1

2
Please read this disclaimer before proceeding:
This document is confidential and intended solely for the educational purpose of
RMK Group of Educational Institutions. If you have received this document
through email in error, please notify the system manager. This document
contains proprietary information and is intended only to the respective group /
learning community as intended. If you are not the addressee you should not
disseminate, distribute or copy through e-mail. Please notify the sender
immediately by e-mail if you have received this document by mistake and delete
this document from your system. If you are not the intended recipient you are
notified that disclosing, copying, distributing or taking any action in reliance on
the contents of this information is strictly prohibited.

3
 ME8692 FINITE ELEMENT ANALYSIS

Department Mechanical Engineering

Batch/Year 2018 – 22/III
Created by Dr. G.Mahendran, ASP/Mech, RMKEC.
Mr. P.Duraimurugan, AP/Mech, RMKEC.
Mr. R. Ashok Kumar, AP/Mech, RMKCET.
Date 18.03.2021

4
 Table of Contents

S.NO CONTENTS PAGE NO

1 Course Objectives 06

2 Pre Requisites 06

3 Syllabus 07

4 Course outcomes 08

5 CO- PO/PSO Mapping 09

6 Lecture Plan 10

7 Activity based learning 11

8 Lecture Notes 15

9 Assignments 107

10 Part A Q & A 109

11 Part B Qs 112

12 Supportive online Certification courses 115

13 Real time Applications in day to day life and to 116

Industry
14 Contents beyond the Syllabus 117

15 Assessment Schedule 118

16 Prescribed Text Books & Reference Books 119

17 Mini Project suggestions 120

5
 I - Course Objectives
01 To introduce the concepts of Mathematical Modeling of
Engineering Problems.
02 To appreciate the use of FEM to a range of Engineering
Problems.

II - Pre Requisite Courses

GE8292 Engineering
Mechanics (Semester 2)
CE8395 Strength of
Materials for Mechanical
Engineers (Semester -4)

5
 III - Syllabus

ME8692 FINITE ELEMENT ANALYSIS L T PC

3 0 03

UNIT I INTRODUCTION 9
Historical Background – Mathematical Modeling of field problems in Engineering
– Governing Equations – Discrete and continuous models – Boundary, Initial and
Eigen Value problems– Weighted Residual Methods – Variational Formulation of
Boundary Value Problems – Ritz Technique – Basic concepts of the Finite Element
Method.

UNIT II ONE-DIMENSIONAL PROBLEMS 9

One Dimensional Second Order Equations – Discretization – Element types-
Linear and Higher order Elements – Derivation of Shape functions and Stiffness
matrices and force vectors- Assembly of Matrices - Solution of problems from
solid mechanics and heat transfer. Longitudinal vibration frequencies and mode
shapes. Fourth Order Beam Equation –Transverse deflections and Natural
frequencies of beams.

UNIT III TWO DIMENSIONAL SCALAR VARIABLE PROBLEMS 9

Second Order 2D Equations involving Scalar Variable Functions – Variational
formulation –Finite Element formulation – Triangular elements – Shape functions
and element matrices and vectors. Application to Field Problems - Thermal
problems – Torsion of Non circular shafts –Quadrilateral elements – Higher Order
Elements.

UNIT IV TWO DIMENSIONAL VECTOR VARIABLE PROBLEMS 9

Equations of elasticity – Plane stress, plane strain and axisymmetric problems –
Body forces and temperature effects – Stress calculations - Plate and shell
elements.

UNIT V ISOPARAMETRIC FORMULATION 9

Natural co-ordinate systems – Isoparametric elements – Shape functions for iso
parametric elements – One and two dimensions – Serendipity elements –
Numerical integration and application to plane stress problems - Matrix solution
techniques – Solutions Techniques to Dynamic problems – Introduction to
Analysis Software.

TOTAL PERIODS : 45

7
 IV - Course Outcomes
(COs)Course Outcome (CO) - Statement
CO. No. Knowledge
Level

CO1# Summarize the basics of finite element

K2
formulation.

CO2# Apply finite element formulations to solve one

K2
dimensional Problems

CO3# Apply finite element formulations to solve two

K2
dimensional scalar Problems..

CO4# Apply finite element method to solve two

K2
dimensional Vector problems.

CO5# Apply finite element method to solve problems on

K2
iso parametric element and dynamic Problems.

CO6# Apply finite element method to solve coupled field

K3
problems.

V - Over all Plan: 45

Hours
Units I II III IV V
No. of Proposed 9 9 9 9 9
Periods Actual

Course Augmentation
Units I II III IV V
Assignement 09.03.2021 23.03.2021 10.04.2021 20.04.2021 07.05.2021

Quiz

Tutorial 08.03.2021 22.03.2021 09.04.2021 19.04.2021 07.05.2021

Video lectures
(NPTEL/ EDUSAT/
Online)

8
 VI - COs-POs Matrices
POs CO PO1 PO 2 PO 3 PO 4 PO 5 PO 6 PO 7 PO 8 PO 9 PO 10 PO 11 PO 12
COs
C305
.1 CO1 3 3 3
C305
.2 CO2 3 3 2
C305
.3 CO3 3 3 3
C305
.4 CO4 3 3 3
C305
.5 CO5 3 3 2
C305
.6 CO6 3 3 3
1: Slight (Low) 2: Moderate (Medium) 3: Substantial (High)

VII - COs-PSOs Matrices

PSOs PSO PSO PSO

1 2 3
COs CO
C305
.1 CO1 3
C305
.2 CO2 3
C305
.3 CO3 3
C305
.4 CO4 3
C305
.5 CO5 3
C305
.6 CO6 3

1: Slight (Low) 2: Moderate (Medium) 3: Substantial (High)

9
 UNIT II: ONE-DIMENSIONAL PROBLEMS
One Dimensional Second Order Equations – Discretization – Element types- Linear and
Higher order Elements – Derivation of Shape functions and Stiffness matrices and force
vectors- Assembly of Matrices - Solution of problems from solid mechanics and heat
transfer. Longitudinal vibration frequencies and mode shapes. Fourth Order Beam
Equation –Transverse deflections and Natural frequencies of beams.

LECTURE PLAN
Session Mode of
Topics to be covered Reference
No. Delivery*
1 One Dimensional Second Order Equations 1,2
2 Discretization 1,2
3 Element types 1,2
4 Linear and Higher order Elements 2,3
5 Derivation of Shape functions and Stiffness Chalk and 2,3
matrices and force vectors Talk/Online
6 Assembly of Matrices Webinars 1,2
7 Solution of problems from solid mechanics and
1,2
heat transfer
8 Solution of problems from solid mechanics and
1,2
heat transfer
Longitudinal vibration frequencies and mode
9 1,2
shapes
Longitudinal vibration frequencies and mode
10 1,2
shapes
11 Fourth Order Beam Equation 1,2
Transverse deflections and Natural frequencies
12 1,2
of beams

CONTENT BEYOND THE SYLLABUS : Role of CAE in Product Design and

Development.
NPTEL REFERENCES :
https://www.digimat.in/nptel/courses/video/112104193/L01.html
OTHER REFERENCES / WEBSITES :
1. Reddy. J.N., “An Introduction to the Finite Element Method”, 3rd Edition, Tata
McGraw- Hill, 2005.
2. P.Seshu, “Text Book of Finite Element Analysis”, Prentice- Hall of India Pvt. Ltd.,
New Delhi, 2007. ISBN-978-203-2315-5.

*Chalk and Talk, PPT / OHP, Quiz, Group Discussion, Student, Seminar,
Case Study, Field Study, Video (NPTEL/EDUSAT/ONLINE), Crossword
Puzzle, Tutorial Session

10
 ACTIVITY BASED LEARNING

1. If the structure is divided into discrete areas or volumes then it is called an

_______
a) Structure
b) Element
c) Matrix
d) Boundaries

2. In finite element modeling every element connects to _______

a) 4 nodes
b) 3 nodes
c) 2 nodes
d) Infinite no of nodes

3. Discretization includes __________ numbering.

a) Element and node

b) Only nodal
c) Only elemental
d) Either nodal or elemental

❖ 4. Global nodes corresponds to _______

a) Entire body
b) On surface
c) On interface
d) On element

❖ 5. Shape function is just a ___________

a) Displacement function
b) Equation
c) Interpolation function
d) Matrix function

❖ 6. In shape functions, _________ must be continuous across the element

boundary.
a) Derivatives
b) Nodes
c) Displacement
d) Shape function

13
 ACTIVITY BASED LEARNING

7. . Continuum is discretized into_______ elements.

a) Infinite
b) Finite
c) Unique
d) Equal

8. Formula for global stiffness matrix is ____________

a) No. of nodes*Degrees of freedom per node
b) No. of nodes
c) Degrees of freedom per node
d) No. of elements

9. Galerkin technique is also called as _____________

a) Variational functional approach
b) Direct approach
c) Weighted residual technique
d) Variational technique

10. Element connectivity is used for _____

a) Traction force
b) Assembling
c) Stiffness matrix
d) Virtual work

11. By the Galerkin approach equation can be written as __________

a) {P}-{K}{Δ}=0
b) {K}-{P}{Δ}=0
c) {Δ}-{p}{K}=0
d) Undefined

12. How is Assembly of stiffness matrix symbolically denoted?

a) K={k}e
b) K←∑eKe
c) K←∑Ke
d) Undefined

16
 ACTIVITY BASED LEARNING

13. From where does the global load vector F is assembled?

a) Element force vectors only
b) Point loads only
c) Both element force vectors and point loads
d) Undefined

14. Dimension of global stiffness matrix is _______

a) N X N, where N is no of nodes
b) M X N, where M is no of rows and N is no of columns
c) Linear
d) Eliminated

15. Global stiffness K is a______ matrix.

a) Identity matrix
b) Upper triangular matrix
c) Lower triangular matrix
d) Banded matrix

16. Types of Boundary conditions are ______

a) Potential- Energy approach
b) Penalty approach
c) Elimination approach
d) Both penalty approach and elimination approach

17. Equilibrium conditions are obtained by minimizing ______

a) Kinetic energy
b) Force
c) Potential energy
d) Load

18. In elimination approach, which elements are eliminated from a matrix ____
a) Force
b) Load
c) Rows and columns
d) Undefined

16
 ACTIVITY BASED LEARNING

19. Penalty approach method is easy to implement in a ______

a) Stiffness matrix
b) Iterative equations
c) Computer program
d) Cg solving

20. Quadratic shape functions give much more _______

a) Precision
b) Accuracy
c) Both Precision and accuracy
d) Identity

Answer:
1.b, 2. c, 3.a, 4.a, 5.c, 6. c, 7. b, 8. a, 9. c, 10. b, 11. a, 12. b, 13. c, 14. b,
15. d, 16. d, 17.c, 18. c, 19.c, 20.c

16
 UNIT II : ONE-DIMENSIONAL PROBLEMS
Lecture Notes

Piece-wise Continuous Trial Functions—Finite Element Method

In the examples so far, we have used a single composite trial function

solution for the entire domain. When we consider the essence of the R-R method,
i.e., assuming a trial function solution and matching it as closely as possible to the
exact solution, we realise that this is essentially a process of "curve fitting". It is well
known that curve fitting is best done piece-wise—the more the number of pieces,
the better the fit. We can thus benefit by using piece-wise defined trial functions
rather than a set of functions valid over the entire domain of the problem. We will
now discuss this technique leading to the formulation of the finite element method
starting from a variational principle. We recall that the evaluation of the total
potential ∏𝑝 involved integration over the entire domain of the problem. With the
discretisation of the structure into sub-domains and use of piece-wise defined trial
functions, we will evaluate these integrals over each sub-domain ("finite element")
and sum up over all the elements.

Bar Element Formulated from the Stationarity of a Functional

Consider a typical bar element as shown in Figure with two nodes and the
axial displacement 𝑢 as the nodal d.o.f.

𝑢 = (1 − 𝑥 ) 𝑢 1 + 𝑥 𝑢2 (1)
ℓ ℓ

Thus the strain energy stored within the element is given by

ℓ 𝐴𝐸 𝑑 𝑢 2 𝐴𝐸 ( 𝑢 2 −𝑢 1 ) 2
𝑈𝑒 = ∫ ( ) 𝑑𝑥 = (2)
0 2 𝑑𝑥 2 ℓ
 If there is a distributed force 𝑞0 acting at each point on the element and
concentrated forces F at the nodes, the potential of the external forces is given by

𝑉 𝑒 = − ∫ℓ 𝑞0 𝑢𝑑𝑥 − 𝐹1 𝑢 1 − 𝐹2 𝑢 2 = −𝑞0 ℓ (𝑢 1 + 𝑢 2 ) − 𝐹1 𝑢 1 − 𝐹2 𝑢 2 ) (3)

0 2

Thus the total potential for the Mi element is given by

∏𝑒𝑝 = 𝑈𝑒 + 𝑉𝑒 =
𝐴𝐸 𝑢 2 −𝑢 2
− 𝑞 0 ℓ (𝑢 1 + 𝑢 2 ) − 𝐹1 𝑢 1 − 𝐹2 𝑢 2 ) (4)
ℓ 2
2

For the entire structure, summing up the contributions of all elements, we

have

𝜕∏𝑝 = ∑𝑛𝑘 =1 𝑒
(5)
∏𝑝

By virtue of the PSTP, the total potential will be stationary with respect to
small variations in the displacement field. The variations in the displacement field in
our case are attained by small variations in the nodal d.o.f, ui . Thus,
𝜕 ∏𝑝
= 0, 𝑖 = 1,2, . . , 𝑛 (6)
𝜕
𝑢𝑖

When applied to Eq. (5) above, this will lead to a set of simultaneous linear
algebraic equations in the nodal d.o.f., in general, designated as [K]{} = {f}. The
contributions of the Mi element to these system level equations can be obtained by
applying Eq. (6) to the element total potential as given by Eq. (4). Thus, we have
the element level contributions as

(𝑢 − 𝑢 ) = 𝑞 0 ℓ + 𝐹
𝜕 ∏𝑒𝑝 𝐴𝐸
=0⇒
𝜕 𝑢1 ℓ 1 2 2 1

(𝑢 − 𝑢 ) = 𝑞 0 ℓ + 𝐹
𝜕 ∏𝑒𝑝 𝐴𝐸
=0⇒ (7)
𝜕 𝑢2 ℓ 2 1 2 2

In matrix form, we write the element level equations as follows:

 −1 𝑢 1 𝑞0 ℓ/2 𝐹
𝐴𝐸
[1 ] {𝑢 } = { } + { 1} (8)
ℓ −1 1 2 𝑞0 ℓ/2 𝐹2

We observe that these are identical to those equations obtained by the weak
form of the Galerkin weighted residual statement, starting from the governing
differential equation of the problem. The process of assembling together individual
element contributions is identical to the assembly procedure we used earlier.

