0% found this document useful (0 votes)

338 views

Finite Element Analysis by G PDF

Uploaded by

Pacha Khan Khogyani

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

338 views

Finite Element Analysis by G PDF

Uploaded by

Pacha Khan Khogyani

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 349

Downloaded From : www.EasyEngineering.

net

ww
w.Ea
syE
ngi
nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

Finite Element Analysis

ww
w.E G. Lakshmi Narasaiah

asy
Prof. & Head Aeronautical Engineering Dept.
MLR Institute of Technology

En Dundigal, Hyderabad 500 043

Formerly Senior Manager,

gi
Corp. R&D, BHEL Vikasnagar, Hyderabad 500 093

nee
rin
g.n
e t
BSP BS Publications
==iiii 4-4-309, Giriraj Lane, Sultan Bazar,
Hyderabad - 500 095 - AP.
Phone: 040-23445688

Downloaded From : www.EasyEngineering.net

Copyright © 2008, by Author

No part of this book or parts thereof may be reproduced, stored in a retrieval

system or transmitted in any language or by any means, electronic, mechanical,
phototcopying, recording or otherwise without the prior written permission of
the Author.

ww "This book references ANSYS software, Release (Ed 5.5) is used by the author

w
and should not be considered the definitive source of information regarding the
ANSYS program. Please consult the ANSYS product documentation for the release

.Ea
of ANSYS with which you are working for the most current information ".

Published by :

BSPBS Publications
syE
ngi
4-4-309, Giriraj Lane, Sultan Bazar,
Hyderabad - 500 095 - A. P.
Phone: 040-23445688
Fax: 91+40-23445611 nee
e-mail: [email protected]

rin
g.n
Pr;ntedat
Adithya Art Printers
Hyderabad.
et

ISBN: 978-81-7800-140-1

Downloaded From : www.EasyEngineering.net

Contents
ww
w.E
asy
Chapter 1

E
Introduction

ngi
1.1 Design and Analysis of a Component.. ............................................................ 1
1-34

nee
1.2 Approximate Method vs. Exact Method ........................................................ 4
1.3 Weighted Residual Methods ........................................................................... 5

rin
1.4 Variational Method or Rayleigh - Ritz Method ............................................ 12

g
1.5 Principle of Minimum Potential Energy ........................................................ 22

.ne
1.6 Origin ofFEM .............................................................................................. 26

t
1.7 Principle ofFEM .......................................................................................... 27
1.8, Classification ofFEM ................................................................................... 31
1.9 Types of Analyses ........................................................................................ 32
1.10 Summary ....................................................................................................... 33
Objective Questions .................................................................................................. 33

Chapter 2
Matrix Operations 35-60
2.1 Types of Matrices ........................................................................................ 35
2.2 MatrixAlgebra ....................................................................... ~ ..................... 37

Downloaded From : www.EasyEngineering.net

Contents (xiv)

2.3 Detenninant .................................................................................................. 39

2.4 Inversion of a Matrix .................................................................................... 41
2.5 Methods of Solution of Simultaneous Equations .......................................... 41
2.5.1 By Inversion of the Coefficient Matrix ............................................. 42
2.5.2 Direct Methods ................................................................................. 44
2.5.3 Iterative Methods .............................................................................. 53

ww 2.6 Eigen Values and Eigen Vectors ................................................................... 55

2.7 Matrix Inversion Through Characteristic Equation ...................................... 59

w.E
2.8 Summary ....................................................................................................... 60

Chapter 3
a
Theory of Elasticity syE 61-84

ngi
3.1 Degrees of Freedom .................................................................................... 61
3.2 Rigid Body Motion ........................................................................................ 62

nee
3.3 Discrete Structures ...................................................................................... 62

rin
3.4 Continuum Structures ................................................................................... 62
3.5 Material Properties ....................................................................................... 63

g .ne
3.6 Linear Analysis ............................................................................................. 63
3.7 Non-linear Analysis ...................................................................................... 63
3.8 Stiffness and Flexibility ................................................................................. 64
3.9 Principle of Minimum Potential Energy ........................................................ 65
t
3.10 Stress and Strain at a Point .......................................................................... 65
3.11 Principal Stresses ......................................................................................... 68
3.12 Mohr's Circle for Representation of 2-D Stresses ...................................... 68
3.13 VonMises Stress ........................................................................................... 71
3.14 Theory of Elasticity ...................................................................................... 72
3.15 Summary ....................................................................................................... 82

