Basics of FEM

Some terminologies used in FEM

1. Discretization (Meshing): The process of dividing the model of the problem continuum
into a finite number of regular subdivisions
2. Elements: Each subdivision is called an “Element”
3. Nodes: The grid (connection) points at which the elements meet each other are called
“nodes”
4. Degrees of Freedom (DOF): The total number of variables (displacements) that are
associated with each node
5. Boundary Conditions: Known values of the variable at the continuum boundary
6. Displacement Function: It is an assumed polynomial expression which closely represents
the anticipated variation of the unknown variable over the element domain.

Problem Dimensions:
One Dimensional Problems: If only one independent coordinate axis is sufficient to
represent the displacement, material and geometry of the problem, then such problems are
known as One Dimensional Problems. Line Elements are used to model 1D problems.

Two Dimensional Problems: If two independent coordinate axes are required to represent
the displacement, material and geometry of the problem, then such problems are known as
Two Dimensional Problems. Area Elements are used to model 2D problems.

Three Dimensional Problems: If three independent coordinate axes are required to represent
the displacement, material and geometry of the problem, then such problems are known as
Three Dimensional Problems. Volume Elements are used to model 3D problems.

Basic types of Elements used in FEM:

Discretization is a process of engineering judgment. A selection of the shape or
configuration of the basic element in FEM is very important. The choice of the element
depends upon the geometry (structure) of the problem and the number of independent space
coordinates necessary to describe the system.
A finite element has usually a One or Two or Three dimensional configuration. The
boundary of the element may be straight (in most of the cases) or curvilinear.

One Dimensional Elements (Line Elements): When the geometry, material and
displacement can be expressed in terms of only independent space coordinate, a 1D element
may be chosen. The coordinate is measured along the axis of the element.

1 3 2

External Node Internal Node External Node

A 1D element can be represented by a straight line whose end points are nodes (nodal
points). The nodal points numbered 1 & 2 are called external nodes because they represent
connecting points with adjacent elements. Some applications may require additional nodes,
such as node 3. It is called an internal node because it doesn’t form a connecting point with
adjacent elements. Additional nodes are usually used to improve the FEM result.

Two Dimensional Elements (Area Elements): When the geometry, material and
displacement can be expressed by minimum two independent space coordinate, a 2D element
Compiled by Vijay G. S., Asst. Prof., 1 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

may be chosen. Problems that can be modeled with 2D elements are plane stress, plane strain
and axisymmetric problems.

A 2D element can be represented by a triangular or

quadrilateral configuration.
3
In the triangular element, the corner nodes (1, 2 & 3) 5
are called primary external nodes. Additional nodes on the
sides (like nodes 4, 5 & 6) are called secondary external 6
7 2
nodes. An internal node (like node 7) may be used in a
triangular element. 4
Similarly, a quadrilateral configuration may have a
minimum of 4 primary external nodes. As well they may 1
contain the additional secondary external nodes and the
internal nodes. A special case of the quadrilateral
configuration is the rectangular configuration.
4 10 9 3
3
7
11 16 15 8
4
6
9 12 13 14 7
8

1 5 2 1 5 6 2
Primary Ext. Nodes – 1, 2, 3 & 4 Primary Ext. Nodes – 1, 2, 3 & 4
Sec. Ext. Nodes – 5, 6, 7 & 8 Sec. Ext. Nodes – 5, 6, 7, 8, 9, 10, 11 & 12
Internal Node - 9 Internal Nodes – 13, 14, 15 & 16

Three Dimensional Elements (Volume Elements): When 4

the geometry, material and displacement can be expressed
by minimum of three independent space coordinate, a 3D 3
element may be chosen.
A tetrahedron element is the basic 3D element that
can be used in FEM. It has 4 primary external nodes. 3D
elements with 8 primary external nodes may be the
hexahedron element or the rectangular prism element. The 2
rectangular prism element is a special case of hexahedron 1
element. 8 Tetrahedron Element
8 7

4
5 7 4 3
6
3 6 5
1
2 1 2
Hexahedron Element
Rectangular Prism Element
Compiled by Vijay G. S., Asst. Prof., 2 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Axisymmetric Elements: Problems involving 3D axisymmetric solids or solids of

revolution, subjected to axisymmetric loading, can be reduced to simple 2D problems.
Because of the total symmetry about ‘z’ axis, all deformations and stresses are independent of
the rotational angle ‘θ’. Thus the problem needs to be looked as a 2D problem in ‘r’ & ‘z’
coordinates only, which defines the revolving area.

z z

r r

Axisymmetric element
Axisymmetric solid or
solid of revolution

Displacement Models:
In FEM after we discretize the problem continuum into finite number of elements, we
assume a displacement function for an element which closely represents the expected
displacement variation over the element domain. These simple functions which are assumed
to approximate the displacements for each element are called as “Displacement Models” or
“Displacement Functions” or “Displacement Fields” or “Displacement Patterns”.
Earlier many types of mathematical functions (like trigonometric functions, algebraic
functions, etc.) were assumed for the displacement functions. But presently mathematical
polynomials of a specific order are used to represent the displacement functions.

Reasons for using polynomials for displacement functions:

1. It is easy to handle the mathematics of polynomials in formulating the element equations

and in performing digital computations. Differentiation and integration of such a
polynomial is easy.
2. The solution can attain a particular level of approximation by selecting a particular order
of the polynomial. So a polynomial of infinite order represents an exact solution. Hence,
by truncating the polynomials at different orders, we can vary the solution approximation.

For example, suppose we want to represent the displacement ‘u’ as a function of ‘ζ’,

u (ζ ) = α1 + α 2ζ + α3ζ 2 + α 4 ζ 3 + α 5ζ 4 + ................ + α ( n +1)ζ n

Greater the number of terms used in the polynomial, closer will be the solution to the exact
solution.

Compiled by Vijay G. S., Asst. Prof., 3 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Exact Solution Exact Solution

u = f(ζ)

u = f(ζ)
u = α1 (consant) u = α1 + α2ζ
(linear)

Region of the Region of the

Element Element
ζ ζ
Fig. (a) Fig. (b)

In figure (a), to approximate the exact

solution, only one term in the polynomial has been Exact Solution

u = f(ζ)
used.
i.e., u = α1 (constant) is used.
We can see the large variation (gap) between the
two. u = α1 + α2ζ + α3ζ2
(quadratic)
In figure (b), the same exact solution is
being approximated by two term polynomial. Region of the
i.e., u = α1 + α2ζ (linear) is used. Element
We can see that the variation (gap) between the two ζ
is less compared to that in figure (1). Fig. (c)
In figure (c), the exact solution is being
approximated by a three term polynomial.
i.e., u = α1 + α2ζ + α3ζ2 (quadratic) is used.
Here, we can note that the gap between the two has further reduced.

So, it can be concluded that a displacement polynomial with larger number of terms
will approximate the solution close to the exact solution.

To represent the 1D problems (where the displacement is along the element axis), the
displacement polynomial should have only one variable in it.
If the 1D element axis is oriented along the horizontal direction (parallel to ‘x’ axis),

u (ζ ) = α1 + α 2 ζ + α3ζ 2 + α 4ζ 3 + α5ζ 4 + ................ (only one variable ‘ζ’ is used)

 α1  1
α  ζ
 2  
α 
{
u (ζ ) = 1 ζ ζ 2 ζ3 }
...........  3  ζ 2 
where, {φ} =  3 
α 4 
 .  ζ 
  .
 .   
 . 
i.e., u (ζ ) = {φ} {α}
T

Compiled by Vijay G. S., Asst. Prof., 4 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

If the 1D element axis is oriented along the vertical direction (parallel to ‘y’ axis),

v(η) = β1 + β 2η + β3η2 + β 4 η3 + β5η4 + ................ (only one variable ‘η’ is used)

β1  1
β  η
 2  
β 
{ } η 
2
v(η) = 1 η η2 η3 ...........  3  where, {θ} =  3 
β4  η 
. .
   
.  . 
i.e., v(η) = {θ} {β}
T

To represent 2D problems, where the displacement has components ‘u’ & ‘v’ along
‘x’ and ‘y’ directions respectively, the displacement polynomial should have two variables
‘ζ’ and ‘η’ in it.
i.e., u (ζ, η) = α1 + α 2 ζ + α3η + α 4 ζη + α5ζ 2 + α 6η2 + α 7 ζ 2 η2 + α8ζ 2η + α9ζη 2
& v(ζ, η) = β1 + β2 ζ + β3η + β 4ζη + β5 ζ 2 + β 6 η2 + β7 ζ 2 η2 + β8ζ 2 η + β9ζη 2

Convergence:
The Finite Element Analysis is an iterative procedure. We don’t get a solution for a
problem being analyzed by FEM just in one go. We have to perform a number of trials or
iterations and record the solution for each iteration. In each iteration we modify the elements
being used for discretization. We then plot a chart known as “Convergence Chart” which is
a plot of solution on ‘y’ axis and element parameter on the ‘x’ axis. In this chart we see that
the FEM solution curve approaches the actual solution line as we continue with iterations.
That is, the error (difference between the actual solution and FEM solution) continuously
decreases. This is known as Convergence. The FEM solution curve will finally go
asymptotic (parallel) to the actual solution line.
Convergence is defined as the monotonic approach of FEM solution to the exact
solution. Monotonic means, with out change in sign or direction.