One-dimensional Heat Transfer Element Based on the

Stationarity of a Functional

Consider a typical one-dimensional heat conduction element as shown in

Figure. with two nodes and the temperature T as the nodal d.o.f. The governing
functional is given by

1 ℓ 𝑑𝑇 2
∏ = 2 ∫0 𝑘 ( 𝑑 𝑥 ) 𝑑𝑥 − ∫ 𝑞 0 𝑇𝑑𝑥 − 𝑄1 𝑇1 − 𝑄1 𝑇2 (9)

Q
where q0 is internal heat source and 1 Q2 represent heat flux at the nodes. We will
use the same interpolation functions as we used previously, i.e., at any point within
the element we have

𝑇(𝑥) = (1 − 𝑥 ) 𝑇1 + 𝑥 𝑇2 (10)
ℓ ℓ

Thus the element level functional is given by

∏𝑒 = 1 ( 𝑘 (𝑇 − 𝑇 )2 ) − 𝑞 0 ℓ (𝑇 − 𝑇 ) − 𝑄 − 𝑄2 𝑇2 (11)
2 ℓ 2 1 2 1 2 1 1
𝑇

For the entire system, summing up the contributions of all elements, we have

∏ = ∑𝑛𝑘 = 1 ∏𝑒 (12)
 The functional will be made stationary with respect to small variations in the
temperature field. The variations in the temperature field in our case are attained by
small variations in the nodal d.o.f. Ti Thus,

𝜕∏
= 0, 𝑖 = 1,2, . . . , 𝑛 (13)
𝜕𝑇
𝑖

When applied to Eq. (12) above, this will lead to a set of simultaneous linear
algebraic equations in the nodal d.o.f., in general designated as [K]{} = {f}. The
contributions of the Mi element to these system level equations can be obtained by
applying Eq. (13) to the element level functional as given by Eq. (11). Thus we have
the element level contributions as

−1 𝑇1 𝑞0 ℓ/2 𝑄
𝑘
[1 ]{ } = { } + { 1} (14)
ℓ −1 1 𝑇2 𝑞0 ℓ/2 𝑄2

We observe that these contributions are identical to those obtained by the

weak form of the Galerkin weighted residual statement, starting from the governing
differential equation of the problem

Meaning of Finite Element Equations

Let us look at the element level equations for a bar element. We have, from
Eq. (8), with qo = 0,

𝐴𝐸
[1 −1 𝑢 1 𝐹
]{ } = { 1} (15)
ℓ −1 1 𝑢2 𝐹2

ie [𝐾]{𝑢} = {𝐹}

The individual elements of the element stiffness matrix have a direct physical
meaning. Suppose we consider a deformation state for the element wherein we have
u1 =1 and u2 = 0. From elementary mechanics of materials we know that the force
required to cause such a deformation will be AE/L to be applied at node 1 and, since
node 2 is to be held at rest, an equal and opposite reaction force develops at node
2. From the above equation if we solve for the forces F\, F2 for this prescribed
displacement, we get the same result. We observe that the elements of the first
column of the stiffness matrix are identical to these forces. Similarly, if we consider a
deformation state wherein u2 = 1 and u1 = 0, the forces required will be given in the
second column of the stiffness matrix. Thus elements of each column of a
stiffness matrix actually represent the forces required to cause a certain
deformation pattern— the ith column standing for a deformation pattern,
where the ith d.o.f. is given unit displacement (translation;] or rotational)
and all other d.o.f. are held zero.

We can actually attempt to derive the finite element equations based on this
direct physical correlation, called the Direct method of formulation of finite
element equations. We will illustrate the usefulness of such an approach for a beam
finite element. Our beam element is a line element having two nodes just like the
bar element, but the d.o.f. permitted at each node are the transverse displacement
v, and the slope (rotational d.o.f.) dv/dx.

We thus expect to derive the beam element equations as

𝑘11 𝑘12 𝑘13 𝑘14 𝑣1 𝐹1

𝑘21 𝑘22 𝑘23 𝜃1 𝑀
[ ] { 𝑣 } = { 1} (16)
𝑘24 2 𝐹2
𝑘31 𝑘32 𝑘33 𝜃2 𝑀2
𝑘34
𝑘41 𝑘42 𝑘43 𝑘44
 Based on the physical interpretation of the columns of the stiffness matrix
above, we observe that the columns of the stiffness matrix should correspond to the
forces (moments) required to cause the deformation states shown in Figure.
For example, the first deformation state requires that the beam act as a clamped-
free beam (x = L being the clamped end) with a force F and a moment M applied at
the end x - 0 such that it undergoes unit transverse deformation but zero rotation.
From elementary mechanics of materials involving the Euler-Bernoulli beam theory,
we find the forces (moments) for causing these deformation states as shown in
Figure. Thus the beam element stiffness matrix derived using the "direct" method is
found to be

12𝐸𝐼
ℓ3
𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐
6𝐸𝐼 4𝐸𝐼
[𝑘] = ℓ2 ℓ
− 6𝐸𝐼 12𝐸𝐼 (17)
−12𝐸 𝐼
3 2 3
ℓ ℓ ℓ
6𝐸𝐼 6𝐸𝐼 −6𝐸𝐼 4𝐸𝐼
[ ℓ2 ℓ ℓ2 ℓ ]

We will derive the beam element equations from the PSTP and show that we
indeed get the same equations. Even though the direct physical meaning of the
individual columns of the stiffness matrix provides useful insight, its use in actually
deriving the element matrices is limited to only simple cases. When we discuss two-
and three-dimensional finite elements and higher order elements with internal
nodes, etc., it is not possible to derive the stiffness matrices based on this
interpretation alone. We need a systematic procedure based on either the Galerkin
WR or the Variational method, to formulate these finite elements. We have
demonstrated that both differential and functional forms yield identical finite element
equations.

General Procedure for Finite Element Analysis

Step 1

➢ Using the governing differential equation, write down the weighted

residual form. Perform integration by parts sufficient number of times
and develop the weak form of the differential equation. Rewrite the
weak form as a summation over n elements.

OR
 ➢ Write the functional representation as a summation over n elements.

Step 2

➢ Define the finite element, i.e. the geometry of the element, its nodes,
and nodal d.o.f.

➢ Derive the shape or interpolation functions.

Step 3

➢ Compute the element level equations by substituting these shape

functions in the weak form or the functional representation.

Step 4

➢ Discretise the given domain into a finite element mesh. For a given
topology of finite element mesh, build up the system of equations by
assembling together element level equations.

➢ Substitute the prescribed boundary conditions and solve for the

unknowns.

SUMMARY

We have discussed the formulation of the finite element equations starting

from either the weak form of the Galerkin weighted residual statement or the
condition of stationarity for a functional and demonstrated that they yield identical
results. For long, it has been common practice to use energy methods in structural
mechanics and, thus, for most problems of structural mechanics, it may be
convenient to use the PSTP and formulate the necessary finite element equations.
The variational principle dealing with the strain energy and the work potential of
external forces has direct physical meaning for the structural analyst. However,
given the functional for any problem (such as the heat transfer example above), we
will be able to develop the necessary finite element equations. The functional for
some problems, such as some problems of heat transfer and fluid flow, may not
have any direct physical meaning, and so it may be more convenient to state the
problem in terms of the governing differential equations. It must be appreciated that
the finite element equations can be derived systematically with equal ease, starting
from either form of representation of governing equations—differential or functional.
In the subsequent chapters of the text, we will use both these forms interchangeably
for developing the necessary finite elements.

One-dimensional Finite Element Analysis

We have already discussed three different methods of deriving finite element

equations, viz., direct method, weak form, and stationarity of a functional. The direct
method was useful for very simple structural elements only, and we illustrated the
equivalence of other two methods with the example of a one-dimensional bar
element and heat transfer element. In this chapter we will study the finite element
analysis of one-dimensional problems in greater detail. We will use the stationarity of
a functional approach (i.e., the Principle of Stationary Total Potential) to develop the
finite element equations for structural mechanics problems. We will first recast the
earlier procedure in a generic form and then work out more example problems using
the bar element. Subsequently, we will discuss the beam and frame elements. For
heat transfer problems, we will use the weak form equation.

General Form of the Total Potential for 1-d

A one-dimensional structure (such as a bar or a beam) can, in general, be

subjected to distributed loading in the form of body forces (such as gravity), forces
distributed over a surface (such as a pressure) or concentrated forces lumped at
certain points. The first two types of forces for one-dimensional problems can be
reduced to force per unit length, q. It may also be subjected to initial strains and
stresses (such as those due to pre-loading, and thermal loading). Under the action
of all these loads, the structure, when properly supported (so as to prevent rigid
body motion), undergoes deformation and stores internal strain energy. We will now
develop an expression for the total potential of such a structure assuming linear
elastic isotropic material obeying Hooke's law. We also assume that all loads and
other parameters are time-invariant.

𝑆𝑡𝑟𝑒𝑠𝑠𝜎𝑥 = 𝐸(𝜀𝑥 − 𝜀𝑥 0 ) + 𝜎𝑥 0 (1)

where  x0 is the initial strain and x0 is the initial stress. This stress (x ) does work in
causing an increment in strain (𝑑𝜀 𝑥 ), which is further stored in the elastic body as an
increment in strain energy. Consider a cube of unit edge length in the body, i.e. unit
volume of material. The increment in strain energy in this unit volume, dU is given
by the incremental work which equals the product of force (F) and incremental
displacement (d). Here, "" stands for the generic displacement e.g. axial
displacement u for a bar. Thus

𝑑𝑈 = [𝐹][𝑑𝛼] = [(𝜎𝑥 )(1)(1)][(𝑑𝜀𝑥 )(1)] = 𝛼𝑥 𝑑𝜀𝑥 (2)

Substituting for stress from Eq. (1) and integrating with respect to strain
yields the strain energy in this unit volume of material. This, when integrated over
the whole structure, yields the total strain energy. Thus,
1
𝑈 = [∫ ( 2) 𝐸𝜀 𝑥2 ] 𝐴𝑑𝑥 − ∫ 𝜀𝑥 𝐸𝜀 𝑥 0 𝐴𝑑𝑥 + ∫ 𝜀𝑥 (3)
𝜎𝑥 0 𝐴𝑑𝑥

The potential V of an external force in any generalised coordinate a is

defined, by convention, in such a way that the generalised force in that coordinate is
given as F= -(∂V/ ∂α). The forces on a one-dimensional structure may be
distributed (q per unit length) or concentrated (P). Therefore, the potential of
external forces is given by

𝑉 = −(∫ 𝛼𝑞𝑑𝑥 + ∑ 𝛼1 𝑝1 ) (4)

where Pt is the concentrated force acting at a point i, a,- is its displacement

and the summation is taken over all the points where concentrated forces are
present. The total potential of the structure is obtained as

∏𝑝 = 𝑈 + (5)
𝑉
We will use this generic form in our further discussion while attempting to
derive the finite element equations in their generic form
Generic Form of Finite Element Equations

Let the displacement a at any point P within the element be interpolated from
the nodal values of displacement {  } e using the shape functions N as follows:

𝛼 = [𝑁]{𝛿}𝑒 (6)

The interpretation of Eq. (6) is as follows: For a one-dimensional problem, we

have one possible displacement at each point (axial displacement u or transverse
displacement v), and { } e is a column vector of nodal d.o.f. of the element and is of
size equal to the product of number of nodes per element (NNOEL) and the d.o.f.
per node (NFREE). For a bar element

NNOEL =2
NFREE =1
𝑢
{𝛿} 𝑒 = { 1 }
𝑢2 (7)

For a beam element

NNOEL =2
NFREE =1

𝑣1
𝑑𝑣 /𝑑𝑥
{𝛿} 𝑒 = { 1 (8)
𝑣2 }
𝑑𝑣 2 /𝑑𝑥

The shape functions [N] represent the interpolation formulae used to obtain
the displacement field within the element, and the size of the matrix [N] will be
consistent with the size of { } e . For example, for the bar element, [N] is given by
 [𝑁] = [(1 − 𝑥 ) ( 𝑥 )] (9)
ℓ ℓ

The nonzero strain at a point P within the element (i.e., ex) is given in terms
of the derivatives of the displacement field and is again represented in terms of
nodal d.o.f. as

𝜀𝑥 = [𝜕](𝛼) = [𝜕](𝑁){𝛿}𝑒 = (10)

(𝐵){𝛿}𝑒

where [𝜕] is a differential operator and [B]=[∂] [N], representing the derivatives of
shape

functions. For a bar element, for example, [𝜕] is simply dldx and [−1 1 ]
ℓ ℓ

We can now rewrite the generic expression for the total potential as follows from Eq.
(3). For an element,
𝑇 𝑇
𝑈 𝑒 = ∫ 1 {𝛿}𝑒 [𝐵 ]𝑇 [𝐵 ]{𝛿}𝑒 𝐸𝐴𝑑𝑥 − ∫{𝛿} 𝑒 [𝐵]𝑇 𝑇
+ ∫{𝛿} 𝑒 [𝐵 ]𝑇 𝜎𝑥 0𝐴𝑑𝑥 (11)
2 𝑥 0 𝐴𝑑𝑥
𝐸𝜀

Again from Eq. (4),

𝑇 𝑇
𝑉 𝑒 = − ∫{𝛿} 𝑒 [𝑁]𝑇 𝑞𝑑𝑥 − ∑{𝛿} 𝑒 [𝑁]𝑇𝑖 (12)
𝑃𝑖
Thus,

𝑒 1 𝑇 𝑇
∏𝑝 = ∫ 2 {𝛿} 𝑒 [𝐵 ]𝑇 [𝐵 ]{𝛿}𝑒 𝐸𝐴𝑑𝑥 − ∫{𝛿}𝑒 [𝐵]𝑇
𝑇 𝑇 𝑇 𝐸𝜀 𝑥 0 𝐴𝑑𝑥
+ ∫{𝛿} 𝑒 [𝐵 ]𝑇 𝜎𝑥 0 𝐴𝑑𝑥 − ∫{𝛿}𝑒 [𝑁]𝑇 𝑞𝑑𝑥 − ∑{𝛿} 𝑒 [𝑁𝑇]𝑖 (13)
𝑃𝑖
where [N]i- is the matrix of shape function values at the location i.

