Computer Implementation of

the Finite Element Method

Mr. M. S. Kadam
Assistant Professor,
Sanjay Ghodawat Group of Institutions,
Atigre
 Q.1: Write a short note on
bandwidth and node numbering?
• In the global stiffness matrix there are several zero (0) terms
besides the diagonal terms. Such a matrix is called as banded
matrix.
I<-B->I

• This is usually a feature of finite element model. A band of non-

zero terms surrounds the diagonal values and remaining locations
are filled with zero’s

2
 Q.1: Continued..

• The smaller the bandwidth, the easier and faster the matrix can be
solved by using elimination method.

• Size of bandwidth depends directly upon the element numbering

and consequently the node numbering.

• Mathematically the bandwidth is given by,

B=(D+1)F
Where,
B = Bandwidth
D = Maximum difference in node number
F = Degrees of freedom 3
 Q.1: Continued..

• Variations in the node numbering can be well illustrated by the

following example,

Assumption: DOF is 1 in above case.

• Numbering is in longer side and in spiral manner,
D=7;B=(7+1)x1=8
• Numbering is in longer side and restarted from initial end,
D=4;B=(4+1)x1=5
4
 Q.1: Continued..

• Numbering is in shorter side and in spiral manner,

D=5;B=(5+1)x1=6
• Numbering is in shorter side and restarted from initial end,
D=3;B=(3+1)x1=4

• Therefore it seen that the bandwidth is minimum if numbering is

done in shorter direction and restarted from the initial end after
reaching the other end.

• In the many standard commercial systems node numbering is

done automatically to keep the bandwidth to minimum.
5
 Q.2: Write a note on element aspect ratio? OR
Explain methods of validation of FEA results
• The shape of element is the major factor which affects the
accuracy of the analysis.
• The aspect ratio is defined as the ratio of largest side of an element
to the smallest side.
• It is recommended that the aspect ratio should be as close to unity
as possible.
• For triangular elements the aspect ratio is defined as the ratio of
height of triangle to the base whereas for rectangular elements it is
the ratio of length of rectangle to the width.

6
 Q.2: Continued..

• The accuracy of FEA is inversely proportional to the aspect ratio

of the element.
• This means that as aspect ratio decreases the accuracy of FEA
increases, but the accuracy is best when the aspect ratio is 1.
• The variation of aspect ratio for the same domain is shown by an
example in the figure below,

1 2 3 4 5 6

1 2 3 4 5 6 7 8 9 10 11 12

7 8 9 10 11 12

(a) (b) 7
 Q.2: Continued..
1 2 3
1 2 3 4

4 5 6
5 6 7 8
7 8 9

9 10 11 12
10 11 12

(c) (d)

1
1 2 2
3
3 4 4
5
5 6 6
7
7 8 8
9
9 10 10
11
11 12 12

(e) (f)
8
Q.2: Continued..

9
 Q.3:Explain the measures of element
distortion in a finite element analysis package
• The finite element method approximates the distribution of a
variable through a component by assuming a known distribution
of that variable in each element.

• However as the elements becomes distorted, so errors are

introduced into the element approximations.

• Higher order elements are less sensitive to element distortion and

the degree of allowable distortion depends on the stress gradient
(or heat flow) experienced by the element.

• The smaller the gradient the more the element can be distorted, so
its in inverse proportion.
10
 Q.3:Continued..

• Element distortions relate to the skew, taper, warpage and aspect

ratio of the faces and sides of the elements.
• In addition incorrect positioning of the mid-side nodes of the
higher order elements and unreasonable internal angles of the
elements indicate distortion.
• The type and number of element checks offered by finite element
packages vary considerably from one package to another. The
most common checks are based on,
1. Aspect ratio
2. Internal angles
3. Warpage angles
4. Mid-side Node position
5. Value of the Jacobian
11
 Q.3:Continued..
1. Aspect ratio:
A.R.= Length / Width (Rectangle)

A.R.=Height / Base (Triangle)

• The aspect ratio of an element is the ratio of the length of the

largest side of the element to the shortest.
• For rectangular elements it is easy to observe the aspect ratio but
for quadrilateral elements which are skewed or tapered and for
triangular elements calculation of aspect ratio is not so obvious.
• Accuracy of FEA increases as aspect ratio approaches towards
unity.
12
 Q.3:Continued..

2. Internal angles:
• Allowable limits for the angles depend on the field variable
distribution and the order of the elements under consideration.
• For rectangular elements: 45°<α<135°
• For triangular elements: 45°<α<75°

3. Warpage angles:
• Warping occurs when the nodes of a quadrilateral elements
(or face of solid element) do not lie in same plane.
• As a guide the warping angle should not exceed 10°.

13
 Q.3:Continued..

4. Mid-side node position:

• The position of mid-side node can affect the accuracy of the
solution.
• Ideally the mid-side node should be placed midway between the
corner nodes.
• As it moves away from mid-position error increases.
• It should not be placed more than 1/8th of the sides length from
the centre position.

Mid-side node off-centre Mid-side node off-line

14
 Q.3:Continued..

5. Value of the Jacobian:

• Element distortion can be related to the Jacobian.
• A general distortion index can be devised for a particular element
by calculating the ratio of the minimum Jacobian to the average
Jacobian.

• Using such schemes a distortion of 1 indicates a perfectly shaped

element.
• Values of the index from 0.6 to 1.0 are generally acceptable,
while ratios of less than 0.4 should be avoided.

15
 Q.4:Write a short note on mesh refinement

• The accuracy of a model can be improved in two ways,

- Either the geometry can be divided into smaller elements, so that
the mesh density is increased.
- Or the accuracy of the elements themselves can be improved by
using higher order interpolation functions.
- These two techniques to improve a model’s accuracy are known
as, h-refinement and p-refinement.