Methods of attaining Convergence:

There are two methods by which Convergence may be attained.
1. h-method &
2. p-method
h-method of Convergence:

1st iteration - Coarse Mesh 2nd iteration - Medium Mesh 3rd - iteration Fine Mesh
No. of Elements = n1 No. of Elements = n2 No. of Elements = n3
Element size = h1 Element size = h2 Element size = h3
Compiled by Vijay G. S., Asst. Prof., 5 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

In h-method, for each iteration, the number of elements is increased or in other words,
the element size is reduced, i.e., the element mesh fineness increases with iterations. But the
order of the displacement polynomial selected for the polynomial is maintained the same for
all iterations.
In the above figure it may be noted that,
n1 < n2 < n3 & h1 > h2 > h3

Actual Solution
S3
Solution

S2
S1 S1, S2 & S3 are FEM
solutions for iterations
1, 2 & 3 respectively
FEM Solution

n1 n2 n3
Number of elements
Convergence by h-method
p-method of Convergence:
In p-method, number of elements and element size are kept constant at suitable
values. The number of nodes per element is increased with iterations. Considering more
number of nodes per element will cause the number of terms in the displacement polynomial
as well as order of the displacement polynomial to increase with iterations. (p stands for
polynomial).

1st iteration 2nd iteration 3rd iteration

No. of nodes/element = p1 No. of nodes/element = p2 No. of nodes/element = p3
In the above figure it may be noted that,
p1 < p2 < p3

Actual Solution
S3
Solution

S2
S1 S1, S2 & S3 are FEM
solutions for iterations
1, 2 & 3 respectively
FEM Solution

p1 p2 p3
Number of nodes/element
Convergence by p-method

Compiled by Vijay G. S., Asst. Prof., 6 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Convergence Requirements or Convergence Criteria to be met by

Displacement Functions:
For FEM to give a solution close to the exact solution, the displacement functions
assumed for the elements must satisfy certain requirements (conditions) called “Convergence
Requirements”.
The displacement functions should satisfy the following three conditions:

First Requirement: The assumed displacement function should be continuous with in the
elements and the displacements must be compatible between adjacent elements.
Variation of
displacement

Q4
Q3
Q1 = 0 Q2
1 1 2 2 3 3 4

q4(3)
q3(2) q3(3)
q1(1)=0 q2(1) q2(2)

1 1 2 2 2 3 3 3 4

Example: Consider a horizontal bar fixed at one end and subjected to an axial point load ‘P’
at the other end. It is required to find the displacement distribution along the length of the bar.
Let it’s FEM model be made up of 3 elements. Then there will be a total of 4 nodes. Let Q1,
Q2, Q3 & Q4 be the global nodal displacements. Since left end is fixed, Q1= 0. Let us assume
that the displacement variation is linear.
According to convergence requirements, the displacement function must be
continuous within each element and compatible between adjacent elements.

Meaning of Continuity:
Displacement is Displacement is
continuous dis-continuous

qj(e) qj(e)
qi(e) qi(e)
i e j i j
e

Compiled by Vijay G. S., Asst. Prof., 7 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

In the above figure we can understand the difference between continuous and
dis-continuous displacement. Dis-continuous displacement occurs if a crack or defect in the
bar is present. Such discrepancies should be properly specified as boundary conditions.

Meaning of Compatibility:
Node 2 is common for element (1) and element (2).
If displacement of node 2 for element (1) = displacement of node 2 for element (2)
i.e., If, q2(1) = q2(2) = Q2, displacement functions of elements (1) & (2) are compatible.

Similarly, if q3(2) = q3(3) = Q3, displacement functions of elements (2) & (3) are compatible.

Second Requirement: The assumed displacement functions should be capable of

representing rigid body displacement of the element. The rigid body displacement is the most
elementary displacement that an element may undergo. i.e., during rigid body displacement
the element should not be strained.
Usually, the first term of the assumed displacement polynomial will be a constant
which represents the rigid body displacement.
For example, for a 2D element, the assumed displacement polynomial may be
u (ζ, η) = α1 + α 2 ζ + α 3η + α 4 ζη + α 5ζ 2 + α 6 η2 + α 7 ζ 2 η2 + α 8 ζ 2 η + α 9 ζη 2

First term
The first term α1 represents rigid body displacement.

Third Requirement: The assumed displacement functions should include constant strain
rates of the element.
Usually, the second term of the assumed displacement polynomial function will represent
the constant strain rate.

For example, for a 1D 2 noded element,

u = α1 + α 2ζ
1 2
2nd term (1st order term)
du
∴ε = = α 2 (cons tan t )
dζ
3
Similarly, for a 3 noded triangular element,
u = α1 + α 2 ζ + α 3η
du du
εζ = = α2 (cons tan t ) & εη = = α3 (cons tan t)
dζ dη
1 2
Note: The elements whose displacement functions satisfy the First Convergence requirement
are said to be Compatible or Confirming Elements.
The elements whose displacement functions satisfy the Second and Third
Convergence requirements are said to be Complete Elements.

Compiled by Vijay G. S., Asst. Prof., 8 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

How to select the order of the assumed displacement polynomial?

The convergence requirements that
• Rigid body displacement should be represented, and
• Constant strain rates must be represented
upto some extent decide the order of the displacement polynomial.

An additional consideration in the selection of the polynomial order is that the

selected pattern (polynomial) should be independent of the orientation of the local coordinate
system in the global coordinate system. This property of the displacement function is called
“Geometric Isotropy” or “Spatial Isotropy” or “Geometric Invariance”.

For polynomials of linear order, the “Geometric Isotropy” condition is inherently

achieved by including the constant strain rate. i.e., u = α1 + α 2 ζ

For polynomials of higher orders, only those higher order coordinates are considered
which satisfy Geometric Isotropy.
The best method to select the higher order coordinates which represent Geometric
Isotropy is to make use of the PASCAL’S TRIANGLE of coordinates. Pascal’s triangle has
symmetric coordinates which take care of Geometric Isotropy.

The Pascal’s Triangle:

1 Constant terms

ζ η Linear terms

ζ2 ζη η2 Quadratic terms

ζ3 ζ2η ζη2 η3 Cubic terms

ζ4 ζ3η ζ2η2 ζη3 η4 Quartic terms

ζ5 ζ4η ζ3η2 ζ2η3 ζη4 η5 Quintic terms

ζ6 ζ5η ζ4η2 ζ3η3 ζ2η4 ζη5 η6

ζ7 ζ6η ζ5η2 ζ4η3 ζ3η4 ζ2η5 ζη6 η7

Axis of symmetry
Using the Pascal’s Triangle, suppose we want to write the displacement polynomial
for a 8 noded quadrilateral element, keeping symmetry in mind, we can include the following
terms.

Compiled by Vijay G. S., Asst. Prof., 9 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Constant terms 1 One term

Linear terms ζ, η Two terms
Quadratic terms ζ2, ζη, η2 Three terms
Cubic terms ζ2η, ζη2 Two terms
or ζ3, η3
Total: Eight terms

∴ u (ζ, η) = α1 + α 2ζ + α3η + α 4ζ 2 + α 5ζη + α 6η2 + α7ζ 2η + α8ζη 2

OR
u (ζ, η) = α1 + α 2ζ + α3η + α 4ζ 2 + α 5ζη + α6η2 + α7 ζ 3 + α8η3

It can be noted that since there are 8 nodes, the displacement polynomial has 8 terms.

Higher Order Elements:

The minimum number of DOFs (number of terms in the displacement polynomial) for
a given element is determined by
• The completeness requirements of convergence
• The requirement of geometrical isotropy
• The necessity of adequate representation of terms in the potential energy functional

By providing additional DOFs, the FEM solution accuracy (nearness to exact

solution) can be improved. Additional DOFs beyond the minimum number may be included
for an element by adding secondary external nodes. So, the elements with additional DOFs
are called “Higher Order Elements”
If the displacement polynomials have terms of the order higher than ONE, such
elements represent higher order elements. i.e., if the terms included have ζ2, ζ2η, ζη2, η2 and
higher, they are called higher order elements.

Following is a table which gives elements which are the Basic elements (not higher order
elements).

Terms of the Pascal’s

Element Displacement Polynomial
triangle used
2 noded 1D element
u = α1+α2ζ 1, ζ
(Linear Element)
3 noded 2D triangular element
u = α1+α2ζ+α3η 1, ζ, η
(CST Element)
4 noded 2D quadrilateral element u = α1+α2ζ+α3η+α4ζη 1, ζ, η, ζη

Note: Displacement polynomials in the table do not have ζ2, η2, ζη2, ζ2η, etc. terms. In the
table the elements are having most minimum number of nodes to represent the basic shape of
the element. Hence, they are known as ‘Basic Elements’.