Downloaded From : www.EasyEngineering.net

(xv) Contents

Chapter 4
Discrete (1-D) Elements 85-125
4.1 Degrees of Freedom of Different Elements ................................................ 85
4.2 Calculation of Stiffness Matrix by Direct Method ....................................... 86
4.3 Calculation of Stiffness Matrix by Variational Principle ............................... 88

ww
,
4.4 Transformation Matrix .................................................................................. 92
4.5 Assembling Element Stiffness Matrices ...................................................... 94

w.E
4.6 Boundary Conditions .................................................................................... 96
4.7 Beam Element Stiffness Matrix by Variational Approach ............................ 98

asy
4.8 General Beam Element .............................................................................. 100

E
4.9 Pipe Element ............................................................................................... 103

ngi
4.10 Summary ..................................................................................................... 104

Objective Questions ................................................................................................ 105

n eer
Solved Problems ...................................................................................................... 108

Chapter 5
Continuum (2-D & 3-D) Elements
ing 127-158
5.1 2-D Elements Subjected to In-plane Loads ................................................ 127.ne
5.2 Simplex, Complex and Multiplex Elements ................................................ 128
53 Stiffness Matrix of a CST Element ............................................................ 129
t
5.3.1 Stiffness Matrix of a Right Angled Triangle ................................... 131
5.4 Convergence Conditions ............................................................................. 133
5.5 Geometric Isotropy ..................................................................................... 136
5.6 Aspect Ratio ............................................................................................... 138
5.7 Inter-Element Compatibility ........................................................................ 139
5.8 2-D Elements Subjected to Bending Loads ................................................ 141
5.9 3-D Elements .............................................................................................. 143

Downloaded From : www.EasyEngineering.net

Contents (xvi)

5. JO Axi-symmetric Elements ............................................................................ 144

5.11 Summary ..................................................................................................... 146

Objective Que!t,tions .............................................................•.................................. 147

Solved Problen'!~ ...................................................................................................... 152

Chapter 6

ww Higher Order and Iso-Parametric Elements

6.1
159-199
Higher Order Elements .............................................................................. 159

w.E
6.2 Isoparametric Elements .............................................................................. 169
6.3 Stiffness Matrices of Some Iso-parametric Elements ................................ 170

asy
6.4 Jacobian ...................................................................................................... 182

En
6.5 Strain-displacement Relations .................................................................... 184
6.6 Summary ..................................................................................................... 188

gi
Objective Questions ................................................................................................ 189

nee
Solvetl Problems ...................................................................................................... 190

Chapter 7
rin
Factors Influencing Solution
g.n
201-236

e
7.1 Distributed Loads ....................................................................................... 201
7.2 Statically Equivalent Loads vs. Consistent Loads ...................................... 202
7.3 Consistent Loads for a Few Common Cases ............................................. 206 t
7.4 Assembling Element Stiffness Matrices .................................................... 209
7.5 Automatic Mesh Generation ...................................................................... 214
7.6 Optimum Mesh Model ................................................................................ 215
7.7 Gaussian Points & Numerical Integration .................................................. 216
7.8 Modelling Techniques ................................................................................ 220
7.9 Boundary Conditions for Continuum Analysis ............................................ 224
7.10 Transition Element ...................................................................................... 228

Downloaded From : www.EasyEngineering.net

(xvii) Contents

7.11 Substructuring or Super Element Approach ............................................... 230

7.12 Deformed and Undeformed Plots .............................................................. 231
7.13 Summary ..................................................................................................... 232

Objective Questions ..........•.....•.•.......•....•................................•.........•.•.......•........... 233

Chapter 8

ww Dynamic Analysis (undamped free vibrations) 237-253

8.1 Normalising Eigenvectors ........................................................................... 239

w 8.2 Modelling for Dynamic Analysis ................................................................ 240

.Ea
8.3 Mass Matrix ............................................................................................... 240

syE
8.4 Summary ..................................................................................................... 251
Objective Questions .•...•.•......••.•.....................•........................................................ 252

Chapter 9
ngi
Steady State Heat Conduction
nee 255-276

rin
9.1 Governing Equations ................................................................................... 255
9.2 1-0 Heat Conduction ................................................................................. 257
9.2.1
9.2.2
Heat Conduction Through a Wall .................................................... 258
g.n
Heat Transfer Through a Fin .......................................................... 268
9.3 2-D heat Conduction in a Plate .................................................................. 273
9.4 Summary ..................................................................................................... 275
et
Objective Questions ....•......••......••......••..•................................•...........................•... 276