Since { } e is a vector of nodal d.o.f., and is not a function of the spatial coordinate
x, it can be taken outside the integration. Thus we get

1 𝑇 𝑇 𝑇
∏𝑒𝑝 = 2 {𝛿} 𝑒 ∫([𝐵 ]𝑇 [𝐵 ]𝐸𝐴𝑑𝑥) {𝛿}𝑒 − {𝛿} 𝑒 ∫[𝐵 ]𝑇 𝐸𝜀𝑥 0 𝐴𝑑𝑥 + {𝛿}𝑒 ∫[𝐵 ] 𝑇 𝜎𝑥 0 𝐴𝑑𝑥

(14)

𝑇 𝑇 𝑇
−{𝛿} 𝑒 [𝑁]𝑇 𝑞𝑑𝑥 − {𝛿} 𝑒 ∑{𝛿} 𝑒 [𝑁]𝑇𝑖
𝑃𝑖
Defining the element stiffness matrix [k]e and the load vector { f } e as
 [𝐾]𝑒 = ∫{𝐵 } 𝑇 [𝐵 ] 𝐸𝐴𝑑𝑥 (15)

{𝑓} 𝑒 ∫[𝐵 ]𝑇 [𝑁]𝑇 𝐸𝜀 𝑥 0 𝐴𝑑𝑥 − ∫[𝐵 ]𝑇 𝜎𝑥 0 𝐴𝑑𝑥 + ∫[𝑁]𝑇 𝑞𝑑𝑥 + ∑[𝑁]𝑇𝑖

𝑖
(16)
𝑃
we can rewrite Eq. (4.14) in the form

1 𝑇 𝑇
∏𝑒𝑝 = 2 {𝛿} 𝑒 [𝐾]𝑒 {𝛿} 𝑒 − {𝛿}𝑒 {𝑓}𝑒 (17)

𝑒
Potential 𝜋𝑝 being a scalar quantity, the total potential for a mesh of finite
elements can be obtained by simple addition. The total potential of the system is
given as
1
∏𝑝 = 𝑁𝑂𝐸𝐿𝐸𝑀
∏𝑒𝑝 = 2 {𝛿} 𝑇 [𝐾]{𝛿} − {𝛿} 𝑇 {𝐹 } (18)
∑1
wherein the global stiffness matrix of the structure [K] and the global load vector
{F} are obtained as

[𝐾] = ∑𝑁𝑂𝐸𝐿𝐸𝑀
𝑛 =1
[ 𝑘] 𝑒 (19)

{𝐹} = ∑𝑁𝑂𝐸𝐿𝐸𝑀
1
{𝑓} 𝑒 (20)

"NOELEM" represents the number of elements in the mesh and {5} contains all the
nodal d.o.f. variables for the entire finite element mesh. The summations indicating
assembly imply that the individual element matrices have been appropriately placed
in the global matrices following the standard procedure of assembly. Using the PSTP,
we set the total potential stationary with respect to small variations in the nodal
d.o.f., i.e.*,

𝜕 ∏𝑝
𝜕 {𝛿 } =0 (21)

Thus the system level equations are given by

[𝐾]{𝛿} = {𝐹} (22)

We will now illustrate the details of the procedure through the familiar bar
element and then derive the element matrices for a beam element. We will follow
this generic notation in our derivation of two-dimensional elements also. Thus the
reader is advised to become fully conversant with this notation, which wil be extensively
used in the rest of the book.

The Linear Bar Finite Element

We recall that our linear bar element is a one-dimensional line element with
two nodes. Each point P within the element is permitted to move only along the axis
of the element and the displacement of the entire cross-section is assumed to be
same. Therefore, we have

Displacement field {𝛼} = 𝑢 (23)

𝑢1
Nodal d.o.f. vector {𝛿} 𝑒 = { }
𝑢2 (24)

𝑥
Shape functions [𝑁] = [𝑁1 𝑁2 ] = [(1 − ℓ𝑥) ℓ ] (25)

𝑑𝑢
Strain {𝜀} = 𝜀𝑥 = (26)
𝑑𝑥

thus differential operator [𝜕]is simply d/dx. Hence

[𝐵] = [𝑑 𝑁 1 𝑑𝑁2
] = [−1
1 (27)
]
𝑑𝑥 𝑑𝑥 ℓ ℓ

Thus the element stiffness matrix is obtained as

𝑇
[𝑘]𝑒 = ∫ℓ [𝐵] [𝐵]𝐸𝐴𝑑𝑥 (28)
0

−1/ℓ
] [−1/ℓ1/ℓ]𝐸𝐴𝑑𝑥
𝑙
= ∫0 [
1/ℓ (29)

= 𝐴𝐸 [ 1 −1
] (30)
ℓ −1 1

which is the same as the result obtained earlier. In the examples that follow, we will
illustrate the use of the bar element in several situations.

Example . A bar subjected to self-weight. Consider a vertically hanging rod (Figure

) of length L, uniform cross-sectional area A, density p, and Young's modulus E. We
wish to find the state of deformation and stress in the rod subjected to gravity load,
using the bar element just described. Through this example, we wish to bring out
several insights into the nature of finite element approximate solutions.

For a bar element, the stiffness matrix is given by Eq. (30). Thus,

[𝐾]𝑒 = 𝐴𝐸
[ 1 −1] (31)
𝑙 −1 1

For the present example, the body force (i.e., self-weight) expressed as "q"
per unit length is given by (Ag). In the absence of any other force, the element
nodal force vector is given by Eq. (16):

{𝑓} 𝑒 = ∫[𝑁]𝑇 𝑞𝑑𝑥 = ∫1 1 − 𝑥/𝑙] (𝑝𝑔)(𝐴)𝑑𝑥

[ 0 𝑥/𝑙
𝜌𝐴𝑙𝑔/2}
= { 𝜌𝐴𝑙𝑔/2 (33)

Since (A ℓg) represents the weight of the element, the equivalent nodal force
vector just obtained is thus simply half the weight of the element distributed at
either node. We will now use one, two, four, eight and 16 finite elements and, in
each case, compare our solution with the exact solution. The exact solution is readily
verified to be given by
2
Displacement 𝑢(𝑥) = 𝜌 𝑔 (𝐿𝑥 − 𝑥 ) (34)
𝐸 2

stress (x)=g(L-x) (35)

We observe that the exact solution corresponds to linear variation of stress and
quadratic variation of displacement, whereas within each bar element we permit only
linear displacement field (Eq. 25) and, consequently, constant stress. Thus we
expect our solution to be poor when the number of elements is very small and
should become more and more accurate as we increase the number of elements.

One-element solution. We use a single bar element to model the entire length of
the rod as shown in Figure(a). We have

𝐴𝐸
[1 −1 𝑢 1 𝐹
] {𝑢 } = { 1 } (36)
1 −1 1 2 𝐹2

Using ℓ = L and substituting for the boundary condition u1 = 0 and for the
known force F2 = AgL/2, we get
𝜌𝐴𝐿𝑔 𝐿
𝑢2 = ( 2
) 𝐴𝐸
𝜌 𝑔 𝐿=2
2𝐸
(37)

Thus,

𝒖(𝒙) = ( 𝝆𝒈𝑳) 𝒙 (38)

𝟐𝑬

𝝆𝒈𝑳
𝑺𝒕𝒓𝒂𝒊𝒏𝜺𝒙 = 𝟐𝑬
(39)

𝝆𝒈𝑳
𝑺𝒕𝒓𝒆𝒔𝒔𝝈𝒙 = 𝑬𝜺 𝒙 = 𝟐
(40)
 Two-element solution. Let us discretise the bar into two equal length
elements ( ℓ = V/2) as shown in Figure (b). For the first element, we have
(1)
𝑢 𝑓
[ 1 −1] { 𝑢 1 } = { 1(1) }
𝐴𝐸
(41)
𝐿 /2 −1 1 2 𝑓2

For the second element,

(2)
𝑢 𝑓
[ 1 −1] { 𝑢 1 } = { 1(2) }
𝐴𝐸
(42)
𝐿 /2 −1 1 2 𝑓2

Assembling the two individual element level equations, we obtain

(1)
1 −1 0 𝑢1 𝑓1
𝐴𝐸
[−1 1 + 1 −1] {𝑢 2 } = { 𝑓1 + 𝑓1(2) }
(1)
(43)
𝐿 /2
0 −1 1 𝑢3
𝑓2(2)

Substituting the boundary condition, u i= 0, and the known forces

(1) 𝜌𝐴𝑔𝐿/2 (2) 𝜌𝐴𝑔𝐿 /2 (2) 𝜌𝐴𝑔𝐿 /2
𝑓 2 2 2
2 , 𝑓1 , 𝑓2
= = =
We get

𝑢 𝜌𝐴𝑔𝐿/2
𝐴𝐸
[ 2 −1] { 𝑢 2 } = { } (44)
𝐿 /2 −1 1 2 𝜌𝐴𝑔𝐿/4

Solving, we obtain

𝑢2 𝜌𝑔𝐿2
{ 𝑢 } = 3/2 { } (45)
3 4𝐸 2

The element level strains and stresses are computed as

𝑢 (1) −1 1 0 3 𝜌𝑔𝐿
𝜀𝑥1 = [𝐵] {𝑢 1 } = [ 𝐿/2 𝐿 /2 ] { 3𝜌 𝑔 𝐿2 } = 4 𝐸 (46)
2 8𝐸

2
𝑢 1 (2) = [ −1 1 ] {3𝜌𝑔𝐿 /8𝐸 } = 1 𝜌 𝑔 𝐿
𝜀𝑥2 = [𝐵 ] {
𝑢2
} (47)
𝐿 /2 𝐿/2 𝜌𝑔𝐿2 /2𝐸 4 𝐸
 3𝜌 𝑔
𝜎𝑥(1) = 𝐸𝜀𝑥(1) = 4
(48)

𝜎 (2) =𝜌 𝑔𝐸𝜀 (2)

= (49)
𝐿
𝑥 𝑥 4

Four-element solution. We further subdivide each of the two elements used

earlier by half Figure (c), i.e., the total rod is discretised into four elements, each of
length (L/4). For each element, we have
𝑢 𝐹
𝐴𝐸
[ 1 −1] { 𝑢 1 } = { 1 } (50)
𝐿 /4 −1 1 2 𝐹2

Assembling all the individual element matrices and substituting for the

boundary condition and the known forces, we obtain

1+ 1 −1 0 0 2 1/2 + 1/2

(51)
𝐴𝐸
[ −1 1 + 1 −1 0] 3
}={
1/2 + 1/2 𝜌𝐴𝐿𝑔
}
𝐿 /4
00 −1
0 1 +
−1 1 −1
1 𝑢5 1/2
{𝑢 1/2 + 1/2 4
4

Solving, we get

2 2
𝑢 = ( 7 ) 𝜌 𝑔 𝐿 , 𝑢 = ( 12 ) 𝜌 𝑔 𝐿 ,
2 16 2𝐸 3 16 2𝐸
15 𝜌𝑔𝐿2
𝑢 4 = ( 16
2
) 2𝐸
, 𝑢 5 = 𝜌2𝐸
𝑔𝐿 (52)

The corresponding stresses in individual elements are obtained as

𝜎 (1) = ( 7 ) 𝜌𝑔𝐿, 𝜎 (2) = ( 5 ) 𝜌𝑔𝐿,

32 32
3 1
𝜎 (3) =( ) 𝜌𝑔𝐿, 𝜎 (4) = ( ) 𝜌𝑔𝐿 (53)
32 32

We observe some very significant points—for example, no matter how few the
elements we take are, our solution for the displacement at the nodes is exact! Even
with just one element (i.e., a single straight-line variation of displacement within the
entire rod), we get the exact value for tip displacement, whereas the exact solution
is clearly a quadratic variation of displacement However, we must note here that the
nodal solution is not exact for most finite element solutions.
 A point of immediate interest is "With nodal displacements being exact, how
could the stresses be inaccurate?". The calculation of stress depends not just on the
nodal values but on the variation of displacement within the element. Since the
element used permits only linear variation of displacement and the exact solution is
quadratic, we expect differences in the stress result. To illustrate this, let us compute
the value of displacement at an interior point as predicted by our finite element
solution. Let us choose to compute the displacement at x = 0.33Z, which is not a
node in any of the finite element solutions we obtained. For this purpose, we first
determine on which element this point P (corresponding to x = 0.33Z) lies, and use
the regular interpolation formula, i.e.,
𝑥𝑝 𝑥𝑝 𝑢𝑖
𝑢 𝑝 = [(1 − ℓ
) ℓ
] (4.54)
{𝑢 }
𝑗

where u and 𝑢𝑗 are the nodal displacements of that element and xp in this equation
refers to the distance from the ith node of the element to the point P (recall that we
used a local coordinate system for each element with its local origin at the ith node,

A summary of our observations based on the above discussion is now given.

➢ Depending on the variation of the field variable (displacement in this

case) and the shape functions used in the finite element, we may require
many elements to model the field accurately, especially at the interior
points.

➢ The field variable itself may be reasonably well captured, but the error in
the derivatives (e.g. stress here) could be considerable.

➢ Stress averaging will be necessary to get a more realistic value of stress

at a node, jointly shared by many elements.

The bar element is capable of modelling only a linear variation of

displacement, and hence we require many such elements to model a physical system
accurately. We can improve this element by permitting quadratic variation of
displacement within the element. We will now discuss a typical higher order element,
viz., quadratic bar element. We cannot achieve this (i.e., a complete quadratic
polynomial) with just two nodes and so we have to add another node to the
element.

It must be observed that we could have used two nodes only but permitted
an incomplete quadratic polynomial shape function, e.g. u(x) - c0 + c2x2. For such
an element,  x - du/dx - 2c2x. Thus, strain and stress will always be zero at one end
of each element (at x = 0)!