Higher Order Quadrilateral Elements:

These can be classified as two types
(1) Lagrange Elements & (2) Serendipity Elements

Compiled by Vijay G. S., Asst. Prof., 10 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Lagrange Elements: Lagrange quadrilateral elements have the additional nodes on the
element edges as well as within the element.
Figure (1) shows the Pascal’s Triangle which highlights the method of choosing
higher order terms to be included and number of terms or nodes in the displacement
polynomial for Lagrange element.
Other Lagrange Family higher order elements can similarly be defined.
1

ζ η
4 noded
ζ2 ζη η2

ζ3 ζ2η ζη2 η3 9 noded

ζ4 ζ3η ζ2η2 ζη3 η4 16 noded

ζ5 ζ4η ζ3η2 ζ2η3 ζη4 η5

25 noded
ζ6 ζ5η ζ4η2 ζ3η3 ζ2η4 ζη5 η6

ζ7 ζ6η ζ5η2 ζ4η3 ζ3η4 ζ2η5 ζη6 η7

ζ8 ζ7η ζ6η2 ζ5η3 ζ4η4 ζ3η5 ζ2η6 ζη7 η8

Fig. (1) Use of Pascal’s triangle for choosing terms for

displacement polynomial for Lagrange Elements
Serendipity Elements: Serendipity quadrilateral elements have additional nodes only on the
element edges and nodes within the element are absent.
Figure (2) shows the Pascal’s Triangle which highlights the method of choosing
higher order terms to be included and number of terms or nodes in the displacement
polynomial for Serendipity elements.
Other Serendipity Family higher order elements can similarly be defined.
1

ζ η
4 noded
ζ2 ζη η2

ζ3 ζ2η ζη2 η3 8 noded

ζ4 ζ3η ζ2η2 ζη3 η4 12 noded

ζ5 ζ4η ζ3η2 ζ2η3 ζη4 η5

16 noded
ζ6 ζ5η ζ4η2 ζ3η3 ζ2η4 ζη5 η6

Fig. (2) Use of Pascal’s triangle for choosing terms for

displacement polynomial for Serendipity Elements
Compiled by Vijay G. S., Asst. Prof., 11 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Higher Order Triangular Elements:

ζ η 3 noded

ζ2 ζη η2 6 noded

ζ3 ζ2η ζη2 η3
10 noded
ζ 4
ζη
3
ζη
2 2
ζη 3
η4

ζ5 ζ4η ζ3η2 ζ2η3 ζη4 η5 15 noded

Fig. (3) Use of Pascal’s triangle for choosing terms for displacement
polynomial for Higher Order Triangular Elements
Higher Order Triangular Elements have the additional nodes on the element edges as well as
within the element.
Figure (3) shows the Pascal’s Triangle which highlights the method of choosing higher order
terms to be included and number of terms or nodes in the displacement polynomial for
Higher Order Triangular Elements.

Shape functions or Interpolation functions:

The Shape Functions (or Interpolation Functions) are mathematical functions which relate the
field variable at any point inside the element to the field variable values at the nodes. In case
of structural problems, the shape functions relate the displacement at any point inside the
element to the nodal displacements. These functions are also used to relate the geometrical
coordinates any point inside the element to the nodal coordinates.

If q1, q2, q3, ………qn are the nodal displacements of the n nodes of an element along a
particular direction, then the displacement ‘q’ at any point inside the element can be obtained
using the Shape functions as follows.
q = N1q1 + N 2 q2 + N 3q3 + N 4 q4 + ............ + N n qn
n
i.e., q = ∑ N i qi
i =1
Similarly, If x1, x2, x3, ………xn are the nodal coordinates of the n nodes of an element, then
the coordinate ‘x’ at any point inside the element can be obtained using the Shape functions
as follows.
x = N1 x1 + N 2 x2 + N 3 x3 + N 4 x4 + ............ + N n xn
n
i.e., x = ∑ N i xi
i =1
Also, a property of the shape function that its value is unity at the node at which it is defined
and zero at all other nodes of the element. The summation of the shape functions of all the
nodes of an element is unity.

Compiled by Vijay G. S., Asst. Prof., 12 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

N1 + N 2 + N 3 + N 4 + ............ + N n = 1
n
i.e., ∑ N i = 1
i =1

Derivation of Shape Function for a 1D linear (2 noded) element:

Horizontal Element:
Consider a 1D linear element with two nodes q1 q2
1 & 2 oriented parallel to the x axis. Its local
1 2 ζ
coordinates in terms of ζ are shown. The
length of the element is 2 units. It is having 1 1
the local coordinate origin located at its
center. Let q1 and q2 be the displacements at ζ = − 1 ζ=0 ζ=1
the nodes 1 and 2 respectively.

Let the displacement ‘u’ at any point within the element be represented as

u = α1 + α 2ζ − − − − −(1)
α  1 
= {1 ζ }  1  where, {P} =  
α 2  ζ 
= {P}T {α } − − − − (2)
Note: Equation (1) contains two terms since the element is having two nodes.

At node 1, ζ = − 1and u = q1 , substituting in (1) we get,

q1 = α1 − α2 --------(3)

At node 2, ζ = + 1and u = q2 , substituting in (1) we get,

q2 = α1 + α2 --------(4)

Writing equations (3) & (4) in matrix form,

 q1  1 −1 α1 
 =  
q2  1 1  α 2 
i.e., {q} = [C ]{α }
∴ {α } = [C ]−1{q} − − − − (5)

Substituting (5) in (2),

u = {P}T [C ]−1{q}
 1 1 
2   q1 
u = {1 ζ }  2  
− 1 1  q2 
 2 2

Compiled by Vijay G. S., Asst. Prof., 13 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

 1 ζ   1 ζ    q 
i.e., u =  −   +    1 
 2 2   2 2   q2 
 1 − ζ   1 + ζ    q1 
=      
 2   2   q2 
q 
= { N1 N2 }  1 
q2 
 1−ζ   1+ ζ 
where, N1 =   & N2 =  
 2   2 
N1 & N2 are the Shape Functions at nodes 1 and 2 respectively.
It may be noted that the expressions for N1 & N2 are linear in ζ. So if we plot them, they
should be straight lines.
Node 1 Node 2
ζ=−1 ζ=+1
N1 1 0
N2 0 1

N1  1−ζ  N2
N1 =   1+ζ 
 2  N2 =  
 2 
1 1

2 1
1 2
1 1 1 1
ζ=−1 ζ=0 ζ=1 ζ=−1 ζ=0 ζ=1

Vertical Element:
Consider a 1D linear element with two nodes 1 & 2 oriented η q2
parallel to the y axis. Its local coordinates in terms of η are shown.
2
The length of the element is 2 units. It is having the local coordinate η=1
origin located at its center. Let q1 and q2 be the displacements at the
nodes 1 and 2 respectively.
1
Let the displacement ‘v’ at any point within the element be
represented as
η=0
v = β1 + β 2η − − − − −(1)
β  1 
= {1 η }  1  where, {P} =   1
 β2  η 
q1
= {P}T {β } − − − − (2)
η=−1
1
Note: Equation (1) contains two terms since the element is having
two nodes.

Compiled by Vijay G. S., Asst. Prof., 14 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

At node 1, η = − 1and v = q1 , substituting in (1) we get,

q1 = β1 − β2 --------(3)

At node 2, η = + 1and v = q2 , substituting in (1) we get,

q2 = β1 + β2 --------(4)

Writing equations (3) & (4) in matrix form,

 q1  1 −1  β1 
 =  
q2  1 1   β 2 
i.e., {q} = [C ]{β }
∴ {β } = [C ]−1{q} − − − − (5)

Substituting (5) in (2),

v = {P}T [C ]−1{q}
 1 1 
2   q1  =  1 − η   1 + η    q1 
v = {1 η} 
2
    
− 1 1  q2   2 2   2 2   q2 
 2 2
 1 − η   1 + η    q1 
i.e., v =      
  2   2    q2 
q 
= { N1
N2 }  1 
q2 
 1−η   1 +η 
where, N1 =   & N2 =  
 2   2 
N1 & N2 are the Shape Functions at nodes 1 and 2 respectively.
It may be noted that the expressions for N1 & N2 are linear in η. So if we plot them, they
should be straight lines.
η Node 1 Node 2
η
η=−1 η=+1
2 1
N1 1 0 2
N2 0 1

1  1 −η 
N1 =   1
 2 

N1 N2

 1 +η 
1 1 N2 =  
 2 

1 1
1

Compiled by Vijay G. S., Asst. Prof., 15 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Derivation of Shape Function for a 1D quadratic (3 noded) element:

Horizontal Element:
Consider a 1D linear element with three nodes q1 q2
q3
1, 2 & 3 oriented parallel to the x axis. Its
local coordinates in terms of ζ are shown. The 1 3 2 ζ
length of the element is 2 units. It is having
the local coordinate origin located at its 1 1
center. The third node is located at the origin. ζ = − 1 ζ=0 ζ=1
Let q1, q2 and q3 be the displacements at the
nodes 1, 2 and 3 respectively.

Let the displacement ‘u’ at any point within the element be represented as

u = α1 + α 2ζ + α 3ζ 2 − − − − −(1)
1
α 1   
  where, {P} =  ζ 
{
= 1 ζ ζ2 } α 2  ζ 2 
α   
 3
= {P}T {α } − − − − (2)

Note: Equation (1) contains three terms since the element is having three nodes.