Chapter 10
,
Design Validation and Other Types of Analysis 277-291
10.1 Compliance with Design Codes ................................................................. 277
10.2 Transient Heat Condition ............................................................................ 281
10.3 Buckling of Columns .................................................................................. 282

Downloaded From : www.EasyEngineering.net

Contents (xviii)

10.4 FatigueAnalysis ......................................................................................... 283

10.5 Creep Analysis ........................................................................................... 285
10.6 Damped Free Vibration .............................................................................. 287
10.7 Forced Vibration ......................................................................................... 288
10.8 Torsion of a Non-circular Rod .................................................................... 289

Chapter 11
ww Computational Fluid Dynamics 293-307
11.1
w.E
Introduction ................................................................................................. 293
11.2 Governing Equation .................................................................................... 294

asy
11.3 Finite Difference Method (FDM) .............................................................. 296
11.4 Elliptic Equations (or boundary value problems) ........................................ 297

E
11.5 Finite Volume Method (FVM) .................................................................... 305

ngi
11.6 FDM vs. FEM .......................................................................................... :. 307

Chapter 12
Practical Analysis Using a Software
n eer 309-329
12.1 Using a General Purpose Software ............................................................ 309
ing
12.2 Some Examples with ANSYS .................................................................... 311
.ne
Objective Questions ...••........•......................•.......•....••.......•..•.•.•....•.•..•••••.•....••..•.•••. 326
Answers .....•••.•.•..•...................•........•.•.•.•••••.....•....................•.....•...•.....•••••••••.•••••.••• 331

References for Additional Reading .•...•.•.••......•.................•............•••.•.••••.••••.••..• 333

t
Index ......................................................................................................................... 335

Downloaded From : www.EasyEngineering.net

CHAPTER 1
INTRODUCTION

ww
1.1.
w.E
DESIGN AND ANALYSIS OF A COMPONENT

Mechanical design is the design of a component for optimum size, shape, etc.,

asy
againstfailure under the application of operational loads. A good design should
also minimise the cost of material and cost of production. Failures that are

E
commonly associated with mechanical components are broadly classified as:

ngi
(a) Failure by breaking of brittle materials and fatigue failure (when
subjected to repetitive loads) of ductile materials.

nee
(b) Failure by yielding of ductile materials, subjected to non-repetitive
loads.
(c) Failure by elastic deformation.

rin
The last two modes cause change of shape or size of the component

g
rendering it useless and, therefore, refer to functional or operational failure.
Most of the design problems refer to one of these two types of failures.
Designing, thus, involves estimation of stresses and deformations of the .ne
components at different critical points of a component for the specified loads
and boundary conditions, so as to satisfY operational constraints ..
Design is associated with the calculation of dimensions of a component to
t
withstand the applied loads and perform the desired function. Analysis is
associated with the estimation of displacements or stresses in a component of
assumed dimensions so that adequacy of assumed dimensions is validated.
Optimum design is obtained by many iterations of modifYing dimensions of the
component based on the calculated values of displacements and/or stresses
vis-a-vis permitted values and re-analysis.
An analytic method is applied to a model problem rather than to an actual
physical problem. Even many laboratory experiments use models. A geometric
model for analysis can be devised after the physical nature of the problem bas
been understood. A model excludes superfluous details such as bolts, nuts,

Downloaded From : www.EasyEngineering.net

2 FINITE ELEMENT ANALYSIS

rivets, but includes all essential features, so that analysis of the model is no~
unnecessarily complicated and yet provides results that describe the actual
problem with sufficient an:uracy. A geometric model becomes a mathematical
",tHlel when its behaviour is described or approximated by incorporating
restrictions such as homogeneity, isotropy, constancy of material properties and
mathematical simplifications applicable for small magnitudes of strains and
rotations.
Several methods, such as method of joints for trusses, simple theory of

ww bending, simple theory of torsion, analyses of cylinders and spheres for axi-
symmetric pressure load etc., are available for designing/analysing simple\
components of a structure. These methods try to o~tain exact solutions of::

w.E
second order partial differential equations and are based on several assumptions.
on sizes of the components, loads, end conditions, material properties, likely !:~
deformation pattern etc. Also, these methods are not amenable for .,'

a
generalisation and effective utilisation of the computer for repetitive jobs.