The Quadratic Bar Element

Determination of Shape Functions

The three noded quadratic bar element is shown in Figure.

We number the two end nodes as 1 and 2 and the middle node is given
number 3. The displacement at any point within the element is now interpolated
from the three nodal displacements using the shape functions as follows:

u(x) = Nx(x)ux + N2 (x)u2 + N3 (x)u3 (55)

The shape functions Ni vary quadratic ally within the element and we now
discuss a systematic procedure to derive such shape functions. Let u(x) be given by
the complete quadratic polynomial
u(x) = c0 + c1 x + c2 x 2 (56)

We know that u(0) = u1 u ( ℓ) = u2 and u (ℓ / 2) = u3 . Hence, from Eq. (4.56),

𝑢 1 = 𝑢 0 ,𝑢 2 = 𝑐0 + 𝑐1 ℓ + 𝑐2 ℓ
2

ℓ ℓ 2
𝑢 3 = 𝑐0 + 𝑐1 ( 2 ) + 𝑐2 ( 2 ) (57)

Solving for ci we obtain

 𝑐0 = 𝑢 1
𝑐1 = (4𝑢3 − 𝑢 2 − 3𝑢 1 )/ℓ
𝑐2 = (2𝑢1 − 2𝑢 2 − 4𝑢 1 )/ℓ
2
(58)

Thus we have

4𝑢 3 −𝑢 2 −3𝑢 1 2𝑢 1 −2𝑢 2 −4𝑢 1

𝑢(𝑥) = 𝑢 1 + ( ℓ
)𝑥+( ℓ2
) 𝑥2 (59)

Rearranging the terms we get

2
2𝑥 2 4𝑥 2
𝑢 (𝑥) = 𝑢 + (1 − 3𝑥 + 2𝑥 ) + 𝑢 (− 𝑥 + ) + 𝑢3 (
4𝑥
+ ) (60)
1 ℓ ℓ2 2 ℓ ℓ2 ℓ ℓ2

Comparing Eqs. (4.55) and (4.60), we obtain

2
2𝑥 2 4𝑥 4𝑥 2
𝑁 = 1 − 3𝑥 + 2𝑥 , 𝑁 − 𝑥 + , 𝑁3 = + (61)
1 ℓ ℓ2 2 ℓ ℓ2 ℓ ℓ2

This is a general technique for determining the necessary shape functions Nt

We first assume a general polynomial field (Eq. (56)). Substituting the nodal
coordinates and requiring that the expression must reduce to nodal d.o.f, we
generate sufficient number of equations (Eq. (57)) to solve for coefficients c,- in
terms of nodal d.o.f. We then substitute for c{ (Eq. (59)) and rearrange the terms to
obtain the desired shape functions (Eq. (60)). Later on we will discuss other
strategies for effective generation of the desired shape function.

Element Matrices

Given the shape functions, we get the strain-displacement relation matrix [B]
as
𝑑𝑢 𝑑
𝜀𝑥 = 𝑑 𝑥 = 𝑑 𝑥 [𝑁]{𝛿}𝑒 = 𝑒
(62)
[𝐵 ]{𝛿}

Where

[𝐵] = [ 𝑑 𝑁 1 𝑑 𝑁 2 𝑑 𝑁 3 ] = [ 2 (2𝑥 − 3ℓ) 2 (2𝑥 − ℓ ) −4 (2𝑥 −

(63)
ℓ)] 𝑑𝑥 𝑑𝑥 𝑙2 2 ℓ2 2 ℓ2
𝑑𝑥

Hence we have the element stiffness matrix given by from Eq. (15) and Eq.
(63):
 [𝐾]𝑒 = ∫[𝐵]𝑇 (64)
[𝐵]𝐸𝐴𝑑𝑥
2 3ℓ
ℓ2
2𝑥 2
2𝑥 − ℓ = [ 2 (2𝑥 − 3ℓ) 2 (2𝑥 − ℓ ) −4 (2𝑥 − ℓ)] 𝐸𝐴𝑑𝑥
1 2
∫0 (65)
ℓ 2 2 ℓ2 2 ℓ2 2 ℓ2
−4
[ ℓ2 (2𝑥 − ℓ)]

On evaluating all the integrals, we obtain the element stiffness matrix as

7 1 −8
[𝐾 ]𝑒 = 𝐴𝐸 [ 1
3ℓ
7 −8] (66)
−8 −8 16

Example 2. A vertically hanging rod. For the present example, the body force due
to gravity expressed as force q per unit length is given by (pAg). Thus, the element
nodal force vector (in the absence of any other force) is given by
3𝑥 2𝑥 2
1 − ℓ + ℓ2
𝑇
{𝑓}𝑒 = ∫0 [𝑁] 𝑞𝑑𝑥 = ∫ℓ 0
1 −𝑥 2𝑥 2
+ 2 𝜌𝐴𝑔𝑑𝑥 (67)
ℓ ℓ
4𝑥 4𝑥 2

[ ℓ − ℓ2 ]

1/6
𝜌𝐴𝑔𝑙 = {1/6} (68)
4/6

Using just one element spanning the entire length of the rod, as shown in
Figure , we obtain the equations
 𝑢2 { 1/6
7 16 ] { 𝑢 3 } = 4/6 } 𝜌𝐴𝑔𝑙
−8
𝐴𝐸
3ℓ = [−8 (69)

where the boundary condition u 1 = 0, and the known forces have already been
substituted and ℓ = L.

Solving, we get
𝜌𝑔𝐿 3𝜌 𝑔 𝐿
𝑢 2 = 2 2𝐸 , 𝑢 3 = 2
8𝐸
(70)

Substituting these equations in Eq. (4.60), we obtain

𝑢(𝑥) = ( −𝑥 + 2 𝑥 2 ) 𝜌 𝑔 ℓ + (4𝑥 − 4𝑥 ) ( 3𝜌 𝑔 ℓ )
2 2 2

ℓ ℓ2 2𝐸 ℓ ℓ2 8𝐸
𝜌 𝑔 ℓ2
= ( 2𝑥 + 𝑥 2) = 𝜌 𝑔 (𝐿𝑥 − 𝑥 2) (71)
2𝐸 ℓ ℓ2 𝐸 2

Comparing with Eq. (34), we observe that our one-element solution (ℓ = V)

tallies with the exact solution completely (i.e. not just at the nodes but interior
points also). It is left as an exercise for the reader to show that the stresses also
match the exact solution.

We thus observe that we could get the exact solution with just one quadratic
bar element where as we needed several linear bar elements to get a reasonably
accurate solution.

Conclusion

A few higher order elements are far superior to several lower order elements.
 Example. A bar subjected to thermal stresses. Consider a uniform rod of
length L, area of cross-section A , Young’s modulus E, and the coefficient of thermal
expansion  heated uniformly through a temperature rise of T °C. Determine the
thermal stresses in the bar for the end conditions shown in Figure (a).

This example problem is indicative of how problems of initial strain { 0} are

solved. The initial strain in this example is due to heating of the bar. It could as well
be due to any other reasons such as bolt pretensioning. While we work out this
example on a bar element, the general methodology remains identical even when
we consider situations involving more compels elements.

Before we analyze the finite element solution to the problem, we will briefly
review the concept of thermal stresses. When a rod is heated through a temperature
rise T, the rod would tend to expand by an amount ( LT). This corresponds to an

initial strain  0 = ( T). If the support restraints on the rod are such that this
expansion is possible, then the rod freely undergoes the thermal expansion ( LT),
and thus the mechanical strain in the rod is given by =LT/L=(T) This is the case
in Figure (a). As there are no forces restraining this thermal expansion, there will be
no internal stresses developed. This is reflected in the material constitutive equation
given by

𝜎 = 𝐸(𝜀 − 𝜀0 ) = 𝐸 (𝛼𝑇 − 𝛼𝑇) = 0 (1)

However, as in Figure 4.(b), if the rod is entirely prevented from undergoing

the free thermal expansion as dictated by the initial strain  0 Then the mechanical
strain in the rod is =0 Thus the rod is subjected to a compressive stress as

𝜎 = 𝐸(𝜀 − 𝜀0 ) = 𝐸(0 − 𝛼𝑇) = −𝐸𝛼𝑇

 This can be visualized as shown in Figure (c), i.e., the rod can be imagined to
have expanded LT), and the support develops
reaction forces which push the rod back (i.e. compress) all the way to its original
position. If, in general, the support is a little distance away from the end of the rod
as shown in Figure (d) so that the rod is allowed partial free expansion, then internal
stresses arise only corresponding to the part of initial strain that was not actually
realized. The structure depicted in Figure (d) is for the purpose of illustration only. In
real structures, when a part of the structure undergoes thermal expansion, it will be
against the elasticity of the other parts. Therefore, only a part of the total thermal
expansion will be permitted.

Thus our general method of modeling initial strain will be as follows: we first
determine the mechanical loads corresponding to the initial strain, i.e. the loads that
would be required in order to cause a mechanical strain equal to the initial strain.
Under these loads, we then find the actual deformation of the structure consistent
with its support restraint conditions. From these actual deformations, we find the
actual mechanical strain. The difference between the actual mechanical strain and
the originally imposed initial strain leads to the internal stresses.

From our general expression for nodal force vector, the nodal loads due to
initial strain are given by

{𝑓} 𝑒 = ∫ [𝐵 ]𝑇
𝑣 𝑥 0 𝐴𝑑𝑥
𝐸𝜀
For the linear bar element

[𝐵 ] [−1 1
]
𝑙 𝑙
𝜀𝑥 0 = 𝛼𝑇

Thus the nodal force vector corresponding to initial strain is given by

{𝑓} 𝑒 = ∫1 [ −1/𝑙] 𝛼𝐸𝑇𝐴𝑑𝑥 = [−𝐴𝐸𝛼𝑇 ]

0 1/𝑙 𝐴𝐸𝛼𝑇

We will use, for the purpose of illustration, two linear bar elements, each of
length (L/2) for modeling the entire rod. For element 1, we have, in the absence of
any other load,

𝑢 (1) 1(1)
𝐴𝐸
[ 1 −1] { 𝑢 1 } = {𝑓 (1) } = {−𝐴𝐸𝛼𝑇 } (2)
𝐿 /2 −1 1 2
𝑓2 𝐴𝐸𝛼𝑇

For element 2, we have

(2)
𝑢 (2) 𝑓
𝐴𝐸
[ 1 −1] { 𝑢 1 } = { 1(2) } = {−𝐴𝐸𝛼𝑇} (3)
𝐿 /2 −1 1 2 𝑓2 𝐴𝐸𝛼𝑇

It is readily seen that these forces, acting on a free bar element, will cause

strains corresponding to the initial strain  0 =  T. Assembling the element matrices,

we get
(1)
1 −1 0 𝑢1 𝑓1
𝐴𝐸
[−1 1 + 1 −1] {𝑢 2 } = { 𝑓2 + 𝑓1(2) }
(1)
𝐿 /2
0 −1 1 𝑢3
𝑓2(2)

If the end x = L is free, then we have, with u1 = 0,

𝑢
𝐴𝐸
[ 2 −1] { 𝑢 2 } = { 𝐴𝐸𝛼𝑇 − 𝐴𝐸𝛼𝑇 } = { 0 }
𝐿 /2 −1 1 3 𝐴𝐸𝛼𝑇 𝐴𝐸𝛼𝑇

Solving, we get

𝑢 2 = 𝐿𝛼𝑇/2, 𝑢 3 = 𝐿𝛼𝑇

For element 1, the actual mechanical strain is

{𝜀} = [𝐵]{𝛿}𝑒 [−1 −1 0

]{ } = 𝛼𝑇
𝑙 𝑙 𝐴𝛼𝑇/2

The internal thermal stress in element 1 is given by

{𝛼} = 𝐸(𝜀 − 𝜀0 ) = 0

Similarly, for element 2, we can show that the internal stresses will be zero. It
is left as an exercise for the reader to show that internal stresses equal to (− T)
will be developed if the end of the rod x = L is also clamped.
Summary: Finite Element Modeling of Initial Strains

Step 1: Compute the element nodal force vector corresponding to the

initial strain { 0}.

Step 2: Solve for the deformations of the structure when subjected to the
“loads due to initial strain” along with any other loads on the structure,
consistent with the boundary conditions.

Step 3: From the deformations obtained in Step 2 above, evaluate the element
level strains {} using {} = [B] {}e.

Step 4: Evaluate the internal stresses developed using the constitutive

equation of the material, viz., {} = E ({} − { o}).

Example. A three-bar truss network. Let us consider the statically indeterminate

problem of three trusses shown in Figure. Unlike all the previous examples discussed
so far, in this case the members of the structure are not all in one line. We use this
example to bring out the issue of “transforming” element level equations in local
coordinates into a common global reference frame before assembly. Since the
structure is subjected to only a concentrated tip load, we expect the deflection to
vary linearly within each member, and hence we expect that one linear bar element
would suffice for each truss member. Figure (a) shows a three-element model of the
structure indicating the element numbers and the node numbers. We also show a
generic truss element in Figure (b), indicating its own local axis and the nodal d.o.f.
along its axis, which are not necessarily along the global axes.

For each truss element in its own local coordinate frame, we have the force-
deflection relations, viz.