At node 1, ζ = − 1 and u = q1 , substituting in (1) we get,

q1 = α1 + α2(−1) + α3(−1)2 = α1 − α2 + α3 --------(3)

At node 2, ζ = + 1 and u = q2 , substituting in (1) we get,

q2 = α1 + α2(1) + α3(1)2 = α1 + α2+ α3 --------(4)

At node 3, ζ = 0 and u = q3 , substituting in (1) we get,

q3 = α1 + α2(0) + α3(0)2 = α1 --------(5)

Writing equations (3), (4) & (5) in matrix form,

 q1  1 −1 1  α1 
    
q2  = 1 1 1  α 2 
 q  1 0 0  α 
 3   3
{q} = [C ]{α }
OR {α } = [C ]−1{q} − − − − (6)

Substituting (6) in (2),

u = {P}T [C ]−1{q}
 0 0 1   q1 
 1   
{
u= 1 ζ ζ2 }  2   −1 1 0  q2 
 
 1 1 −2   q3 

Compiled by Vijay G. S., Asst. Prof., 16 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

 q1 
1 1 1   
i.e., u =  (0 − ζ + ζ 2 ) (0 + ζ + ζ 2 ) (2 + 0 − 2ζ 2 )  q2 
2 2 2  
 q3 
 q1 
 −ζ (1 − ζ ) ζ (1 + ζ )   
= 1 − ζ 2   q2 
 2 2  
 q3 
 q1 
 
= { N1 N2 N 3 }  q2 
q 
 3
−ζ (1 − ζ ) ζ (1 + ζ )
where, N1 = , N2 = & N3 = 1 − ζ 2
2 2
N1, N2 & N3 are the Shape Functions at nodes 1, 2 and 3 respectively.
It may be noted that the expressions for N1, N2 & N3 are quadratic in ζ. So if we plot them,
they should be curvilinear lines.

Node 1 Node 2 Node 3

ζ=−1 ζ=+1 ζ=0
N1 1 0 0
N2 0 1 0
N3 0 0 1
−ζ (1 − ζ )
N1 = ζ (1 + ζ )
2 N2 =
2
N1 N2

1 1

2 1
1 3 3 2
ζ=−1 ζ=0 ζ=1 ζ=−1 ζ=0 ζ=1

N3

N3 = 1 − ζ 2

1 3 2
ζ=−1 ζ=0 ζ=1

Compiled by Vijay G. S., Asst. Prof., 17 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Vertical Element:
In a similar way the shape functions for a 1D quadratic (3 noded) element oriented
along the y axis can be derived as follows. η q 2
2
η=1
−η (1 − η )
N1 = ,
2 1
η (1 + η )
N2 = & q3
2 3 η=0
N3 = 1 −η 2

1
q1
η=−1
1

Derivation of Shape Functions for a 2D linear (3 noded) Triangular

Element: (Constant Strain Triangular Element)
q6
Consider a three noded triangular element. Each
q5
node is having two degrees of freedom. The (ζ3, η3) 3
nodal displacements of nodes 1, 2 & 3 in the ‘ζ’
direction are q1, q3 & q5 and in ‘η’ direction they q4
are q2, q4 & q6 respectively. v
u q3
2
Let P(ζ, η) be any point inside the CST element. η P(ζ, η)
Then, the displacement at P(ζ, η) can be assumed (ζ2, η2)
q2
to be a linear function of ‘ζ’ & ‘η’ where ‘ζ’ &
1
‘η’ are local coordinates parallel to x and y axes
(ζ1, η1) q1
respectively.
ζ
u = α1 + α 2ζ + α 3η − − − − −(1)
α 1  1 
   
= {1 ζ η } α 2  where, {P} = ζ 
α  η 
 3  
= {P} {α } − − − − (2)
T

At node 1, ζ = ζ1, η = η1 and u = q1 , substituting in (1) we get,

q1 = α1 + α2ζ1 + α3η1 --------(3)

At node 2, ζ = ζ2, η = η2 and u = q3 , substituting in (1) we get,

q3 = α1 + α2ζ2 + α3η2 --------(4)

At node 3, ζ = ζ3, η = η3 and u = q5 , substituting in (1) we get,

q5 = α1 + α2ζ3 + α3η3 --------(5)

Compiled by Vijay G. S., Asst. Prof., 18 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Writing equations (3), (4) & (5) in matrix form,

 q1  1 ζ 1 η1  α1 
    
q3  = 1 ζ 2 η 2  α 2 
q  1 ζ η  α 
 5  3 3 3

{q} = [C ]{α }
OR {α } = [C ]−1{q} − − − − (6)

−1
1 ζ 1 η1  ζ 2η3 − ζ 3η 2 ζ 3η1 − ζ 1η3 ζ 1η 2 − ζ 2η1 
−1 
[C ] = 1 ζ 2 η 2  =  1 
η 2 − η3 η3 − η1 η1 − η2  − − − −(7)
C 
1 ζ 3 η3   ζ 3 − ζ 2 ζ1 − ζ 3 ζ 2 − ζ 1 
C = 1( ζ 2η3 − ζ 3η2 ) − ζ 1 (η3 − η2 ) + η1 (ζ 3 − ζ 2 )

Let a1 = ζ 2η3 − ζ 3η 2 a2 = ζ 3η1 − ζ 1η3 a3 = ζ 1η 2 − ζ 2η1

b1 = η 2 − η3 b2 = η3 − η1 b3 = η1 − η2
c1 = ζ 3 − ζ 2 c2 = ζ 1 − ζ 3 c3 = ζ 2 − ζ 1

∴ C = a1 + b1ζ 1 + c1η1 − − − − (8)

Substituting (6), (7) & (8) in (2),

 a1 a2 a3   q1 
 
η}  b1 b3  q3 
1
u= {1 ζ b2
(a1 + b1ζ 1 + c1η1 )
 c1 c2 c3  q5 
 q1 
 a + b ζ + c1η a2 + b2ζ + c2η a3 + b3ζ + c3η   
= 1 1   q3  − − − − (9)
 a1 + b1ζ 1 + c1η1 a1 + b1ζ 1 + c1η1 a1 + b1ζ 1 + c1η1   
q5 

It is known from analytical geometry that if (ζ1, η1), (ζ2, η2) and (ζ3, η3) are the coordinates
of the corners of a triangle, then the area A of the triangle is given as:
1 ζ 1 η1
1 1
A = 1 ζ 2 η 2 = (a1 + b1ζ 1 + c1η1 )
2 2
1 ζ 3 η3
1
∴A = C OR 2 A = C = a1 + b1ζ 1 + c1η1
2 (ζ3, η3) 3

It must be noted that first row of the above matrix [C] A1

represents the first corner of the triangle. A2 P(ζ, η) 2
If we consider any arbitrary point P(ζ, η) inside the
A3 (ζ2, η2)
triangular element and join P to the nodes 1, 2 & 3, three
1
triangles are formed. (ζ1, η1)

Compiled by Vijay G. S., Asst. Prof., 19 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

They are:
∆ P23 opposite to node 1 whose area is A1.
∆ P13 opposite to node 2 whose area is A2.
∆ P12 opposite to node 3 whose area is A3.

1 ζ η
2 A1 = 1 ζ 2 η 2 = 1(ζ 2η3 − ζ 3η2 ) − ζ (η3 − η 2 ) + η (ζ 3 − ζ 2 ) = (a1 + b1ζ + c1η )
1 ζ 3 η3

1 ζ η
2 A2 = 1 ζ 1 η1 = 1(ζ 1η3 − ζ 3η1 ) − ζ (η3 − η1 ) + η (ζ 3 − ζ 1 ) = (a2 + b2ζ + c2η )
1 ζ 3 η3

1 ζ η
2 A3 = 1 ζ 1 η1 = 1(ζ 1η2 − ζ 2η1 ) − ζ (η 2 − η1 ) + η (ζ 2 − ζ 1 ) = (a3 + b3ζ + c3η )
1 ζ 2 η2

1 ζ 1 η1
2 A = 1 ζ 2 η2 = 1(ζ 2η3 − ζ 3η2 ) − ζ 1 (η3 − η1 ) + η1 (ζ 3 − ζ 2 ) = (a1 + b1ζ 1 + c1η1 )
1 ζ 3 η3

Substituting these in (9),

 q1 
 2A 2 A2 2 A3   
u= 1  q3 
 2A 2A 2A  
q5 
 q1 
 
= { N1 N2 N 3 } q3 
q 
 5
∴ u = {N }{q}

Similarly, it can be derived that,

q2 
2A 2 A2 2 A3   
v= 1  q4 
 2A 2A 2A   
q6 
q2 
 
= { N1 N2 N 3 } q4 
q 
 6
∴ v = {N }{q}

Compiled by Vijay G. S., Asst. Prof., 20 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

A1 a1 + b1ζ + c1η
where, N1 = = ,
A a1 + b1ζ 1 + c1η1
A2 a2 + b2ζ + c2η
N2 = = &
A a1 + b1ζ 1 + c1η1
A3 a3 + b3ζ + c3η
N3 = =
A a1 + b1ζ 1 + c1η1
η
are the Shape functions of the triangular element.

A standard (master) CST element is shown in figure. 2 (0,1)

(ζ1, η1) = (1,0)

(ζ2, η2) = (0,1) 1
(ζ3, η3) = (0,0)
ζ
3 (0,0) 1 (1,0)
a1 = 0, a2 = 0, a3 = 1 1
b1 = 1, b2 = 0, b3 = −1
c1 = 0, c2 = 1, c3 = −1

A = 0 + (1)(1) + 0 = 1
A1 = 0 + (1) ζ +0 = ζ
A2 = 0 + 0 +(1)η = η
A3 = 1 − ζ − η

N1 = ζ, N2 = η, N3 = 1 − ζ − η are the Shape Functions for a master

CST element.

