of the beam. Similar to all other
structures, to develop the FEM equations, shape functions for the interpolation
of the variables from the nodal variables would first have to be developed. The
transverse displacement model for the beam element is assumed to be a cubic
polynomial involving four constants.

50
 V
(x)
0 x
Node Node
1 l 2
The cubic interpolation function for the displacement of a beam element
is expressed as;
v)x( = α1 +α2x+α3x2 +α4x3

with the nodal degrees of freedom defined as v1 = v(x = x1), ϑ1 = (dv/dx) (x =

x1), v2 = v(x = x2), and ϑ 2 = (dv/dx)(x = x2), where x1 and x2 denote the x
coordinates of nodes 1 and 2 of the element. The shape functions are defined in
local coordinates system as;

And the displacement function for the element e can be written as;
[ ]{ } …(4.1)
Where {d} is displacement vector defined as;

{ } { }

4.4 Strain Matrix

After the obtaining of the shape functions, the next step would be to obtain the
element strain matrix. The relationship between the strain (εXX) and the
deflection (v) is;

…(4.2)

51
Substituting Eq. (4.1) into Eq. (4.2), which gives the relationship between the
strain and the deflection;
[ ]{ }
where the strain matrix B is given by;

[ ] [ ]

[ ] [ ]

4.5 Element Matrix

The element stiffness matrix is defined as;

[ ] ∫ [ ] [ ][ ] …(4.4)

By substituting Eq. (4.3) into Eq. (4.4), the stiffness matrix can be obtained as;

[ ] ∫ [ ] [ ][ ] ∫ ∫ [ ] [ ]

∫ [ ] [ ] …(4.5)

where ∫ dA) is the second moment of area (or moment of inertia) of

the cross section of the beam with respect to the z-axis. Eq.(4.5) can be
rewritten as;

[ ] ∫

[ ]

52
Evaluating the integrals in the above equation leads to;

6 3L 6 3L 
3L 2L2 3L L2 
2EI
 K e   3 Z   …(4.6)
L 6 3L 6 3L 
 
3L L
2
3L 2L 2 

4.6 Force Vector

Suppose the element is loaded by an external distributed force fy along the x-

axis, two concentrated forces fs1 and fs2, and concentrated moments ms1 and ms2,
respectively, at nodes 1 and 2; the total nodal force vector becomes;

{ } ∫ [ ] { } ∫ [ ] { }

{ ⁄ }
4.7 Beam Finite Element Equation

The final finite element equation of a beam is defined as;

[ ]{ } { } …(4.8)

Where [ke] and { } are defined using Eqs. (4.6) and (4.7), respectively.

53
 Chapter Five

FEM for Frames

5.1 Frame Element

A frame element is formulated to model a straight bar of an arbitrary

cross-section, which can deform not only in the axial direction but also in the
directions perpendicular to the axis of the bar. The bar is capable of carrying
both axial and transverse forces, as well as moments. Therefore, a frame
element is seen to possess the properties of both truss and beam elements.

Frame elements are applicable for the analysis of skeletal type systems of
both planar frames (two-dimensional frames) and space frames (three-
dimensional frames). Frame members in a frame structure are joined together by
welding so that both forces and moments can be transmitted between members.
Herein, it is assumed that the frame elements have a uniform cross sectional
area. If a structure of varying cross-section is to be modeled using the
formulation given below, then the structure must be divided into smaller
elements of different constant cross-sectional area so as to simulate the varying
cross-section. If the variation in cross-section is too severe for accurate
approximation, then the equations for a varying cross-sectional area can also be
formulated without much difficulty using the same concepts and procedure.

5.2 FEM Equations for Planar Frames

Consider a frame structure whereby the structure is divided into frame

elements connected by nodes. Each element is of length le = 2a, and has two
nodes at its two ends. The elements and nodes are numbered separately in a
convenient manner. In a planar frame element, there are three degrees of
freedom (DOFs) at one node in its local coordinate system, as shown in

54
Fig.(5.1). They are the axial deformation in the x direction, u; deflection in the y
direction, v; and the rotation in the x–y plane and with respect to the z-axis, θz.
Therefore, each element with two nodes will have a total of six DOFs.

Fig.(5.1). Planar frame element and the DOFs.