𝐴𝐸
[1 −1 𝑢 𝑖 𝐹𝑖
] {𝑢 } =
1 −1 1 𝑗
{𝐹 }
𝑗

In our particular example, the length of the central element is L while that of
the inclined rods is L/cos. Thus the individual element equations will be essentially
similar to above equation except for the appropriate value of element length.
Considering the components of deflection as shown in Figure (b), we can write

𝑢 𝑖 = 𝑢 𝑖 𝑋 𝑐𝑜𝑠 𝜃 + 𝑢 𝑖𝑌 𝑠𝑖𝑛 𝜃
𝑢𝑗 = 𝑢 𝑖 𝑋 𝑐𝑜𝑠 𝜃 + 𝑢 𝑖𝑌 𝑠𝑖𝑛 𝜃

Thus a point can now move along the X, Y directions as given by uX, uY, but
these are not independent—the resultant deformation can only be along the axis of
the element since that is the only deformation permitted for a truss. The relationship
between the nodal d.o.f. in the local coordinate frame of the element and the
generic d.o.f. in the global frame can be written as
𝑢𝑖 𝑋
𝑢𝑖 𝑐𝑜𝑠 𝜃 𝑠𝑖𝑛 𝜃 0 0 𝑢
{𝑢 } = ] = { 𝑢𝑗𝑖𝑌 }
𝑗 0 0 𝑐𝑜𝑠 𝜃 𝑠𝑖𝑛 𝜃
[
𝑢𝑗
Thus we have

{𝛿} 𝑒 = [𝑇]{𝛿}𝑒
ℓ 𝑔

where the subscript “ℓ” stands for local and the subscript “g” stands for global
coordinate frame. If we write the element level equations as

[𝐾]𝑒 {𝛿} 𝑒 = {𝐹 }𝑒
ℓ ℓ ℓ

Then

[𝐾]𝑒 [𝑇]{𝛿}𝑒 = {𝐹}𝑒

ℓ 𝑔 ℓ

Premultiplying by [T]T, we get

([𝑇 ]𝑇 [𝑘]𝑒 [𝑇]){𝛿}𝑒 = [𝑇 ]𝑇 {𝐹 }𝑒

ℓ 𝑔 ℓ
The RHS can be expanded as

𝑐𝑜𝑠 𝜃 0 𝐹𝑖 𝑐𝑜𝑠 𝜃
𝐹𝑖 𝐹𝑖 𝑠𝑖𝑛 𝜃
[𝑇]𝑇 {𝐹 }𝑒 = [𝑠𝑖𝑛 𝜃 ] [𝐹 ] = 𝐹
𝜃 𝑗 𝑐𝑜𝑠 𝜃
0 𝜃 { 𝐹𝑗 𝑠𝑖𝑛 𝜃 }
ℓ 0
From Figure (b),𝑐𝑜𝑠 we see that these are generic forces at the nodes in the
0 𝑠𝑖𝑛
global directions. Thus we have

[𝑇]𝑇 {𝐹}𝑒 = {𝐹}𝑒

ℓ 𝑔

If we define the element stiffness matriz in global reference frame as

[𝐾]𝑒 = [𝑇]𝑇 [𝑘]𝑒 = [𝑇 ]

𝑔 ℓ

then upon arranging,

[𝐾]𝑒 [𝛿]𝑒 = {𝑓}𝑒

𝑔 𝑔 𝑔

The above figure contrasts the element level equations in local and global
coordinates for a generic element.

Element 1. The element stiffness matrix in its own coordinate frame (xy) is given
by
 [𝐾]𝑒ℓ = 𝐴𝐸 1 −1
[ ]
𝐿/ 𝑐 𝑜 𝑠 −1 1
𝛼

In the global reference frame (XY), we have, with  =

𝐺𝑙𝑜𝑏𝑎𝑙 (1𝑋) (1𝑌) (3𝑋) (3𝑌)

𝑙𝑜𝑐𝑎𝑙 (𝑖𝑋) (𝑖𝑌) (𝑗𝑋) (𝑗𝑌) 𝐿𝑜𝑐𝑎𝑙𝐺𝑙𝑜𝑏𝑎
𝑙
(1)
[𝐾]𝑔
𝑐𝑜𝑠2 𝛼 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 − 𝑐𝑜𝑠2 𝛼 − 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠2 𝛼 (𝑖𝑋) (1𝑋)
𝐴𝐸 𝑐𝑜𝑠 𝛼 𝑠𝑖𝑛2 𝛼 − 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 − 𝑠𝑖𝑛2 𝛼 (𝑖𝑌) (1𝑌)
= [ 𝑐𝑜𝑠2 𝛼 𝑠 𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 ] (𝑗𝑋)
𝐿 (3𝑋)
𝑆𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑠𝑖𝑛2 𝛼 (𝑗𝑌) (3𝑌)

Element 2. The local frame (xy) and global frame (XY) are identical and hence
there is no transformation involved. We have

(2) (3)

(i) (j)
(2) 𝐴𝐸 1 −1] (𝑖)(2)
[𝑘]𝑔 = [ 1 (𝑗)(3)
𝐿 −1

Equation of element 2, when rewritten as a (4 × 4) matrix, gets modified as

𝐺𝑙𝑜𝑏𝑎𝑙 (2𝑋) (2𝑌) (3𝑋) (3𝑌)

𝑙𝑜𝑐𝑎𝑙 (𝑖𝑋) (𝑖𝑌) (𝑗𝑋) (𝑗𝑌)
𝐿𝑜𝑐𝑎𝑙𝐺𝑙𝑜𝑏𝑎
𝑙
1 0 −1 0 (𝑖𝑋) (2𝑋)
[𝐾]𝑔(2) = 𝐴𝐿𝐸 [ 0 0 0 0 ] (𝑖𝑌) (2𝑌)
−1 0 1 0 (𝑗𝑋) (3𝑋)
0 0 0 0 (𝑗𝑌) (3𝑌)

Element 3. The transformation matrix would be[𝑇(−𝛼)], 𝑖. 𝑒

[𝑇] = [𝑐𝑜𝑠 𝛼 − 𝑠𝑖𝑛 𝛼 0 0 ]

𝑛𝛼
0 0 𝑐𝑜𝑠 𝑠𝑖
𝛼
Therefore,
−
(4𝑋) (4𝑌) (3𝑋) (3𝑌)
(𝑖𝑋) (𝑖𝑌) (𝑗𝑋) (𝑗𝑌)
 𝑐𝑜𝑠2 𝛼 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 − 𝑐𝑜𝑠2 𝛼 − 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 (𝑖𝑋) (4𝑋)
(3) 𝐴𝐸 𝑠𝑖𝑛2 𝛼 − 𝑠𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 − 𝑠𝑖𝑛2 𝛼 ] (𝑖𝑌) (4𝑌)
[𝐾]𝑔 = [ −𝑠 𝑖𝑛 𝛼 𝑐𝑜𝑠 𝛼 (𝑗𝑋)
𝐿\ 𝑐𝑜𝑠 𝛼 𝑐𝑜𝑠2 𝛼 (3𝑋)
𝑆𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑠𝑖𝑛2 𝛼 (𝑗𝑌) (3𝑌)

Having obtained the element matrices in a common global frame XY, we are
ready to assemble the equations at the structure level. For the convenience of
assembly, we have marked the element level and structure level locations of the
matrix coefficients. Since we have four nodes and uX and uY at each node, we have
eight variables, and our global stiffness matrix is of size (8 × 8). The assembled
equations are given in Figure.

By virtue of the boundary conditions due to supports at nodes 1, 2 and 3, we

have

𝑢 1𝑋 = 0 = 𝑢1𝑌 , 𝑢 2𝑋 = 0 = 𝑢 2𝑌 , 𝑢 4𝑋 = 0 = 𝑢 4𝑌

Thus from the 5th and 6th equations of the above Figure,

3
[1 + 2 𝑐𝑜𝑠 𝛼 ] { 𝑢 3𝑋} = { 0𝑝}
2 𝑠𝑖𝑛2 0
𝐴𝐸
𝐿
𝛼 𝑐𝑜𝑠 𝛼 𝑢 3𝑌
0

Solving, we get

𝑢 3𝑋 = 𝑃𝐿 1
𝐴𝐸 1+2 𝑐 𝑜 𝑠 3 𝛼
, 𝑢 3𝑌 =
0

which tallies with the well-known result.

Having obtained the deflection of the structure, we can use the first six
equations of the above Figure to find the support reactions at nodes 1−3, which
ensure that the displacements at these nodes are zero. We now summarise the
general procedure for assembling element level equations.

Summary: General Method of Assembly

Step 1: Identify the local (or element) level and global (on structure) level
coordinate frames of reference.

Step 2: Obtain the element matrices in the local reference frame.

Step 3: Obtain the coordinate transformation matrix [T] between the local and
global frames of reference.

Step 4: Transform the element matrices into the common global reference
frame.

Step 5: Identify the global locations of the individual coefficients of the

element matrices based on local and global node numbers.

Step 6: Assemble the element matrices by placing the coefficients of the

element matrices in their appropriate places as identified in Step 5.

4.5 Beam Element

4.5.1 Selection of Nodal d.o.f.

A truss or rod undergoes only axial deformation and we assumed that the entire
cross-section undergoes the same displacement u. Thus the bar element was just a line
element with two nodes and each node having u as the nodal d.o.f. A beam, on the
other hand, undergoes transverse deflection denoted by v. The Figure below shows a
typical beam section undergoing transverse deflection. We assume that the cross-
section is doubly symmetric and the bending takes place in a plane of symmetry.
According to the classical theory of beam bending (Euler−Bernoulli theory), the
entire cross-section has the same transverse deflection v as the neutral axis;
sections originally perpendicular to the neutral axis remain so after bending.
Hence the deflections are small and we assume that the rotation of each section is
the same as the slope of the deflection curve at that point, i.e., ( dv/dx).

Therefore, the axial deformation of any point P away from the neutral axis
(due to bending alone) is given by uP =

– (z)(dv/dx). Thus, if v(x) is determined, then the entire state of deformation of the
body is completely determined.

Our beam element is therefore also a simple line element, representing the
neutral axis of the beam. However, in order to ensure the continuity of deformation
at any point (i.e. those on the neutral axis as uell as auay from it), we have to
ensure that v and (dv/dx) are continuous. We achieve this by taking two nodal
d.o.f., viz., v and  (= dv/dx). Thus when two beam elements are joined together,
they share a common node and so have the same transverse deflection and slope. If
we did not have the slope d.o.f., then we would only have been able to ensure that
transverse deflection is continuous across elements, but not guarantee that the
slope also would be continuous. Thus the points away from the neutral azis could
actually separate (thereby creating a void) or overlap (creating a kink). We recall
that for this problem (fourth order differential equation in v), the essential boundary
conditions consist of both v and dv/dx. Thus, taking these nodal d.o.f. permits us to
readily assign prescribed essential boundary conditions (e.g. fized end v = 0 =
dv/dx). Moreover, a prescribed value of moment load can readily be taken into
account with the rotational d.o.f  The Euler− Bernoulli beam element is shown in
Figure.
Determination of Shape Functions

The displacement field v (x) assumed for the beam element should be such

that it takes on the values of deflection and the slope at either end as given
by the nodal values vi, i, vj, j, Let v(x) be given by

𝑣(𝑥) = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + 𝑐3 𝑥 3

Differentiating with respect to x, we have

𝑑𝑣/𝑑𝑥 = 𝑐1 + 2𝑐2 𝑥 + 3𝑐3 𝑥 2

Atx=0 and L, we have

𝑣𝑖 = 𝑐0 , 𝜃𝑖 = 𝑐1 , 𝑣𝑗 = 𝑐0 + 𝑐1 𝐿 + 𝑐2 𝐿2 + 𝑐3 𝐿3 , 𝜃𝑗 = 𝑐1 + 2𝑐2 𝐿 +
3𝑐3 𝐿2
We can solve for the coefficients c0, c1, c2, c3 in terms of the nodal
d.o.f.𝑣𝑖 , 𝜃𝑖 , 𝑣𝑗 , 𝜃𝑗 and substitute in the first equation to get

𝑣𝑖 (𝑥) = 𝑁1 𝑣𝑖 + 𝑁2 𝜃𝑖 + 𝑁3 𝑣𝑗 + 𝑁4 𝜃𝑗

where the shape functions N are given by

𝑁1 = 1 − 3𝑥 2 /𝐿2 + 2𝑥 3 /𝐿3 ,𝑁2 = 𝑥 − 2𝑥 2 /𝐿2 + 𝑥 3 /𝐿3

𝑁3 = 3𝑥 2 /𝐿2 − 2𝑥 3 /𝐿3 ,𝑁4 = −𝑥 2 /𝐿 + 𝑥 3 /𝐿3

(Note: The reader is advised to carry out the steps to gain familiarity with this
method of deriving the shape functions.)

Rewriting , we get
𝑣𝑖
𝜃𝑖
𝑣𝑖 (𝑥) = [𝑁]{𝛿}𝑒 = [𝑁1 𝑁2 𝑁3 𝑁4 ] { 𝑣 }
𝑗
𝜃𝑗
THE FINITE ELEMENT METHOD
or NODAL APPROXIMATION METHOD:

➢ The basic concept behind the Finite element method is “going from part to whole”
➢ Name “FINITE ELEMENT” coined by Clough
➢ Fitting of a number of piecewise continuous polynomials to approximate the variation of
the field variable over the entire domain

STEPS INVOLVED IN THE FINITE ELEMENT METHOD:

Discretisation of the structure: In this step the given structure is divided into subdivisions or
elements. Depending upon the problem we may choose I D, II D or IIID elements.

I D elements
Constant strain triangular element Bilinear Rectangular element

Linear strain triangular element Eight noded quadratic quadrilateral elements

II D elements

3
4

1 2

Linear Quadrilateral element

III D elements
Selection of suitable displacement model:
We make an assumption as to the variation of the unknown solutions over the element. In
general, the field variable (example, temperature, displacement etc.) is assumed to vary
linearly or quadratically or cubically.

Displacement model associated with each element

Field variable u

Length l

Derivation of elemental matrices and load vectors:

From the assumed displacement model, the elemental stiffness matrix [K]e and load vector
[P]e of the element are to be derived using either equilibrium methods or a suitable
variational principle.
Assembly of elemental equations to obtain overall stiffness matrix: the individual element
stiffness matrices and load vectors are to be assembled in a suitable manner to get the
overall stiffness equation which is expressed as
[K]{u} = {P}
where [K] is the assembled stiffness matrix
{u} is the vector of unknowns or nodal
displacements
{P} is the vector of nodal forces for the
complete structure

Imposition of boundary conditions: The Boundary conditions could now be

incorporated to get the reduced equations.

Solutions for the unknown nodal displacements: The elemental matrices, on assembly,
yield a set of equations, which could be expressed as a set of matrices, which could be
solved using any iterative procedure or numerical method.

Computation of elemental strains and stresses: From the unknown displacements, the
element strains and stresses can be computed by using the necessary equations of solid or
structural mechanics.