Derivation of Shape Function for a 4 noded Quadrilateral Element:

Consider a master quadrilateral element with 4 4 3
nodes. It is a perfect square with its geometric center at
origin of ζ, η space and sides of the square being 2 units 1
long so that η
ζ
At node 1, (ζ1, η1) ≡ (−1, −1),
At node 2, (ζ2, η2) ≡ (1, −1),
At node 3, (ζ3, η3) ≡ (1, 1) & 1
At node 4, (ζ4, η4) ≡ (−1, 1)
1 2
Each node is having two degrees of freedom. The nodal 1 1
displacements of nodes 1, 2, 3 & 4 in the ‘ζ’ direction
are q1, q3, q5 & q7 and in ‘η’ direction they are q2, q4, q6
& q8 respectively.

Let P(ζ, η) be any point inside the quadrilateral element. Then, the displacement at P(ζ, η)
can be assumed to be a function of ‘ζ’ & ‘η’ where ‘ζ’ & ‘η’ are local coordinates parallel to
x and y axes respectively.

Compiled by Vijay G. S., Asst. Prof., 21 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

From the Pascal’s triangle, we can write the displacement functions of any point within the
quadrilateral element.

u = α1 + α 2ζ + α 3η + α 4ζη − − − − −(1)
α 1  1
α  ζ 
   
= {1 ζ η ζη }  2  where, {P} =  
α 3  η 
α 4  ζη 

= {P}T {α } − − − − (2)

At node 1, ζ = −1, η = −1 and u = q1 , substituting in (1) we get,

q1 = α1 + α2(−1)+ α3(−1) + α4(−1)(−1) = α1 − α2 − α3 + α4 --------(3)

At node 2, ζ = +1, η = −1 and u = q3 , substituting in (1) we get,

q3 = α1 + α2(1)+ α3(−1) + α4(1)(−1) = α1 + α2 − α3 − α4 --------(4)

At node 3, ζ = +1, η = +1 and u = q5 , substituting in (1) we get,

q5 = α1 + α2(1)+ α3(1) + α4(1)(1) = α1 + α2 + α3 + α4 --------(5)

At node 4, ζ = −1, η = +1 and u = q7 , substituting in (1) we get,

q7 = α1 + α2(−1)+ α3(1) + α4(−1)(1) = α1 − α2 + α3 − α4 --------(6)

Writing equations (3), (4), (5) & (6) in matrix form,

 q1  1 −1 −1 1  α1 
 q  1 1 −1 −1 α 
 3    2 
 =
q5  1 1 1 1  α 3 
q7  1 −1 1 −1 α 4 
{q} = [C ]{α }
OR {α } = [C ]−1{q} − − − − (7)

Substituting (7) in (2),

u = {P}T [C ]−1{q}
1 1 1 1   q1 

 1   −1 1 1 −1  q3 
= {1 ζ η ζη}    
 4   −1 −1 1 1   q5 
 
 1 −1 1 −1 q7 
 q1 
q 
1  
u = {(1 − ζ − η + ζη ) (1 + ζ − η − ζη ) (1 + ζ + η + ζη ) (1 − ζ + η − ζη )}  3 
4  q5 
q7 

Compiled by Vijay G. S., Asst. Prof., 22 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

 q1 
 
 (1 − ζ )(1 − η ) (1 + ζ )(1 − η ) (1 + ζ )(1 + η ) (1 − ζ )(1 + η )   q3 
i.e., u =   
 4 4 4 4   q5 
q7 
 q1 
q 
 3
∴ u = { N1 N2 N3 N4 }   − − − − (8)
q5 
q7 
Similarly, it can be derived that,
q2 
q 
 
v = { N1 N2 N3 N4 }  4 
 q6 
 q8 

where,
(1 − ζ )(1 − η ) (1 + ζ )(1 − η )
N1 = N2 =
4 4
(1 + ζ )(1 + η ) (1 − ζ )(1 + η )
N3 = & N4 =
4 4
are the shape functions at 4 nodes of the quadrilateral element.

At Node 1 At Node 2 At Node 3 At Node 4

(-1, -1) (1, -1) (1, 1) (-1, 1)
N1 1 0 0 0
N2 0 1 0 0
N3 0 0 1 0
N4 0 0 0 1

Finding Shape Functions using Lagrange Polynomial:

If only continuity of the displacement is to be satisfied, Lagrange Polynomials can be used to

derive the Shape Functions. Lagrange Polynomial in one dimension is defined as:

Lk = ∏
n
( x − xm )
m =1 ( xk − xm )
m≠ k

i.e., Lk =
( x − x1 )( x − x2 ) ...... ( x − xk −1 )( x − xk +1 ) ....... ( x − xn )
( xk − x1 )( xk − x2 ) ...... ( xk − xk −1 )( xk − xk +1 ) .......( xk − xn )

Compiled by Vijay G. S., Asst. Prof., 23 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Problem 1: Derive the Shape Functions for a 1D element with 5 nodes.

Solution:
1 3 5 4 2
x1 x3 x5 x4 x2

N1 = L1 =
( x − x2 )( x − x3 )( x − x4 )( x − x5 )
( x1 − x2 )( x1 − x3 )( x1 − x4 )( x1 − x5 )
N 2 = L2 =
( x − x1 )( x − x3 )( x − x4 )( x − x5 )
( x2 − x1 )( x2 − x3 )( x2 − x4 )( x2 − x5 )
N 3 = L3 =
( x − x1 )( x − x2 )( x − x4 )( x − x5 )
( x3 − x1 )( x3 − x2 )( x3 − x4 )( x3 − x5 )
N 4 = L4 =
( x − x1 )( x − x2 )( x − x3 )( x − x5 )
( x4 − x1 )( x4 − x2 )( x4 − x3 )( x4 − x5 )
N 5 = L5 =
( x − x1 )( x − x2 )( x − x3 )( x − x4 )
( x5 − x1 )( x5 − x2 ) ( x5 − x3 )( x5 − x4 )
Problem 2: Derive the shape functions for a 9 noded quadrilateral element (9 noded
Lagrange Element)
Solution:
The 9 noded Lagrange Element has 4 corner nodes (1, 2, 3 & 4), 4 mid-side nodes (5,
6, 7 & 8) and one internal node (node 9) at element center.
The shape functions for such an element can be derived starting from the polynomial
displacement function as done for 1D 2 noded element, CST element & 4 noded quadrilateral
element. Also, the shape functions can be derived by superimposing the shape functions of
horizontal and vertical 1D quadratic elements into 2D space. This can be done as follows:

N1ζ N3ζ N2ζ

The shape functions for
1 1 the horizontal element are:
ζ (1 − ζ )
4 7 3 N1ζ = −
N2η 2
ζ (1 + ζ )
N 2ζ =
2
1 N 3ζ = 1 − ζ 2
η
8 ζ The shape functions for
N3η 6
9 the vertical element are:
η (1 − η )
N1η = −
1 2
η (1 + η )
N 2η =
2
N1η N 3η = 1 − η 2
1 5 2

Compiled by Vijay G. S., Asst. Prof., 24 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

The shape functions for the nine nodes of the quadrilateral element is given by:

 ζ (1 − ζ )  η (1 − η ) 
N1 = N1ζ × N1η =  −  − 
 2  2 
 ζ (1 + ζ )  η (1 − η ) 
N 2 = N 2ζ × N1η =   − 
 2  2 
 ζ (1 + ζ )   η (1 + η ) 
N 3 = N 2ζ × N 2η =   2 
 2  
 ζ (1 − ζ )   η (1 + η ) 
N 4 = N1ζ × N 2η =  −  
 2  2 
 η (1 − η ) 
N 5 = N 3ζ × N1η = (1 − ζ 2 )  − 
 2 
 ζ (1 + ζ ) 
N 6 = N 2ζ × N 3η =   (1 − η )
2

 2 
 η (1 + η ) 
N 7 = N 3ζ × N 2η = (1 − ζ 2 )  
 2 
 ζ (1 − ζ ) 
N 8 = N1ζ × N 3η =  −
 2 

(1 −η 2 )

N 9 = N 3ζ × N 3η = (1 − ζ 2 )(1 − η 2 )

Problem 3: Derive the shape functions for a 1D cubic (4 noded) horizontal element using the
Lagrange polynomial. Using these shape functions derive the shape functions for a 16 noded
lagrange quadrilateral element.

Solution: We first derive the shape functions for 1D 4 noded (cubic) horizontal element.