5.2.1 Equations in Local Coordinate System

Considering the frame element shown in Fig.(5.1) with nodes labeled 1

and 2 at each end of the element, it can be seen that the local x-axis is taken as
the axial direction of the element with its origin at the middle of the element. As
mentioned, a frame element contains both the properties of the truss element
and the beam element. Therefore, the element matrices for a frame element can
be simply formulated by combining element matrices for truss and beam
elements. Recall that the truss element has only one degree of freedom at each
node (axial deformation), and the beam element has two degrees of freedom at
each node (transverse deformation and rotation). Combining these will give the
degrees of freedom of a frame element, and the element displacement vector for
a frame element can thus be written as;

55
 …(5.1)

To construct the stiffness matrix, the stiffness matrix for truss elements is

extended at first to a 6×6 matrix corresponding to the order of the degrees of

freedom of the frame element in the element displacement vector in Eq. (5.1):

…(5.2)

Next, the stiffness matrix for the beam element is also extended to a 6×6
matrix corresponding to the order of the degrees of freedom of the frame
element in Eq. (5.1):

…(5.3)

56
 The two matrices in Eqs. (5.2) and (5.3) are now superimposed together
to obtain the stiffness matrix for the frame element:

…(5.4)

The same simple procedure can be applied to the force vector as well.
The element force vectors for the truss and beam elements are extended into
6×1 vectors corresponding to their respective DOFs and added together. If the
element is loaded by external distributed forces fx and fy along the x-axis;
concentrated forces fsx1, fsx2, fsy1 and fsy2; and concentrated moments ms1 and ms2,
respectively, at nodes 1 and 2, the total nodal force vector becomes;

…(5.5)

5.2.2 Equations in Global Coordinate System

The matrices formulated in the previous section are for a particular frame
element in a specific orientation. A full frame structure usually comprises
numerous frame elements of different orientations joined together. As such,

57
their local coordinate system would vary from one orientation to another. To
assemble the element matrices together, all the matrices must first be expressed
in a common coordinate system, which is the global coordinate system. The
coordinate transformation process is the same as that described for truss
structures.

Assume that local nodes 1 and 2 correspond to the global nodes i and j,
respectively. The displacement at a local node should have two translational
components in the x and y directions and one rotational deformation. They are
numbered sequentially by u, v and θz at each of the two nodes, as shown in Fig.
(5.2). The displacement at a global node should also have two translational
components in the X and Y directions and one rotational deformation. They are
numbered sequentially by D3i−2, D3i−1 and D3i for the ith node, as shown in
Fig.(5.2). The same sign convention also applies to node j. The coordinate
transformation gives the relationship between the displacement vector de based
on the local coordinate system and the displacement vector De for the same
element, but based on the global coordinate system:

Fig.(5.2). Coordinate transformation for 2D frame elements.

58
 5 6)

Where;

…(5.7)

and T is the transformation matrix for the frame element given by;

…(5.8)

in which;

…(5.9)

where α is the angle between the x-axis and the X-axis, as shown in Fig.(5.2),
and;

…(5.10)

Note that the coordinate transformation in the X–Y plane does not affect
the rotational DOF, as its direction is in the z direction (normal to the x–y

59
plane), which still remains the same as the Z direction in the global coordinate
system. The length of the element, le, can be calculated by;

…(5.11)

The matrix T for a frame element transforms a 6 × 6 matrix into another 6

× 6 matrix. Using the transformation matrix, T, the matrices for the frame
element in the global coordinate system become;

Note that there is no change in dimension between the matrices and vectors in
the local and global coordinate systems.

60
 Chapter Six

FEM for Plates

6.1 Plate Elements

A plate structure is geometrically similar to the structure of the 2D plane

stress problem, but it usually carries only transversal loads that lead to bending
deformation in the plate. For example, higher floors of a building are a typical
plate structure that carries most of us every day. The plate structure can be
schematically represented by its middle plane laying on the x–y plane, as shown
in Fig.(6.1). The deformation caused by the transverse loading on a plate is
represented by the deflection and rotation of the normals of the middle plane of
the plate, and they will be independent of z and a function of only x and y. The
element to be developed to model such plate structures is known as the plate
element.

Fig.(6.1): A plate and its coordinates system.