54
Example:

L1 = 10 cm 1
L2 = 10cm
E= 2x107N/cm2 2
EBC:
2 U1 = 0
Pl = 10kN
3
A 1 = 2sq.cm A
2 = 1sq.cm

10 kN

 1 − 1
K  1
=
EA1
 
l1  −1 1 
 4  105 − 4  105 
K  1
=  
5 
 − 4  10 5
4  10 
EA2  1 − 1
K 2 =  
l2  − 1 1 
 2  105 − 2  105 
K  2
=  5 

 − 2  10 5
2  10 
The assembled stiffness matrix is given by

 4 −4 0 
 
K g =10  − 4 4 + 2 − 2 
5

 0 − 
 2 2 
The load vectors are

R
P =  
1

0
where R is the reaction at the fixed end
0 
P 2
= 
1 
R
 
P = 0
10
 
The overall equilibrium equation is given by
[K] {u} = {P}

or

 2 − 2 0  u1   R 
     
2  10  − 2 3 − 1 u2  =  0 
5

 0 − 1 1  u  10
   3  

 2 − 2 0  u1   R 
    
2  10  − 2 3 − 1 u2  =  0 
5

 0 − 1 1  u  10
   3  
5 3 − 1 u2  0
2  10     =  
 − 1 1  u3  1

u2 = 0.25 x 10-4 cm
u3 = 0.75 x 10-4 cm
 The stresses in the elements are given by

Stress in element 1 = 1 = Є 1E1

= (0.25 x 10-5) (2 x 107)
= 5 kN/cm2
Stress in element 2 = 2 = Є 2E2
= (0.50 x 10-5) (2 x 107)
= 10 kN/cm2

COMPUTATION OF REACTION AT FIXED END:

2 x 105 [2 * u1 – 2 * u2] = R
Substituting for u1and u2 we get
Reaction R= 10kN
 Example

x
A( x) = A1 − ( A1 − A2 )
l
x
i.e. A( x) = 80 − (80 − 20)
300
= (80 − 0.2 x)
Specific weight  = 0.075 N/cm3
Young's Modulus E = 2 x 107 N/cm2

l
dN1 dN1
K 11 =
0 EA(x) dx dx dx
 A1 - A2   1  2
= l


0
E  A1 -
 l
x   −  dx
  l

= E A1 A2 E(A1 + A2 )
( + )=
l 2 2 2l
 l
dN1 dN 2

=
K 12 EA(x) dx
0
dx dx

 A1 - A2   − 1   1  dx
l


=
E  A1 - x  l   l 
0  l 

E ( A1 + A2 )
E A1 A2 = −
=
− ( + ) 2l
l 2 2

K 12 = K 21
l
dN 2 dN 2

=
K 22 EA(x) dx
0
dx dx

 A -A 
l 2
1

=
E  A1 - 1 2    dx
0  l  l
E A A E ( A1 + A2 )
= ( 1 + 2 ) =
l 2 2 2l
Therefore the element stiffness matrix will be

e E A1 + A2 1 -1 
[K ] =

l 2 -1
 1
60
Similarly the element nodal load vector will be

l
r1 =  γA(x) N1 dx
0

 (A - A ) 
l
x
=

0
[γ  A1 - 1 2  (1 - ) ] dx
 l  l

=  A1 A2 
 l + l
l
3 6 
r2 =  γA(x) N 2 dx
0

 (A - A )  x
l
=

0
[γ  A1 - 1 2  ( ) ] dx
 l  l
 A1 A2 
=  l + l
6 3 
Therefore the assembled load vector will be

r  e = γl  2A1

6 2A2
A2 

A1 

Case - I: Discretize the Tapered Bar into 3 elements.

The length of each element = 100 cm.
 E A1 + A2 1 -1  E  70 − 70 
K =
1
-1 =
l1 2  1 100 − 70 70 


E A2 + A3 1 -1  E  50 − 50 
K =
2
-1 =
l2 2  1 100 − 50 50 


E A3 + A4 1 -1  E  30 − 30
K =
3
-1 =
l3 2  1 100 − 30 30 


The global stiffness matrix will become

[K 1 ] 
K  = 
 [K 2 ]


 [K 3 ] 

 70 -70 
E − 70 70+50 -50 
=  
100  -50 50+30 -30 
 
 -30 30
 70 -70 0 0 
E − 70 120 -50 0 
= 
100  0 -50 80 -30 
 
 0 0 -30 30 
r 1  γl 2A1 + A2 
 =  
 r2  6  2A2 + A 1 
 220 
r  =
1 
 x 100  6 

200


 6 
 100 
r  =
 160 
r  =
 6 
  3
γ x 100 
80 
2
γ x 100  6 
140  
   6 
 6 

Similarly the assembled global load vector will become

|r |1
 P1 
R =

2
|r |
+


 P2


  P 3 
 |r 3| 
The global load vector is

 220   R
 6   

R =   +  
 200 160
+   O
 6 6   
γ x 100    
 140 100 
+  O
 6 6   
 80   
   P
 6   
 220  R 
   
   
 360  O 
γ x 100    
=   +  
6  240  O 
   
   
 80  P 
   

65
Now the total system of equation will be

 70 -70   u1   220 R 
-70  u    O 
E  120 -50   2 = γ x 100  360  
       
100  50 80 -30  u3  6  240  O 
 -30 30   u4   80   P 
Now applying the Boundary conditions i.e. u1 = 0 ..
Delete the first row and first column of elements and the system of equation will reduce to

 120 -50   u2   360 O 

    = γ x 100    
 -50 80 -30   u3   240 O 
6  80  P 

 30 30

u     
 4 66

The data are E = 2 x 107 N/cm2  = 0.075 N/cc and P = 1 x 105 N.

On solving the above equation we get
u4 = 0.035501997 cm
u3 = 0.018818567 cm
u2 = 0.008778557 cm
The deflection at mid section of the bar by interpolation is
= 0.01379856 cm

u2 + u3
U x=50 =
2
Example 3

Element 1,
u1 u2
2EA  1 − 1
k1 =  
L − 1 1 

Element 2,
u2 u3
EA  1 − 1
k2 =
L − 1 1 
  2 − 2 0  u1   F1 
EA 
− −  u   
=  F2 
L 
2 3 1  2
 0 − 1 1  u3  F 
 3

68

EA
3u3  = P
L
PL
Thus, u2 =
3EA

Stress in element 1 is
u2 − u1
σ 1 = Eε1 = E
L
E  PL  P
=  − 0 =
L  3EA  3A
Similarly, stress in element 2 is
u3 − u 2
σ 2 = Eε2 = E
L
E PL  P
= 0 −  =−
L  3EA  3A

Which indicates that bar 2 is in compression.

 1 −1 0   0   F1 
EA      
 − 1 2 − 1 u 2 = P 
L
 0 − 1 1   Δ  F 
 3
WEAK FORM OF GOVERNING EQUATION FOR THERMAL
PROBLEMS

where
k = Thermal conductivity coefficient
h = Thermal convection coefficient
A = Area of cross section subjected to CONDUCTION
p= Perimeter is the area exposed to CONVECTION
T = Atmospheric Temp. , T = Variable
Q = Heat Source
(q + dq) − q + hp dx (T − T ) = 0
 by dx we get
dq
+ hp (T − T ) = 0
dx
d  dT 
 − kA( x)  + hp (T − T ) = 0
dx  dx 

Boundary conditions:
i) At x= 0 T = To

ii) At the free end any one of the following three possible boundary conditions could be
specified
1. If free end is insulated _ kA dT/dx = 0
2. If free end is open to atmosphere
_ kA dT/dx|=l = hA(T- T)
3. Specified temperature T(l) = Tl
 The governing equation for heat transfer in a one dimensional problem is given by

d  dT 
 − KA  + hp(T − T) = 0
dx  dx 
The weak form can be obtained by

 w(x)R(x)dx =0
For a bar of length ‘l’ with wall temperature ‘T’ the weak form of the governing equation becomes

d  dT  
l

0  dx 
w( x ) − KA
dx 
+ hp (T − T ) dx = 0



d  dT 
l l

0 w( x ) 
dx 
− KA 
dx 
dx +  w( x)hp(T − T)dx = 0
0

d  dT 
Let l
I 1 =  w( x)  − KA dx
0
dx  dx 
and
u = w(x ) du = dw
d  dT  dT
dv = − KA dx v = − KA
dx  dx  dx
 I 1 = uv −  vdu
l
 dT   dT  dw
l
I 1 = w( x) − KA  −  − KA  dx
 dx  0 0  dx  dx
Substituting the above term in equation 1, we get

l
 dT   dT  dw
l l
w( x)− KA  −  − KA  dx +  w( x)hp(T − T)dx = 0
 dx  0 0  dx  dx 0

l
 dT 
l l l
dT dw
w( x)− KA  +  KA dx +  hpw ( x)T ( x)dx −  hpw ( x)Tdx = 0
 dx  0 0 dx dx 0 0

Boundary term B1(T,w) B2(T,w) l(w)

l l l
dT dw
0 dx dx dx + 0 hpw( x)T ( x)dx = 0 hpw( x)Tdx
KA

− w( x)hA(TL − T)
 Substituting in the weak form

And w(x) as N1 first and then N2 we get a system of two equations in two unknowns
namely T1 and T2 which can be written as

T ( x) = N1T1 + N 2T2

 K 11 K 12   K 11 K 12 
   1 
T  T1  q1 
  +   = 
 K 21 K 22  T2   K 21 K 22  T2   q 2 
  cond   conv

l
dN i dN j
=  kA(x)
Where e
K ijcond dx
0
dx dx
l
K ije conv =  hp(x)
0
N i N j dx
l

q ej =  hpT
0
 N j dx
Let the elements be of equal length l
The element matrices are

KA 1 -1  hP l 2 1 0 0 
[K e ] = -1 1 1 2 0 hA
+ +
l  6  

e hPl T 1 0 
[f ] =  +  
2 1 hA T
Boundary conditions:
at x = 0, T(0) = T
dT
at x = L, − KA l = hA (Tl - T )
dx

conduction = convection loss

For a typical linear element

NI = 1 - (x/l)
NJ = (x/l)

Let the elements be of equal length l = 2 cm

The element matrices are

e kA 1 -1  hp l 2 1 0 0 
[K ] = -1 1
+ 1  + 
hA
l  6  2 0

e
hpl T 1 0 
[q ] =   +  
2 1 hA T
 The element matrices for ELEMENT (1),
(2) & (3) are
e  6 − 6 20 
[K ]cond =  ; {qe } =  
− 6 6  20 

e 0.667 0.333  20

[K ]conv =  ; {qe } =  
0.333 0.667  20

e 6.666 -5.667  20

[K ] therm =   ; {qe } =  
-5.667 6.666  20
The element matrix for ELEMENT (4) is
e 6 − 6 
[K ]cond = 
 − 6 6 
0.667 0.333  0 0 
e
[K ]conv =  + 
0.333 0.667  0 0.4 
20  0 
{qe } =   +  
20  8 
e 6.666 -5.667  20 
[K ]therm =  ; {qe } =  
-5.667 7.066  28 
77
 On assembly we get

 6.667 -5.667 0 0 0  T1   20 

-5.667 13.33 -5.667 0 0  T2  20+20
     
 0 -5.667 13.33 -5.667 0  * T3  = 20+20
 
 0 0 -5.667 13.33 -5.667  T4  20+20
   
 0 0 0 -5.667 
7.066  T5   28 

By applying Boundary condition at

at x = 0 T = T0 = 80º
 13.33 -5.667 0 0  T2  40+5.667*80 
-5.667 0  T3  40 
 13.33 -5.667    
*   =  
 0 -5.667 13.33 -5.667  T4
   40 
  T5  28 
 0 0 -5.667 7.066 

By solving we get
T2 = 53.95 0
C; T3 = 39 .88 0 C;

T4 = 32 .82 0
C; T5 = 30.29 0 C;
BEAM ELEMENTS
 Beam in Bending-Continued
• Considering the equilibrium of vertical forces and
moments, we have the governing equation:

 Fy = 0
− Q + Q + dQ + qdx = 0 80

dQ
+ q( x) = 0 − − − − − − − − 1
dx
M = 0
M + Qdx − M − dM
dM
= Q − − − − − − − − − − − −2
dx
 dM
Sub Q = in eqn. 1
dx
d  dM 
   + q( x) = 0
dx  dx 
d 2M
2
+ q( x) = 0
dx
d 2w
Now M = − EI
dx 2
d2  d 2w 
 EI
2 
 − q( x) = 0
2 
dx  dx 
Governing Differential Equation
d 4 w(x)
EI 4
= q(x); q is the distributed loading
dx
Boundary conditions could involve
specification of any of the following variables
w = transverse displacement
dw
θ= = Slope
dx
d 2w
M = EI 2 = Moment
dx
d 3w
Q = EI 3 = Shearforce
dx
Boundary conditions
w = transverse displacement
Primary
dw
θ= = Slope variables
dx
d 2w
M = EI 2 = Moment
dx Secondary
d 3w variables
Q = EI 3 = Shearforce
dx

Possible loads
Distributed load (uniform or non-uniform),
Transverse loads, Transverse moments or
combination loading in transverse direction
 Shape functions for beam element

Sign conventions

w (x) = a0 + a1 x + a2 x 2 + a3 x 3  (1)
a0 
a 
 
w (x) = 1 x x2 x 3   1   (2)
a 2 

 a3 

θ (x) = a1 + 2a2 x + 3a3 x 2  (3)
a0 
a 
 1
θ (x) = 0 x 2x 3x     (4)
2

a 2 

 a3 

At x=0 w=w1 and θ= θ1
At x=l w=w2 and θ= θ2

at x = 0 w1 = a0 + a1 0 + a2 0 + a3 0
θ1 = 0 + a1 + 2a2 0 + 3a3 0
x=l w2 = a0 + a1l + a2l 2 + a3l 3
θ2 = 0 + a1 + 2a2l + 3a3l 2

 w1  1 0 0 0  a0 
 θ  0 0   a1 
 1  1 0
 =  
w2  1 l l2 l 3  a 2 
2 
 θ2  0 1 2l 3l  a3 
−1
a0  1 0 0 0   w1 
 a  0 0  θ 
 1  1 0  1
 =  
a 2   1 l l2 l3  w2 
2 

 a3 
 0 1 2l 3l  
 θ2 

 −1
1 0 0 0   w1 
0 1 0  θ 
0  1
w (x) = 1 x x 2
x3    
1 l l2 l3  w2 
  
0 1 2l 3l 2   θ2 


 w1 
θ 
 1
w(x) = N 1 N2 N3 N4   
w2 
 θ2 
 3x 2   2x3 
N 1 = 1 −  2  +  3 
 l   l 
 2x 2   x3 
N 2 = x −   +  2 
 l  l 
 3x 2   2x3 
N 3 =  2  −  3 
 l   l 
 x2   x3 
N 4 = −  +  2 
 l  l 
These shape functions are based on
Hermitian polynomials
 N1 & N2 associated with displacement
N2 & N4 associated with slopes