Consider a standard (master) 1D cubic element 1 1

of length 2 units. The origin of the ζ axis is at
the element center. The nodal coordinates are: 1 3 4 2 ζ
1 1
3 3
At node 1, ζ = -1 2 2 2
At node 2, ζ = +1 3 3 3

At node 3, ζ = -1/3
At node 4, ζ = +1/3

Applying Lagrange polynomial function to each node,

(ζ − ζ 2 )(ζ − ζ 3 )(ζ − ζ 4 ) (ζ − 1)(ζ + 13 )(ζ − 13 ) 9

N1ζ = L1 = = = − (ζ − 1)(ζ 2 − 19 )
(ζ 1 − ζ 2 )(ζ 1 − ζ 3 )(ζ 1 − ζ 4 ) (−1 − 1)(−1 + 3 )( −1 − 3 )
1 1
16
(ζ − ζ 1 )(ζ − ζ 3 )(ζ − ζ 4 ) (ζ + 1)(ζ + 13 )(ζ − 13 ) 9
N 2ζ = L2 = = = (ζ + 1)(ζ 2 − 19 )
(ζ 2 − ζ 1 )(ζ 2 − ζ 3 )(ζ 2 − ζ 4 ) (1 + 1)(1 + 3 )(1 − 3 )
1 1
16
(ζ − ζ 1 )(ζ − ζ 2 )(ζ − ζ 4 ) (ζ + 1)(ζ − 1)(ζ − 3 ) 1
27 2
N 3ζ = L3 = = 1 = (ζ − 1)(ζ − 13 )
(ζ 3 − ζ 1 )(ζ 3 − ζ 2 )(ζ 3 − ζ 4 ) ( − 3 + 1)( − 3 − 1)( − 3 − 3 ) 16
1 1 1

Compiled by Vijay G. S., Asst. Prof., 25 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

(ζ − ζ 1 )(ζ − ζ 2 )(ζ − ζ 3 ) (ζ + 1)(ζ − 1)(ζ + 13 ) 27 2

N 4ζ = L4 = = 1 = − (ζ − 1)(ζ + 13 )
(ζ 4 − ζ 1 )(ζ 4 − ζ 2 )(ζ 4 − ζ 3 ) ( 3 + 1)( 3 − 1)( 3 + 3 )
1 1 1
16
η
Similarly, we can derive the shape functions for 1D 4 noded (cubic)
vertical element. 2
2
3
9
N1η = − (η − 1)(η 2 − 19 ) 1
16 4
9 1
N 2η = (η + 1)(η 2 − 19 ) 3
2
16 3
1
27 2
N 3η = (η − 1)(η − 13 ) 3
16 3
27 1
N 4η = − (η 2 − 1)(η + 13 ) 2
16 3

To obtain the shape functions of the 16 noded Lagrange element, we superimpose the shape
functions from the horizontal element & vertical element into 2D space as follows.
N1ζ N3ζ N4ζ N2ζ

1 1

N2η 4 10 9 3

1
11 16 15
N4η η 8
ζ

N3η 7
1 12 13 14

N1η
1 5 6 2
The shape functions for the sixteen nodes of the quadrilateral element is given as:

 9   9 
N1 = N1ζ × N1η =  − (ζ − 1)(ζ 2 − 19 )  ×  − (η − 1)(η 2 − 19 ) 
 16   16 
9   9 
N 2 = N 2ζ × N1η =  (ζ + 1)(ζ 2 − 19 )  ×  − (η − 1)(η 2 − 19 ) 
16   16 
9  9 
N 3 = N 2ζ × N 2η =  (ζ + 1)(ζ 2 − 19 )  ×  (η + 1)(η 2 − 19 ) 
16  16 
 9  9 
N 4 = N1ζ × N 2η =  − (ζ − 1)(ζ 2 − 19 )  ×  (η + 1)(η 2 − 19 ) 
 16  16 

Compiled by Vijay G. S., Asst. Prof., 26 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

 27   9 
N 5 = N 3ζ × N1η =  (ζ 2 − 1)(ζ − 13 )  ×  − (η − 1)(η 2 − 91 ) 
 16   16 
 27   9 
N 6 = N 4ζ × N1η =  − (ζ 2 − 1)(ζ + 13 )  ×  − (η − 1)(η 2 − 19 ) 
 16   16 
9   27 
N 7 = N 2ζ × N 3η =  (ζ + 1)(ζ 2 − 19 )  ×  (η 2 − 1)(η − 13 ) 
16   16 
9   27 
N 8 = N 2ζ × N 4η =  (ζ + 1)(ζ 2 − 19 )  ×  − (η 2 − 1)(η + 13 ) 
16   16 
 27  9 
N 9 = N 4ζ × N 2η =  − (ζ 2 − 1)(ζ + 13 )  ×  (η + 1)(η 2 − 19 ) 
 16  16 
 27  9 
N10 = N 3ζ × N 2η =  (ζ 2 − 1)(ζ − 13 )  ×  (η + 1)(η 2 − 19 ) 
 16  16 
 9   27 
N11 = N1ζ × N 4η =  − (ζ − 1)(ζ 2 − 19 )  ×  − (η 2 − 1)(η + 13 ) 
 16   16 
 9   27 
N12 = N1ζ × N 3η =  − (ζ − 1)(ζ 2 − 19 )  ×  (η 2 − 1)(η − 13 ) 
 16   16 
 27   27 
N13 = N 3ζ × N 3η =  (ζ 2 − 1)(ζ − 13 )  ×  (η 2 − 1)(η − 13 ) 
 16   16 
 27   27 
N14 = N 4ζ × N 3η =  − (ζ 2 − 1)(ζ + 13 )  ×  (η 2 − 1)(η − 13 ) 
 16   16 
 27   27 
N15 = N 4ζ × N 4η =  − (ζ 2 − 1)(ζ + 13 )  ×  − (η 2 − 1)(η + 13 ) 
 16   16 
 27   27 
N16 = N 3ζ × N 4η =  (ζ 2 − 1)(ζ − 13 )  ×  − (η 2 − 1)(η + 13 ) 
 16   16 

Problem 4: Derive the shape functions for an 8 noded Serendipity quadrilateral element.

4 (-1, 1) 7 (0, 1) η = +1 3 (1, 1)

η
ζ
8 (-1, 0) 6 (1, 0)

ζ = −1 ζ = +1

η = −1
1 (-1,-1) 5 (0, -1) 2 (1, -1)

Compiled by Vijay G. S., Asst. Prof., 27 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

First considering only corner nodes.

For Node 1:
Conditions to be satisfied by N1 are:
(a) N1 = 0 along ζ = 1 ⇒ 1−ζ=0
(b) N1 = 0 along η = 1 ⇒ 1−η=0
(c) N1 = 0 along the line 5-8 ⇒ 1+ζ+η=0
(d) N1 = 1 at node 1, i.e., at ζ = −1 & η = −1

Putting conditions (a), (b) & (c) in the form of a single equation,
N1 = C(1 − ζ) (1 − η) (1 + ζ + η) ------(1)
When coordinates of nodes 2, 3, 4, 5, 6, 7 & 8 are substituted in (1), we get N1 = 0 at these
nodes, which is the requirement to be satisfied by N1.

When condition (d), i.e., coordinates of node 1 is substituted in (1),

1 = C(1 + 1) (1 + 1) (1 −1 − 1)
⇒ C = -1/4
1
∴ N1 = − (1 − ζ )(1 − η )(1 + ζ + η )
4

For Node 2:
Conditions to be satisfied by N2 are:
(a) N2 = 0 along ζ = −1 ⇒ 1+ζ=0
(b) N2 = 0 along η = 1 ⇒ 1−η=0
(c) N2 = 0 along the line 5-6 ⇒ 1−ζ+η=0
(d) N2 = 1 at node 1, i.e., at ζ = +1 & η = −1

Putting conditions (a), (b) & (c) in the form of a single equation,
N2 = C(1 + ζ) (1 − η) (1 − ζ + η) ------(2)
When coordinates of nodes 1, 3, 4, 5, 6, 7 & 8 are substituted in (2), we get N2 = 0 at these
nodes, which is the requirement to be satisfied by N2.

When condition (d), i.e., coordinates of node 2 is substituted in (2),

1 = C(1 + 1) (1 + 1) (1 −1 − 1)
⇒ C = -1/4
1
∴ N 2 = − (1 + ζ )(1 − η )(1 − ζ + η )
4

For Node 3:
Conditions to be satisfied by N3 are:
(a) N3 = 0 along ζ = −1 ⇒ 1+ζ=0
(b) N3 = 0 along η = −1 ⇒ 1+η=0
(c) N3 = 0 along the line 6-7 ⇒ 1−ζ−η=0
(d) N3 = 1 at node 1, i.e., at ζ = +1 & η = +1

Putting conditions (a), (b) & (c) in the form of a single equation,
N3 = C(1 + ζ) (1 + η) (1 − ζ − η) ------(3)
When coordinates of nodes 1, 2, 4, 5, 6, 7 & 8 are substituted in (3), we get N3 = 0 at these
nodes, which is the requirement to be satisfied by N3.