Plate elements are normally used to analyze the bending deformation of
plate structures and the resulting forces such as shear forces and moments. In
this aspect, it is similar to the beam element, except that the plate element is
two-dimensional whereas the beam element is one-dimensional. A plate element
can also be triangular, rectangular or quadrilateral in shape. In this section, we
61
the development of the rectangular element is covered only, as it is often used.
Matrices for the triangular element can also be developed easily using similar
procedures.

There are a number of theories that govern the deformation of plates. In

this section, rectangular elements based on the Reissner–Mindlin plate theory
that works for thick plates will be developed. There are a number of higher
order plate theories that can be used for the development of finite elements. It is
assumed that the element has a uniform thickness h. If the plate structure has a
varying thickness, the structure has to be divided into small elements that can be
treated as uniform elements. However, the formulation of plate elements with a
varying thickness can also be done, as the procedure is similar to that of a
uniform element. Consider now a plate that is represented by a two-dimensional
domain in the x–y plane, as shown in Fig.(6.1). The plate is divided in a proper
manner into a number of rectangular (linear quadrilateral) elements of four
nodes, as shown in Fig.(6.2). At a node, the degrees of freedom include the
deflection w, the rotation about x axis θx , and the rotation about y axis θy,
making the total DOF of each node three. Hence, for a rectangular element with
four nodes, the total DOF of the element would be twelve. Following the
Reissner–Mindlin plate theory, its shear deformation will force the cross-section
of the plate to rotate in the way shown in Fig.(6.3). Any straight fiber that is
perpendicular to the middle plane of the plate before the deformation rotates,
but remains straight after the deformation. The two displacement components
that are parallel to the middle surface (u and v) at a distance z from the
centroidal axis can be expressed by;

…(6.1)

…(6.2)
62
 where ϑx and ϑy are, respectively, the rotations of the fiber of the plate
with respect to the x and y axes. The in-plane strains can then be given as;
{ } { } …(6.3)

where {χ} is the curvature vector, given as;

⁄
{ } { ⁄ } …(6.4)
⁄ ⁄

Fig.(6.2): 2D domain of a plate

Fig.(6.3): Shear deformation in a plate.
meshed by rectangular elements.

The off-plane shear strain is then given as;

{ } { } 5
{ }

The potential (strain) energy expression for a thick plate element is;

∫ [{ } { } { } { }] …(6.6)

or;

63
 ∫ ∫ { } { } ∫ ∫ { } { } …(6.8)

The first term on the right-hand side of the above equation is for the in-
plane stresses and strains, whereas the second term is for the transverse stresses
and strains. {τ} is the average shear stresses vector relating to the shear strain in
the form;

{ } { } [ ]{ } [ ]{ }

where G is the shear modulus, and κ is a constant that is usually taken to

be π2/12 or 5/6. Substituting Eqs. (6.3) and (6.9) into Eq. (6.8), the potential
(strain) energy becomes;

∫ { } [ ]{ } ∫ { } [ ]{ } …(6.10)

6.2 Shape Functions

For four-node rectangular thick plate elements, the deflection and

rotations can be approximated as;

∑ …(6.11-a)

∑ …(6.11-b)

∑ …(6.11-c)

where the shape function Ni in natural coordinates system is;

…(6.12)

Rewriting Eqs. (6.11a-c) into a single matrix equation;

64
 { }=[ ]{ } …(6.13)

where {d} is the displacement vector for all the nodes in the element, arranged
in the order;

{ }

{ }

6.3 Element Matrices

To obtain the stiffness matrix [ke], Eq. (6.13) is substituted into Eq.
(6.10);

[ ] ∫ [ ] [ ][ ] ∫ [ ] [ ][ ]

The first term in the above equation represents the strain energy
associated with the in-plane stress and strains. The strain matrix [BI ] has the
form of ;

[ ] [ ]

where;

⁄
[ ⁄ ], j=1,…,4
⁄ ⁄

65
The strain matrix [Bo] has the form;

[ ] [ ]

where;

⁄
[ ], j=1,…,4
⁄

The integration in the stiffness matrix [ke], can be evaluated numerically

using the Gauss integration scheme. The simplest and most practical means to
solve this problem is to use 2 ×2 Gauss points for the integration of the first
term, and use only one Gauss point for the second term in Eq. (6.14).

6.4 Force Vector

The force vector is written as;

{ } ∫ [ ] { }dA

66