Ritz Weak Formulation

l
 d 4 w( x) 
0  dx 4
EI − q ( x )  v( x)dx = 0 v( x) = is the weighting function

l l
d 4 w( x)
0 EI dx 4 v( x)dx − 0 q( x)v( x)dx = 0
Integration by parts,
d 4 w( x) d 3 w( x)
u = v (x ); dv = EI v = EI
dx 4 dx 3
l
 d 3w d 3 w dv
v( x) EI 3  −  EI 3 dx −  q( x)v( x)dx = 0
 dx  0 dx dx
 dv d 2v
Now u = , and2du =
dx dx

d 3w d 2w
dv = EI 3 , and v = EI 2
dx dx
l l
 d 3 w   dv d 2w d 2 w d 2v
v( x) EI 3  −  EI 2  +  EI 2 2
dx −  q( x)v( x)dx = 0
 dx  0  dx dx  0 dx dx

Rearranging,
l l
l 2
d wd v 2 l
 dv d w 
2
d w
3

 EI
0 dx 2 dx 2
dx = 0 q ( x ) v ( x ) dx + 
 dx
EI
dx 2 
−
0 
v ( x ) EI
dx 3 
0

l l
l 2
d wd v 2 l
 dv d 2w  d 3w
0 EI dx 2 dx 2 dx = 0 q( x)v( x)dx +  dx EI dx 2  − v( x) EI dx 3 
0 0

l l
l
d 2 w d 2v
l
 dv d 2w  d 3w
0 EI dx 2 dx 2 dx = 0 q( x)v( x)dx +  dx EI dx 2  − v( x) EI dx 3 
0 0

Slope Momen Shear force

t
 l l
d 2 w d 2v
0 EI dx 2 dx 2 dx =  q(x)v(x)dx+
0

M(l)θ(l) − M(0)θ(0) − Q(l)w(l) − Q(0)w(0)

Strain Energy = Work Done by UDL +

Work done by moment +
Work done by shear
force

➢ From the quadratic functional we see that specification of w and dw/dx= θ

constitutes the essential boundary conditions.

➢ Specification of Q and M constitutes the natural boundary conditions

➢ Since a quadratic functional exists minimizing it will lead to the equilibrium

equations in either the direct form or in the variational (weak) form
Substituting for w(x) and v(x) as given below
 w1 
 
 
w( x) = N1 N 2 N 3 N 4   1 
w2 
 2 
ie
w( x) = N 1 w1 + N 21 + N 3 w2 + N 4 2
and
v( x) = N1 , N 2 , N 3 , N 4

Substituting for the displacement in the weak form of the equation,

and taking the weighting functions as the shape functions, we get a
system of 4 equations in 4 unknowns.

K u =  f 
K u =  f 

l 2
d 2 Ni d N j
K ij =  EI 2 2
dx  w1 
0 dx dx  
 
where u =  1 
l w2 
f j =  q(x)N j (x)dx  2 
0

Stiffness Matrix for beam element

l 2
d 2 Ni d N j
K ij =  EI 2 2
dx
0 dx dx
 3x 2   2x 3 
N 1 = 1 −  2  +  3 
 l   l 
 2x 2   x3 
N 2 = x −   +  2 
 l  l 
 3x 2   2x 3 
N 3 =  2  −  3 
 l   l 
 x2   x3 
N 4 = −  +  2 
 l  l 
dN 1 6x 6x 2 d 2 N1 6 12x
=− 2 + 3 = − +
dx l l dx 2 l2 l3
dN 2 4x 3x 2 d 2 N2 4 6x
= 1− + 2 = − +
dx l l dx 2 l l2
dN 3 6x 6x 2 d 2 N3 6 12x
= 2 − 3 = −
dx l l dx 2 l2 l3
dN 4 2x 3x 2 d 2 N4 2 6x
=− + 2 = +
dx l l dx 2 l l2

l
d 2 N1 d 2 N1
K11 =  EI 2 2
dx
0 dx dx

2
l
 6 12 x 
K 11 =  EI  − 2 + 3  dx
 l l 
0  
EI
= 12 3
l
 l
d 2 N1 d 2 N 2
K 12 =  EI 2 2
dx
0
dx dx
l
 6 12x  4 6 
=  EI  − 2 + 3  − + dx
 l l  l l 
0   2 

EI
= 6 2 = K 21
l
l
 6 12x  6 12x 
K 13 =  EI  − 2 + 3  −
 l 2 l 3 
dx
 l l
0  
EI
= −12 3 = K 31
l
6EI
K 14 = 2 = K 41
l

4EI 6EI
K 22 = K 23 = − 2 = K 32
l l
 2EI
K 24 = − = K 42
l

12EI
K 33 = −
l3
6EI
K 34 = − 2 = K 43
l

4EI
K 44 = −
l
 12 6l − 12 6l 
 4l 2 − 6l 2l 2 
EI  6l
StiffnessMatrixK  = 3
e

l − 12 − 6l 12 − 6l 
 2 
 6l 2l 2
− 6l 4l 

Now the load vector is given by

l
f j =  q(x)N j (x)dx
0
 l l
  3x 2   2x 3  
f 1 =  q(x)N1(x)dx =  q(x) 1 −  2  +  3  dx
0 0   l   l 
ql
=
2
l l
  2x 2   x 3  
f 2 =  q(x)N2(x)dx =  q(x) x −   +  2  dx

0 0   l   l  
2
ql
=
12

l l
  3x 2   2x 3  
f 3 =  q(x)N3(x)dx =  q(x)  2  −  3  dx
0 0  l   l 
ql
=
2
l l
  2x 2   x 3  
f 4 =  q(x)N4(x)dx =  q(x) x −   +  2  dx

0 0   l   l

2
ql
=−
12
 Load Vector is given by

 1 
 
ql  l/6 
 f e =  
2  1 
− l/6 

Hence the element stiffness and load vector

for the beam element are given by
 12 6l − 12 6l 
 4l 2 − 6l 2l 2 
EI  6l
StiffnessMatrixK  = 3
e

l − 12 − 6l 12 − 6l 
 2 
 6l 2l 2
− 6l 4l 
 1  R
   
ql  l/6   0 
 f e =  + 
2  1  0 
− l/6  M 
 Beam Element
For a classical beam element,

 w1 
 
 
w( x) = N1 N 2 N3 N4  1 
 w2 
 2 
 w1 
 
du d  dw  d 2 w d 2 N1 d 2 N 2 d 2 N3 d 2 N 4  1 
 xx = =  z  = z 2 = z  
dx dx  dx  dx dx 2 dx 2 dx 2 2
dx  w2 
 2 
Example 1: Cantilever Beam subjected to
point load at the tip

Boundary conditions for this beam are

At x = 0, w1 = 0 and θ1 = 0
d 3w d 2w
At x = l, EI 3 = P and EI 2 = M = 0
dx dx
The Equilibrium Equation is given by
 12 6l − 12 6l  w1    R 
  θ   M 
EI  6L 4l
2
− 6L 2l 2  1   
  =  
l 3  − 12 − 6l 12 − 6l w2   − P 
 6l 2l 2
− 6l 4l 2  θ2    O 
 
Imposing the essential Boundary conditions
we can strike off columns 1 & 2 & Rows 1 & 2
which leaves us with
EI  12 − 6l   w2  − P

3 
  
2 
=  
l  − 6l 4l   θ2   O 
 Which gives the equations.
12EI 6EI
3
w2 − 2 θ2 = − P
l l
6EI 4EI
− 2 w2 + θ2 = 0
l l

Solving for θ 2 & w2 we get

Pl 2 Pl 3
2 = and w2 =
2 EI 3EI

Example 2: Simply supported beam with

uniformly distributed load

The above beam can be idealized by using

one element. The entire beam need not be
modeled. Instead, taking advantage of
symmetry we can model one half of the beam
The boundary conditions in this case are

d 2w
At x = 0, w1 = 0 and EI 2 = 0
dx
d 2w
At x = l ,  2 = 0 and EI 2 = 0
dx
The stiffness matrix is given by
 12 6l − 12 6l   w1   1  R
EI 
 6l 4l 2 − 6l 2l 2   θ1  fl  l/6   0 
 =  + 
l 3 − 12 − 6l 12 − 6l  w2  2  1   0 
 2 
 6l 2l 2
− 6l 4l   θ2  − l/6  M 

Where R is the reaction at left end and M is

the moment at mid section.
The reduced stiffness matrix after imposing
Boundary conditions are given by
 12 6l − 12 6l   w1   1  R

EI  6l 4l 2 − 6l 2l 2   1  fl  l / 6   0 
 =  + 
l 3 − 12 − 6l 12 − 6l  w2  2  1   0 
 
 6l 2l 2 − 6l 4l 2   2  − l / 6 M 

EI  4l 2 − 6l   1  fl l / 6 0
   =   +  
l3 − 6l 12  w2  2  1  0
 4EIθ1 6EIw2 fl 2
− 2
=
l L 12
6EIθ1 12EIw2 fl
− 2 + 3
=
l l 2
8EIθ1 12EIw2 fl
2
− 3
=
l l 6
3
fl
1 =
3EI
4
5 fl
w1 =
24 EI

Substitute l = L 2
3
fl
We get θ1 =
24EI

4
5fL
w1 =
384EI
Example 3: Fixed – Fixed beam with central
load

The above beam can be modeled taking

advantage of symmetry as a single element

Boundary conditions:
At x = 0, w1 = 0 and 1 = 0
d 3w P
At x = l ,  2 = 0 and EI 3 = −
dx 2

Deleting 1st , 2nd and 4th rows and columns

of the stiffness matrix the equilibrium
equation is given by
  12 6l − 12 6l   w1   R 

EI  6l 4l 2 − 6l 2l 2   1   M 
  = 
l 3 − 12 − 6l 12 − 6l  w2  − P / 2
 
 6l 2l 2 − 6l 4l 2   2   M 

Deleting 1st , 2nd and 4th rows and

columns of the stiffness matrix the
equilibrium equation is given by

EI P
12 3 w2 = −
l 2

P l3
or w2 = −
2 12EI Downwards
Pl 3
=
24 EI
L
Substituti ng l = we get
2
PL3
w2 =
192EI
EXAMPLE 4: The beam shown in fig is fixed
at both ends and supported between the ends
with a simple support that allows rotation.
Compute the rotation and reaction at the
supports. Also determine the moments and
shear forces.

The given beam can be discretized into two

elements as shown below

The stiffness matrix & equations are given by

Element 1
 12 6L − 12 6L   w1   1 
 2    
EI  6L 4L − 6L 2L   θ1  fL  L/6 
2

 =  
 
L − 12 − 6L 12 − 6L w2  2  1 
3

 2 
−   θ2  − L/6 
2
 6L 2L 6L 4L

Element 2
 12 6(2L) − 12 6(2L)   w1  0 
 2    
EI 6(2L) 4(2L) − 6(2L) 2(2L)   θ1  0 
2

 = 
(2L)  − 12 − 6(2L)
3
12 − 6(2L) w2  0 
 2 
6(2L) 2(2L) − 6(2L) 4(2L)   θ2  0 
2

The global stiffness matrix is a (6 x 6) matrix.

Boundary conditions are
w1 = w2 = w3 = θ 1 = θ 3 = 0
The global equations now reduces to one
equation and one unknown, θ 2 [Remove 1st,
2nd, 3rd, 5th, & 6th rows & columns].

( )
2
EI fL
3
4 L2
+ 2 L2
2 =
L 12
fL3
or  2 =
74 EI
Now to compute reactions and moments for
each span we utilize the local stiffness matrix
for that span. Let the reactions and moments
for the span 1-2 be R1, M1, R2 and M2.

 12 6L − 12 6L   0   1   R1 
 2 
     
EI  6L 4L − 6L 2L   0  fL  L/6   M 1 
2

 0 =  + 
3

L − 12 − 6L 12 − 6L  3  2  1   R2 
 2 
f L
−  72EI  − L/6  M 2 
2
 6L 2L 6L 4L

Solving we get
7 fL fL2 1 5 fL WL2
R1 = ; M1 = ; R2 = ;M2 =−
12 9 12 36
R2 represents the reaction at node 2 which is
the sum of shear forces at 2nd node of
element (1) and that at the 1st node of
element (2). Thus R2 = R12 + R22.
The stiffness matrix for element (2) can be
used to compute R22, M2, R3 and M3.
  12 12L − 12 12L   0 3   R22 
 12L 16L2 − 12L 8L2   fL   
EI   72EI  = M 2 
8L3 − 12 − 12L 12 − 12L  0   R3 
 2 
 12L 8L
2
− 12L 16L   0   M 3 

Solving we get
fL fL
R =
2
2 R3 = −
48 48
fL2 fL2
M2 = M3 =
36 72
R2 = R21 + R22
 ASSIGNMENT
S.No Question K Level CO
.
01 Axial load of 500N is applied to a stepped shaft, at the K2 CO2
interface of two bars. The ends are fixed. Calculate the
nodal displacements and stresses when the element is
subjected to all in temperature of 100˚C. Take E1 = 70*103
N/mm2, E2 = 200*103 N/mm2, A1 = 900mm2, A2 =
1200mm2, α1 = 23*10-6 / ˚C, α2 = 11.7*10-6 / ˚C, L1 =
200mm, L2 = 300mm.

02 The stepped bar shown in fig is subjected to an increase in K2 CO2

temperature, T=80o C. Calculate the displacements,
element stresses and support reactions.

03 An axial load P = 300 x 103 N is applied at 200C to the rod K2 CO2

as shown below. The temperature is then raised to 600C
a) Assemble the stiffness (K) and load (F)
matrices.
b) Calculate the nodal displacements and
element stresses.