Compiled by Vijay G. S., Asst. Prof., 28 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

When condition (d), i.e., coordinates of node 3 is substituted in (3),

1 = C(1 + 1) (1 + 1) (1 −1 − 1)
⇒ C = -1/4
1
∴ N 3 = − (1 + ζ )(1 + η )(1 − ζ − η )
4

For Node 4:
Conditions to be satisfied by N4 are:
(a) N4 = 0 along ζ = +1 ⇒ 1−ζ=0
(b) N4 = 0 along η = −1 ⇒ 1+η=0
(c) N4 = 0 along the line 7-8 ⇒ 1+ζ−η=0
(d) N4 = 1 at node 1, i.e., at ζ = −1 & η = +1

Putting conditions (a), (b) & (c) in the form of a single equation,
N4 = C(1 − ζ) (1 + η) (1 + ζ − η) ------(4)
When coordinates of nodes 1, 2, 3, 5, 6, 7 & 8 are substituted in (4), we get N4 = 0 at these
nodes, which is the requirement to be satisfied by N4.
When condition (d), i.e., coordinates of node 4 is substituted in (4),
1 = C(1 + 1) (1 + 1) (1 −1 − 1)
⇒ C = -1/4
1
∴ N 4 = − (1 − ζ )(1 + η )(1 + ζ − η )
4

Now, considering the mid side nodes:

For node 5:
Conditions to be satisfied by N5 are:
(a) N5 = 0 along ζ = +1 ⇒ 1−ζ=0
(b) N5 = 0 along ζ = −1 ⇒ 1+ζ=0
(c) N5 = 0 along η = +1 ⇒ 1−η=0
(d) N5 = 1 at node 5, i.e., at ζ = 0 & η = −1

Putting conditions (a), (b) & (c) in the form of a single equation,
N5 = C(1 − ζ) (1 + ζ) (1 − η) = C(1 − ζ2) (1 − η) ------(5)
When coordinates of nodes 1, 2, 3, 4, 6, 7 & 8 are substituted in (5), we get N5 = 0 at these
nodes, which is the requirement to be satisfied by N5.
When condition (d), i.e., coordinates of node 5 is substituted in (5),
1 = C (1 − 0) (1 +1)
⇒ C = 1/2
1
∴ N 5 = (1 − ζ 2 )(1 − η )
2

For node 6:
Conditions to be satisfied by N6 are:
(a) N6 = 0 along ζ = −1 ⇒ 1+ζ=0
(b) N6 = 0 along η = −1 ⇒ 1+η=0
(c) N6 = 0 along η = +1 ⇒ 1−η=0
(d) N6 = 1 at node 6, i.e., at ζ = 1 & η = 0

Compiled by Vijay G. S., Asst. Prof., 29 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Putting conditions (a), (b) & (c) in the form of a single equation,
N6 = C(1 + ζ) (1 + η) (1 − η) = C(1 + ζ) (1 − η2) ------(6)
When coordinates of nodes 1, 2, 3, 4, 5, 7 & 8 are substituted in (6), we get N6 = 0 at these
nodes, which is the requirement to be satisfied by N6.
When condition (d), i.e., coordinates of node 6 is substituted in (6),
1 = C (1 + 1) (1 − 0)
⇒ C = 1/2
1
∴ N 6 = (1 + ζ )(1 − η 2 )
2

For node 7:
Conditions to be satisfied by N7 are:
(a) N7 = 0 along ζ = −1 ⇒ 1+ζ=0
(b) N7 = 0 along ζ = +1 ⇒ 1−ζ=0
(c) N7 = 0 along η = −1 ⇒ 1+η=0
(d) N7 = 1 at node 7, i.e., at ζ = 0 & η = 1

Putting conditions (a), (b) & (c) in the form of a single equation,
N7 = C(1 + ζ) (1 − ζ) (1 + η) = C(1 − ζ2) (1 + η) ------(7)
When coordinates of nodes 1, 2, 3, 4, 5, 6, & 8 are substituted in (7), we get N7 = 0 at these
nodes, which is the requirement to be satisfied by N7.
When condition (d), i.e., coordinates of node 7 is substituted in (7),
1 = C (1 − 0) (1 + 1)
⇒ C = 1/2
1
∴ N 7 = (1 − ζ 2 )(1 + η )
2

For node 8:
Conditions to be satisfied by N8 are:
(a) N8 = 0 along ζ = +1 ⇒ 1−ζ=0
(b) N8 = 0 along η = +1 ⇒ 1−η=0
(c) N8 = 0 along η = −1 ⇒ 1+η=0
(d) N8 = 1 at node 8, i.e., at ζ = −1 & η = 0

Putting conditions (a), (b) & (c) in the form of a single equation,
N8 = C(1 − ζ) (1 − η) (1 + η) = C(1 − ζ) (1 − η2) ------(8)
When coordinates of nodes 1, 2, 3, 4, 5, 6, & 8 are substituted in (8), we get N8 = 0 at these
nodes, which is the requirement to be satisfied by N8.
When condition (d), i.e., coordinates of node 8 is substituted in (8),
1 = C (1 + 1) (1 − 0)
⇒ C = 1/2
1
∴ N8 = (1 − ζ )(1 − η 2 )
2

Compiled by Vijay G. S., Asst. Prof., 30 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Problem 5: Derive the shape functions for a 12 noded Serendipity quadrilateral element.
η = +1
4 10 9 3 Node Coordinates
1 (-1, -1)
ζ = −1 2 (+1, -1)
3 (+1, +1)
11 8 4 (-1, +1)
η 5 (-1/3, -1)
ζ 6 (+1/3, -1)
(0, 0) 7 (+1, -1/3)
8 (+1, +1/3)
12 7 9 (+1/3, +1)
10 (-1/3, +1)
11 (-1, +1/3)
ζ = +1 12 (-1, -1/3)
1 5 6 2
η = −1
It may be noted that the non-corner nodes on the edges (nodes 5, 6, 7, 8, 9, 10, 11 & 12) are
equidistant from the ζ, η origin and hence can be considered to be lying on a circle with
2
1  10
center at (0, 0) and radius =  − 0  + (1 − 0)2 = .
3  3
10
∴ The equation of the circle is ζ 2 + η 2 = and the nodes 5 through 12 satisfy the equation
9
of the circle.

First considering only corner nodes.

For Node 1:
Conditions to be satisfied by N1 are:
(a) N1 = 0 along ζ = 1 ⇒ 1−ζ=0
(b) N1 = 0 along η = 1 ⇒ 1−η=0
10 10
(c) N1 = 0 along the circle ζ 2 + η 2 = ⇒ ζ 2 +η 2 − =0
9 9
(d) N1 = 1 at node 1, i.e., at ζ = −1 & η = −1

Putting conditions (a), (b) & (c) in the form of a single equation,
 10 
N1 = C(1 − ζ) (1 − η)  ζ 2 + η 2 −  ------(1)
 9 
When coordinates of all other nodes except that of node 1 are substituted in (1), we get N1 = 0
at these nodes, which is the requirement to be satisfied by N1.

When condition (d), i.e., coordinates of node 1 is substituted in (1),

 10 
1 = C(1 + 1) (1 + 1)  1 + 1 − 
 9
⇒ C = 9/32

Compiled by Vijay G. S., Asst. Prof., 31 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

9  10 
∴ N1 = (1 − ζ )(1 − η )  ζ 2 + η 2 − 
32  9

Similarly we can derive for nodes 2, 3 & 4:

9  10 
N 2 = (1 + ζ )(1 − η )  ζ 2 + η 2 − 
32  9

9  10 
N3 = (1 + ζ )(1 + η )  ζ 2 + η 2 − 
32  9

9  10 
N4 = (1 − ζ )(1 + η )  ζ 2 + η 2 − 
32  9

Considering the non-corner nodes on the edges

For node 5:
Conditions to be satisfied by N5 are:
(a) N5 = 0 along ζ = +1 ⇒ 1−ζ=0
(b) N5 = 0 along ζ = −1 ⇒ 1+ζ=0
(c) N5 = 0 along η = +1 ⇒ 1−η=0
(d) N5 = 0 at node 6, i.e., at ζ = +1/3 & η = −1 ⇒ ζ − 1/3 = 0 OR 1 − 3ζ = 0
(e) N5 = 1 at node 5, i.e., at (-1/3, -1)

Putting conditions (a), (b), (c) & (d) in the form of a single equation,
N5 = C(1 − ζ) (1 + ζ) (1 − η)( 1 − 3ζ ) ------(2)
When coordinates of all other nodes except that of node 5 are substituted in (2), we get N5 = 0
at these nodes, which is the requirement to be satisfied by N5.

When condition (e), i.e., coordinates of node 5 is substituted in (2),

1 = C (1 + 1/3) (1 – 1/3)(1 + 1)(1+3/3)
⇒ C = 9/32
9
∴ N 5 = (1 − ζ 2 )(1 − η )(1 − 3ζ )
32

Similarly we can derive for nodes 6, 7, 8, 9, 10, 11 & 12:

9 9
∴ N6 = (1 − ζ 2 )(1 − η )(1 + 3ζ ) ∴ N10 = (1 − ζ 2 )(1 + η )(1 − 3ζ )
32 32

9 9
∴ N7 = (1 − η 2 )(1 + ζ )(1 − 3ζ ) ∴ N11 = (1 − η 2 )(1 − ζ )(1 + 3ζ )
32 32

9 9
∴ N8 = (1 − η 2 )(1 + ζ )(1 + 3ζ ) ∴ N12 = (1 − η 2 )(1 − ζ )(1 − 3ζ )
32 32

9
∴ N9 = (1 − ζ 2 )(1 + η )(1 + 3ζ )
32

Compiled by Vijay G. S., Asst. Prof., 32 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Iso-parametric, Sub-parametric and Super-parametric Elements

Iso-parametric elements:
The elements for which the geometry and the displacements are represented by the same
shape functions are called iso–parametric elements.

Example: Suppose N1, N2, N3 and N4, represents the shape functions of a 4 noded
quadrilateral elements,

 x = N1 x1 + N 2 x2 + N 3 x3 + N 4 x4
The geometry  ---------(1)
 y = N1 y1 + N 2 y2 + N3 y3 + N 4 y4
u = N1 q1 + N 2 q3 + N 3 q5 + N 4 q7
and displacement  ---------(2)
v = N1 q2 + N 2 q4 + N 3 q6 + N 4 q8

are represented by same shape functions N1, N2, N3 and N4. Hence, the 4 noded quadrilateral
element is an iso-parametric element.

It may be noted that equations (1) and (2) are of the same order and contain same
number of terms. i.e., all the nodes (terms in displacement polynomial) of the element are
used to represent geometry and displacement.