• The later appears the simpler of the two methods for the user of a
finite element package, but in practice commercial programs only
offer linear and quadratic (and occasionally cubic) elements, so
that the opportunity for p-refinement is limited.
16
 Q.4:Continued..

• Also increasing the element order from linear to quadratic leads

to significant increase in the computer time needed to analyze the
structure.

• To confirm the convergence of a model by progressive h-

refinement of the mesh, all previous meshes should be contained
in the finer meshes. This is generally referred to as a reducible
net.

• Let us consider an example of reducible meshes for examining

convergence. (progressive mesh size)

17
 Types of Meshing

• Free Meshing:
Random discretization of the domain using irregular elements

• Mapped Meshing:
Regular and uniform discretization of the domain into elements.

Let us consider an example of meshing for area domain by

above two techniques

18
 Q.5:Describe the concept:
‘simplification through symmetry’
• There are four common types of symmetry encountered in
engineering problems.

• If the geometry or the configuration of the body and the external

conditions (i.e. boundary conditions) are similarly symmetric,
then only the repeated part of the structure needs to be modeled.

• It is vital however that the loading and constraint conditions are

applied to the parts model in such a way that they truly reflect the
symmetry of the problem.

19
 Q.5:Continued..

1. Axial symmetry:
• Since axial symmetry is
encountered so frequently, axi-
symmetric elements are included
in finite element packages.
• Such symmetry is mainly used
for cylindrical walls, ring shaped
solids, pressure vessels, conical
frustums etc.

20
 Q.5:Continued..

2. Planar symmetry:
• Planar symmetry is well
illustrated by the case of a flat
plate with a circular hole in it,
loaded uniformly.
• These are generally exposed by
shell element in FEA packages.
• These are 2D area element
having thickness in z direction.

21
 Q.5:Continued..

3. Cyclic symmetry:
• Cyclic symmetry is present in
spline fittings, propellers,
couplings etc.
• The displacement are zero ‘0’ in
circumferential direction for the
stress problems.
• These can utilise the cylindrical
coordinate system, and can be
rotated to 1/2, 1/4th and 1/8th
etc.

22
 Q.5:Continued..

4. Repetitive symmetry:
• For repetitive symmetry
problems the common
boundaries of the repeated
segments are constrained in a
perpendicular direction.
• In this the structures are
modeled assuming plane strain
systems e.g. chain, train etc.

23
 Q.6:Explain various attempts made to reduce
memory requirements in storing the stiffness matrix

• In FEM the size of matrices can goes as high as 1000x1000 or

even more, hence memory requirement for storing stiffness
matrix is very high.
• If matrix is solved by elimination method then huge storage
memory is required, there are several different techniques are
used to reduce the memory requirement.

24
 Q.6:Continued..

1. Use of symmetry of banded matrix:

• Since the stiffness matrix is always symmetric and banded in
nature this technique have been used to store only the semi-
bandwidth of non-zero elements and get the solution.
• If Bs is the semi-bandwidth of NxN matrix then it is required to
store only NxBs elements as shown in figure.

25
 Q.6:Continued..

2. Partitioning matrix:
• For layer systems the method of symmetry or banded matrix is
not sufficient, in such cases partitioning of the matrix is made.
• In this only few triangular sub-matrices need to be stored in
computer core at a given time while the remaining locations are
kept in peripheral storage like hard disk.

26
 Q.6:Continued..

3. Skyline storage technique:

• The drawn through the matrix such that all the elements above
that line are zero and below the line there may be the
combination of zero and non-zero terms, is known as skyline.
• Hence in this technique all the elements above the line are sent
in peripheral devices to reduce the memory requirement.
• All the elements below the line are stored in computer core.

27
Q.7:Explain in detail the steps to be carried out in
commercial FEA software for obtaining the displacement
in stepped bar each of length L/2, fixed at one end and
subjected to axial force P at the other end.

A1
A2

1 2 P

L/2 L/2

28
 Q.7:Continued..

• Consider that we have two steps 1 & 2, hence we have two

elements. Therefore assuming constant material for the bar,
For elements 1,
A1= Area of cross section for element 1
E= Modulus of elasticity of bar
L1=L/2= Length of element 1
For elements 2,
A2= Area of cross section for element 2
E= Modulus of elasticity of bar
L2=L/2= Length of element 2

29
 Q.7:Continued..

• FEA procedure of commercial software:

Go to Preferences – select structural – click on OK
Select Element type – add/edit/delete – 1D-link – click on OK
Go to Real constant – 1. Area of c/s: A1, 2. Area of c/s: A2 – OK
Define material properties – Material model – elastic – isotropic:
Modulus of elasticity: E, Poissions ratio: γ – OK
Go to Modeling – create nodes: 1, 2, 3 – coordinates of these
nodes will be,
Node X y
1 0 0
2 L/2 0
3 L 0
30
 Q.7:Continued..

Create elements – through nodes – OK

e i j
1 1 2
2 2 3

Select loads – Define loads –

type – structural – displacement – Select node 1: All DOF = 0 (This is the
boundary condition because node 1 is fixed
type – structural –force – Select node 3: Fx = P This is the
applied load (positive/tensile)
Go to Solution phase – click on solve – current LS (i.e. current load
system) – OK

31
 Q.7:Continued..

Select Postprocessor – for finding other important information like –

Reactions at support, Elemental stresses, strains, displacement vector
sum etc.
Go to plot – deformed and un-deformed shape – OK
List – Results – Nodal solution –
1: Displacement – OK
2: Von-Mises stresses – OK
3: Equivalent strains – OK

Click on (X) close – (pop up window) – save & exit.

These are the steps to be followed for solving a stepped bar problem.

32