40
 ASSIGNMENT
S.No Question K Level CO
.
04 A tapered steel plate of uniform thickness 40 mm with K2 CO2
Youngs modulus 2*105 N/mm2 and weight density
0.82*10-4 N/mm3. The Plate is subjected to a point load of
500N at mid-point. Compute the nodal displacement,
element stress and reaction force.

05 Determine the displacements and slopes at the nodes for K2 CO2

the beam shown in fig. Take k = 200 kN/m, E = 70 GPa
and I = 2*10-4 m4

40
 PART - A QUESTIONS AND ANSWERS
UNIT – II INTRODUCTION
S.N O QUESTION AND ANSWER K CO
LEVEL
1 What is meant by finite element? K1 CO2
A small units having definite shape of geometry and
nodes is called finite element.

2 What is meant by node or joint? K1 CO2

Each kind of finite element has a specific structural
shape and is inter- connected with the adjacent element by
nodal point or nodes. At the nodes, degrees of freedom are
located. The forces will act only at nodes at any others
place in the element.
3 What is the basic of finite element method? K1 CO2

Discretization is the basis of finite element method.

The art of subdividing a structure in to convenient
number of smaller components is known as
discretization.
4 State the three phases of finite element method. K1 CO2

Preprocessing
Analysis
Post Processing
5 What is meant by degrees of freedom? K1 CO2

When the force or reaction act at nodal point node is

subjected to deformation. The deformation includes
displacement rotation, and or strains. These are
collectively known as degrees of freedom.
6 What is meant by discretization and assemblage? K1 CO2

The art of subdividing a structure in to convenient

number of smaller components is known as
discretization. These smaller components are then put
together. The process of uniting the various elements
together is called assemblage.

7 What are the h and p versions of finite element K1 CO2

method?
It is used to improve the accuracy of the finite
element method. In h version, the order of polynomial
approximation for all elements is kept constant and the
numbers of elements are increased. In p version, the
numbers of elements are maintained constant and the
order of polynomial approximation of element is
increased.
10
9
 PART - A QUESTIONS AND ANSWERS
UNIT – I INTRODUCTION
S.N O QUESTION AND ANSWER K CO
LEVEL
8 What is truss element? K1 CO2

The truss elements are the part of a truss structure

linked together by point joint which transmits only axial
force to the element.
09 What are the h and p versions of finite element K1 CO2
method?

It is used to improve the accuracy of the finite

element method. In h version, the order of polynomial
approximation for all elements is kept constant and the
numbers of elements are increased. In p version, the
numbers of elements are maintained constant and the
order of polynomial approximation of element is
increased.
10 During discretization, mention the places where K1 CO2
it is necessary to place a node?

Concentrated load acting point Cross-section

changing point Different material interjections Sudden
change in point load

11 Differentiate between global and local axes. K1 CO2

Local axes are established in an element. Since it is

in the element level, they change with the change in
orientation of the element. The direction differs from
element to element.

Global axes are defined for the entire system. They

are same in direction for all the elements even though
the elements are differently oriented.
12 Write down the general finite element equation. K1 CO2

{F }=[K ]{u}

13 Define shape function. K1 CO2

Approximate relation φ (x,y) = N1 (x,y) φ1 + N2 (x,y)

φ2 + N3 (x,y) φ3 Where φ1, φ2, and φ3 are the values
of the field variable at the nodes N1, N2, and N3 are the
interpolation functions.
N1, N2, and N3 are also called shape functions because
they are used to express the geometry or shape of the
element.
11
0
 PART - A QUESTIONS AND ANSWERS
UNIT – I INTRODUCTION
S.N O QUESTION AND ANSWER K CO
LEVEL
14 What are the characteristic of shape functions? K1 CO1

It has unit value at one nodal point and zero value

at other nodal points. The sum of shape function is
equal to one.
15 Why polynomials are generally used as shape K1 CO1
function?

Differentiation and integration of polynomial are quit

easy. The accuracy of the result can be improved by
increasing the order of the polynomial. It is easy to
formulate and computerize the finite element equations

16 How do you calculate the size of the global K1 CO1

stiffness matrix?

Global stiffness matrix size = Number of nodes X

Degrees of freedom per node

17 State the properties of stiffness matrix. K1 CO1

▪ It is a symmetric matrix
▪ The sum of elements in any column must be equal
to zero
▪ It is an unstable element. So, the determinant is
equal to zero.
18 Write down the expression of shape function N K1 CO1
and displacement u for one dimensional bar
element.
U= N1u1+N2u2
N1= 1-X / l N2 = X / l

19 List the two advantages of post processing. K1 CO1

Required result can be obtained in graphical form.

Contour diagrams can be used to understand the
solution easily and quickly.
20 What is the difference between static and
dynamic analysis?
Static analysis: The solution of the problem does not
vary with time is known as static analysis.
Dynamic analysis: The solution of the problem varies
with time is known as dynamic analysis

11
1
 Part B Questions

S.No Question K Level CO

.
01 i. Derive the shape functions for One-Dimensional Linear K2 CO2
Bar element.

ii. Explain with suitable example the steps involved

in FEA
02 Consider the bar as shown below. Calculate the nodal K2 CO2
displacements, element stresses and support reactions. (E
= 200 x 109N/m2)

03 Calculate the nodal displacement, reaction forces and K2 CO2

elemental stresses for the bar shown in Figure. The bar is
fixed at the ends.

04 A composite wall consists of three materials as shown in K2 CO2

figure. The inside wall temperature is 200º C and the
outside air temperature is 50ºC with a convection coefficient
of 10W/m2K. Compute the temperature along the
composite wall.

40
S.No. Question K Level CO
05 A composite wall is made of three different materials. The K2 CO2
thermal conductivity of the various sections are k1 = 2
W/cm ˚C, k2 = 1 W/cm ˚ C, k3 = 0.2 = W/cm ˚C. The
thickness of the wall for the section is 1cm, 5cm and 4cm
respectively. Calculate the temperature values at nodal
points within the wall. Assume the surface area to unity.
The left edge of the wall is subjected to a temperature of
30˚C and the right side of the wall is at 10˚C.

07 Compute the temperature distribution in stainless steel fin K2 CO2

shown in figure. The region can be discretized into four
elements and five nodes.

08 The figure shows a uniform Aluminium fin of diameter 25 K2 CO2

mm. The root (left end) of the fin is maintained at a
temperature of T∞= 120 °C, convection takes place from
the lateral (circular) surface and the right (flat) edge of the
fin. Assuming k = 200 W /m °C, h = 1000 W /m2 °C and T
= 20 °C, Calculate the temperature distribution in the fin
using one dimensional element, considering two elements.

42
S.No. Question and Answer K Level CO
09 Consider a three-bar truss as shown in Fig. Take E = 2 x K2 CO2
105 N/mm2. Calculate the nodal displacements. Take Al =
2000 mm2 , A2 = 2500 mm2, A3 = 2500 mm2

250

10 K2 CO2
Determine the deflection in the beam, loaded as
shown in Fig., at the mid-span . Determine also the
reactions at the fixed ends. E = 200 GPa. Il = 20 x
10 m4, 12 = 10 x 10¯6 m4.

11 Find the eigen vales and eigen vectors of a cantilever beam K2 CO2
of length ‘L’, Youngs modulus ‘E’, Moment of Inertia ‘I’ and
mass density ‘℮’. Use two element idealization.
12 Find the eigen vales and eigen vectors of an axial bar of K2 CO2
length ‘L’, Youngs modulus ‘E’, Moment of Inertia ‘I’ and
mass density ‘℮’. Use two element idealization.

43
 SUPPORTIVE ONLINE CERTIFICATION COURSES

S.N Course name Agency Link

o
1 Basics of Finite NPTEL https://nptel.ac.in/courses/112/10
Element Analysis – I 4/112104193/
2 Finite Element NPTEL https://nptel.ac.in/courses/105/10
Analysis 5/105105041/
3 Finite Element Method NPTEL https://nptel.ac.in/courses/112/10
5/112105206/
4 Basics of Finite NPTEL https://nptel.ac.in/courses/112/10
Element Analysis – II 4/112104205/
5 Introduction to Finite NPTEL https://nptel.ac.in/courses/112/10
Element Method 6/112106135/
6 Finite Element Method NPTEL https://nptel.ac.in/courses/112/10
4/112104115/
7 Finite Element Method NPTEL https://nptel.ac.in/courses/112/10
3/112103295/
8 Finite Element Udemy https://www.udemy.com/course/fi
Analysis (FEA) Theory- nite-element-analysis-crash-
Crash Course course/
9 Finite element analysis Udemy https://www.udemy.com/course/fi
course using Ansys nite-element-analysis-course-
program using-ansys-program/
10 SimScale for Coursera https://www.coursera.org/projects
Engineering /simscale-for-engineering-
Simulations - FEA for simulations-fea-for-beginners
Beginners

11
5
REAL TIME APPLICATION IN DAY TO DAY LIFE AND TO
INDUSTRY

To calculate the deflection and bending moment of steel structures

To find the temperature distribution on bar element
Deflection of spring using potential energy method

11
6
 CONTENT BEYOND SYLLABUS
Effect of Self Weight on a Cantilever Beam
Introduction
This tutorial was completed using ANSYS 7.0 The purpose of the tutorial is to show the required steps
to
account for the weight of an object in ANSYS.
Loads will not be applied to the beam shown below in order to observe the deflection caused by the
weight of
the beam itself. The beam is to be made of steel with a modulus of elasticity of 200 GPa.
ANSYS Command Listing
/Title, Effects of Self Weight
/PREP7
Length = 1000
Width = 50
Height = 10
K,1,0,0 ! Create Keypoints
K,2,Length,0
L,1,2
ET,1,BEAM3 ! Set element type
R,1,Width*Height,Width*(Height**3)/12,Height !** = exponent
MP,EX,1,200000 ! Young's Modulus
MP,PRXY,1,0.3 ! Poisson's ratio
MP,DENS,1,7.86e-6 ! Density
LESIZE,ALL,Length/10, ! Size of line elements
LMESH,1 ! Mesh line 1
FINISH
/SOLU ! Enter solution mode
ANTYPE,0 ! Static analysis
University of Alberta ANSYS Tutorials -
www.mece.ualberta.ca/tutorials/ansys/CL/CIT/Density/Print.html
Copyright © 2001 University of Alberta
DK,1,ALL,0, ! Constrain keypoint 1
ACEL,,9.8 ! Set gravity constant
SOLVE
FINISH
/POST1
PLDISP,2 ! Display deformed shape
University of Alberta ANSYS Tutorials -
www.mece.ualberta.ca/tutorials/ansys/CL/CIT/Density/Print.html
Copyright © 2001 University of Alberta

11
7
 ASSESSMENT SCHEDULE

Unit Test I : March 2021

Internal Assessment Test I : March 2021

Unit Test II : April 2021

Internal Assessment Test II : April 2021

Model Examination : May 2021

11
8
 PRESCRIBED TEXT BOOKS & REFERENCE BOOKS

TEXT BOOKS:

T1) Reddy. J.N., “An Introduction to the Finite Element Method”, 3rd Edition, Tata
McGraw Hill, 2005

T2) Seshu, P, “Text Book of Finite Element Analysis”, Prentice-Hall of India Pvt. Ltd.,
New Delhi, 2007.

REFERENCES:

R1) Rao, S.S., “The Finite Element Method in Engineering”, 3rd Edition, Butterworth
Heinemann, 2004

R2) Logan, D.L., “A first course in Finite Element Method”, Thomson Asia Pvt. Ltd.,
2002

R3) Robert D. Cook, David S. Malkus, Michael E. Plesha, Robert J. Witt, “Concepts
and Applications of Finite Element Analysis”, 4th Edition, Wiley Student Edition,
2002.

R4) Chandrupatla & Belagundu, “Introduction to Finite Elements in Engineering”, 3rd

Edition, Prentice Hall College Div, 1990

11
9
 MINI PROJECT SUGGESTIONS

Temperature Distribution Analysis of Rectangular Plate

To find the nodal Temperature Distribution of this Rectangular Plate

Thermal conductivity of the plate, Kxx=401 W/(m-K).

Preprocessing: Problem Definition

1. Change jobname. File -> Change Jobname

2. Define element type: Preprocessor -> Element Type -> Add/Edit/Delete

Click on “Add”. The “Library of Element Types” menu appears, as shown. Scroll
down to highlight “Thermal Solid”, and “Quad 4 node 55” as shown. Click on “OK”,
then “Close”.

3. Define Material Properties:

Preprocessor -> Material Properties -> -Constant- Isotropic

“OK” for material set number 1, then enter 401 for KXX (thermal conductivity), then
click on “OK”. KXX is the only material property needed for this analysis.

4. Create a rectangular area:

Preprocessor -> -Modeling- Create -> -Areas- Rectangle -> By Dimensions

5. Specify mesh density controls.

We will specify numbers of element divisions along lines.

Choose:

Preprocessor -> -Meshing- Size Controls -> -Lines- Picked Lines

6. Mesh the rectangle to create nodes and elements.

Preprocessor -> -Meshing- Mesh -> Areas -> Mapped -> 3 or 4 Sided

A picking menu appears. Select “Pick All”. The rectangle will be meshed.

12
0
 MINI PROJECT SUGGESTIONS

Solution Phase: Assigning Loads and Solving

7.Apply temperatures around the edges:

Solution -> -Loads- Apply -> -Thermal- Temperature -> On Lines

Solution -> -Loads- Apply -> -Thermal- Temperature -> On Keypoints

8.Solve the problem: Solution -> -Solve- Current LS

Postprocessing:

9. Plot the temperature distribution:

General Postproc -> Plot Results -> -Contour Plot- Nodal Solu

12
1
 Thank you
Disclaimer:

This document is confidential and intended solely for the educational purpose of RMK Group of
Educational Institutions. If you have received this document through email in error, please notify the
system manager. This document contains proprietary information and is intended only to the
respective group / learning community as intended. If you are not the addressee you should not
disseminate, distribute or copy through e-mail. Please notify the sender immediately by e-mail if you
have received this document by mistake and delete this document from your system. If you are not
the intended recipient you are notified that disclosing, copying, distributing or taking any action in
reliance on the contents of this information is strictly prohibited.

115