Sub-parametric elements: The elements (including higher order elements) for which the
geometry is represented by lesser terms and of lesser order when compared to the
representation of displacement.

Example: For a beam element with 2 nodes, the shape functions N1 and N2, used for linear
1D element are utilized for representing geometry.

i.e., x = N1x1 + N2x2 1 2 x

q2 q4
q1 q3
But for representing displacements (vertical deflection and slope) some other shape
functions H1, H2, H3 and H4 are used which are called Hermite cubic shape functions. These
shape functions conveniently represent deflection and slope at a node. Note that deflection is
perpendicular to beam axis. It can also be noted that H1, H2, H3 and H4 are 4 terms and of
higher order (cubic). Where as N1 and N2 are 2 terms and of lower order (linear).

Another example is the nine noded quadrilateral element. When curved edges are
involved, the shape functions for all nodes are used to represent geometry. When straight
(linear) edges are involved, only 4 corner nodes shape functions are used to represent
geometry which is sub-parametric representation.

Compiled by Vijay G. S., Asst. Prof., 33 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM
η

η 7
4 3
3
7

4
6
6 8 ς
ς 9
8 9

2 1 5 2
1 5

All shape functions are used to Only corner node shape functions
represent geometry (shape) are used to represent geometry
(Iso-Parametric representation) (Sub-Parametric representation)

Super-parametric elements: These are elements which are different from the iso-parametric
and sub-parametric elements. i.e., higher order functions with larger number of terms are
required to represent geometry. This means, representation of displacement is done by lesser
number of terms and of lower order.

Note: All iso-parametric and sub-parametric elements satisfy the convergence requirements,
where as only a few super-parametric elements are able to satisfy convergence requirements.

Types of Boundary Conditions

In structural problems we normally come across two types of boundary conditions

1. Specified displacement boundary condition
2. Multi point constraint boundary condition

(1) Specified displacement boundary condition:

If the boundary conditions are of the type Q1 = a1, Q2 = a2, Q3 = a3…………Qr = ar they
are called as specified displacement boundary condition. There can be two types of Specified
displacement boundary conditions,
(a) Homogeneous Boundary Condition and
(b) Non- Homogeneous Boundary Condition

If the specified displacement is zero it is called Homogeneous Boundary Condition. In the

figure shown below, the left end of the bar is fixed. So the displacement boundary condition
at the left end is zero. That is, it is a Homogeneous Boundary Condition.

Compiled by Vijay G. S., Asst. Prof., 34 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

If the specified displacement is non-zero it is called Non-Homogeneous Boundary

Condition. In the figure shown below, the left end of the bar is fixed. So the displacement
boundary condition at the left end is zero. That is, it is a Homogeneous Boundary Condition.
But the right end of the bar is free to displace by a non-zero amount δ. Hence, the
displacement boundary condition at the left end is a Non-Homogeneous Boundary Condition.

(2) Multi point constraint boundary condition:

If the boundary conditions are of the type β1Q1+β2Q2 = β0 where β0, β1, β2 are known
constants then they are called as multi point constraints. Multi point constraints are used for
modeling, inclined roller support, rigid connections or shrink fits, etc.

Steel Bar
Rigid Aluminium
Link Bar b
a

Q1 Q2
a
b
Q1 = (a/b)Q2
P
In the example of rigid link shown in the above figure, Q1 & Q2 are the displacements of the
free ends of the steel and aluminium bars. They are dependent on each other as per the
relation Q1 = (a/b)Q2. This condition is of the type β1Q1+β2Q2 = β0. Hence, the type of
boundary conditions used here are multi point constraints.

Another way of classifying Boundary Conditions are:

(1) Natural BCs
(2) Essential BCs

u
In the above figure the loads P & Q are the Natural Boundary Conditions and the specified
displacement Q1 = 0 & Q3 = δ are the Essential Boundary Conditions. u
Compiled by Vijay G. S., Asst. Prof., 35 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Methods of handling boundary conditions:

Boundary conditions can be handled by two methods:
1. Elimination approach
2. Penalty method

(1) Elimination approach:

First we write the {F} = [K]{Q}relationship

 F1   K11 K12 K13 .. .. K1n   Q1 
F   K K 22 K 23 .. .. K 2 n   Q2 
 2   21
 F3   K 31 K 32 K 33 .. .. K 3n   Q3 
 =   
.   .. .. .. .. .. ..   
.   .. .. .. .. .. ..   
    
 Fn   K n1 K n2 K n3 .. .. K nn  QN 

Then we apply the displacement boundary conditions. Now we cancel all the rows
and columns of the above relationship corresponding to the degrees of the freedom which are
specified.

Example: Suppose in a problem we have total of 4 degrees of freedom, then the size of the
global stiffness matrix is 4×4. So the {F} = [K]{Q} relation ship will be:
 F1   K11 K12 K13 K14   Q1 
F   K K 24  Q2 
 2   21 K 22 K 23
 =  
 F3   K 31 K 32 K 33 K 34  Q3 
 F4   K 41 K 42 K 43

K 44  Q4 
Suppose two degrees of freedom Q1 and Q3 are specified to be a1 and a3, then rows and
column corresponding to degrees of freedom 1 and 3 are eliminated.
 F1   K11 K12 K13 K14   Q1 = a1 
F   K K 24   Q2 
 2   21 K 22 K 23
 =  
 F3   K 31 K 32 K 33 K 34  Q3 = a3 
 F4   K 41 K 42 K 43

K 44   Q4 
The modified {F} = [K]{Q} relationship becomes:

F2 − K 21 a1 − K 23 a3   K 22 K 24  Q2 
  =  
F4 − K 41 a1 − K 43 a3   K 42 K 44  Q4 

Using one of the matrix methods of solution the unknown displacements Q2 & Q3 are
evaluated.
If the specified displacements a1 and a3 are zeroes, then the force vector does not get
modified.

The reaction forces at the ith support is evaluated as:

Ri = (ith row of [K]){Q} − Fi

Compiled by Vijay G. S., Asst. Prof., 36 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

where, [K] is the global stiffness matrix, {Q} is the global force vector and Fi the external
force applied at DOF number i.

The reaction forces at the supports, i.e., at DOFs 1 and 3 can be evaluated as follows:
Reaction force at DOF number 1 is
 Q1 
Q 
 
R1 = {K11 K12 K13 K14 } ×  2  − F1
Q3 
Q4 
i.e., R1 = K11Q1 + K12Q2 + K13Q3 + K14Q4 − F1

Similarly, the reaction force at DOF number 3 is

 Q1 
Q 
 
R3 = { K 31 K 32 K 33 K 34 } ×  2  − F3
Q3 
Q4 
i.e., R3 = K 31Q1 + K 32 Q2 + K 33Q3 + K 34 Q4 − F3

(2) Penalty approach:

First we write the {F} = [K]{Q}relationship

 F1   K11 K12 K13 .. .. K1n   Q1 
F   K K 22 K 23 .. .. K 2 n   Q2 
 2   21
 F3   K 31 K 32 K 33 .. .. K 3n   Q3 
 =   
.   .. .. .. .. .. ..   
.   .. .. .. .. .. ..   
    
 Fn   K n1 K n2 K n3 .. .. K nn  QN 

Then we apply the displacement boundary conditions. Then we define a high stiffness
value C as C = |Max (Kij)| × 104. The global stiffness matrix [K] is modified by adding C to
those diagonal elements of [K] corresponding to the specified degree of freedom. The global
force vector {F} is also modified by adding Cai to Fi, where, ai is the specified displacement
at DOF number i.

Example: Suppose in a problem we have total of 4 degrees of freedom, then the size of the
global stiffness matrix is 4×4. So the {F} = [K]{Q} relation ship will be:
 F1   K11 K12 K13 K14   Q1 
F   K K 24  Q2 
 2   21 K 22 K 23
 =  
 F3   K 31 K 32 K 33 K 34  Q3 
 F4   K 41 K 42 K 43

K 44  Q4 

Compiled by Vijay G. S., Asst. Prof., 37 Dept of Mech. & Mfg. Engg., MIT, Manipal
 Basics of FEM

Suppose two degrees of freedom Q1 and Q3 are specified to be a1 and a3, then C is
added to K11 and K33. Also the corresponding elements of the global force vector will be
modified by adding Ca1 to F1 and Ca3 to F3. The above {F} = [K]{Q} relationship will be
modified as follows.
F1 + Ca1   K11 + C K12 K13 K14  Q1 
F   K K 24  Q2 
 2   K 22 K 23
 
=
21
 
 F3 + Ca3   K 31 K 32 K 33 + C K 34
 Q3 
F4 
  K 41 
K 42 K 43 K 44  Q4 

Using one of the matrix methods of solution the unknown displacements are
evaluated. If the specified displacements a1 and a3 are zeroes, then the global force vector
does not get modified.

The reaction forces at the ith support is evaluated as:

Ri = −C (Qi − ai)
where, Qi is the evaluated displacement at DOF number i and ai is the specified displacement
at DOF number i.

Hence, the reaction forces at the supports, i.e., at DOFs 1 and 3 can be evaluated as follows:
R1 = −C (Q1 − a1)
R3 = −C (Q3 − a3)

Compiled by Vijay G. S., Asst. Prof., 38 Dept of Mech. & Mfg. Engg., MIT, Manipal