syE
Strength of materials approach deals with a single beam member for •
different loads and end conditions (free, simply supported and fixed). In a space -,.
frame involving many such beam members, each member is analysed \f

ngi
independently by an assumed distribution of loads and end conditions. '

nee
For example, in a 3-member structure (portal frame) shown in Fig. 1.1, the
(horizontal) beam is analysed for deflection and bending stress by strength of
materials appnl1ch considering its both ends simply supported. The load and

rin
moment reactions obtained at the ends are then used to calculate the deflections
and "tresses in the two columns separately.
p

2
I
r-:t
P
t
t'+M2
R,
g
. .ne .~
R2

t
M, 2 M, M2
R, R2
3 + 3

/17'
FIGURE 1.1 Analysis of a simple frame by strength of materials approach

Simple supports for the beam imply that the columns do not influence slope
of the beam at its free ends (valid when bending stiffness of columns = 0 or the
column is highly flexible). Fixed supports for the beam imply that the slope of
the beam at its ends is zero (valid when bending stiffness of columns = 00 or the
column is extremely rigid). But, the ends of the horizontal beam are neither
simply sup .. ,}lied nor fixed. The degree of fixity or influence of columns on the
slope of the lJeam at its free ends is based on a finite, non-zero stiffness value.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODuCTION 3

Thus, the maximum deflection of the beam depends upon the relative stiffness
of the beam and the columns at the two ends ofthe beam.
For example, in a beam of length 'L', modulus of elasticity 'E', moment of
inertia 'I' subjected to a uniformly distributed load of'p' (Refer Fig. 1.2).

Deflection, 8 = 5 pL4 with simple supports at its two ends (case (a»
384EI

=~ with fixed supports at its two ends (case (b»

384 EI

ww i
w.E Case (a) : Simple supports Case (b) : Fixed supports

FIGURE 1.2 Deflection of a beam with different end conditions

a
If, in a particular case,
syE
L= 6 m, E= 2 x 1011 N/m2, Moment of inertia for beam IB =·0.4!t x W-4 m4
Moment of inertia for columns Ie = 0.48 x 10-4 m4 and distributed load
p=2 kN/m,
ngi
Dmax = 3.515 mm with simple supports at its two ends
and Dmax = 0.703 mm with fixed supports at its two ends
nee
whereas, deflection of the same beam, when analysed along with columns by
FEM,
Dmax = 1.8520 mm, when IB = Ie (Moments of inertia for beam & columns)
rin
= 1.0584 mm, when 5 IB = Ie
and = 2.8906 mm, when IB = 5 Ie
All the three deflection values clearly indicate presence of columns with
g .ne
finite and non-zero stiffness and, hence, the deflection values are in between
those of beam with fr~e ends and beam with fixed ends.
Thus, designing a single beam member of a frame leads to under-designing
t
if fixed end conditions are assumed while it leads to over-designing if simple
supports are assumed at its ends. Simply supported end conditions are,
therefore, normally used for a conservative design in the conventional approach.
Use of strength of materials approach for designing a component is, therefore,
associated with higher factor of safety. The individual member method was
acceptable for civil structures, where weight of the designed component is not a
serious constraint. A more accurate analysis of discrete structures with few
members is carried out by the potential energy approach. Optimum beam design
is achieved by analysing the entire structure which naturally considers finite

Downloaded From : www.EasyEngineering.net

4 FINITE ELEMENT ANALYSIS

stiffness of the columns, based on their dimensions and material, at it.~ ends.
This approach is followed in the Finite Element Method (FEM).

:1..2 ApPROXIMATE METHOD VS. EXACT METHOD

An analytical solution is a mathematical expression that gives the values of the
desired unknown quantity at any location of a body and hence is valid for an
infinite number of points in the component. However, it is not possible to obtain
analytical mathematical solutions for many engineering problems.

ww For problems involving complex material properties and boundary

conditions, numerical methods provide approximate but acceptable solutions
(with reasonable accuracy) for the unknown quantities - only at discrete or

w.E
finite number of points in the component.-Approximation is carried out in two
stages:
(a) In the formulation of the mathematical model, w.r.t. the physical

asy behaviour of the component. Example : Approximation of joint with

multiple rivets at the junction of any two members of a truss as a pin

E
joint, assumption that the joint between a column and a beam behaves

ngi
like a simple support for the beam,.... The results are reasonably
accurate far away from the joint.
(b) In obtaining numerical solution to the simplified mathematical model.

n eer
The methods usually involve approximation of a functional (such as
Potential energy) in terms of unknown functions (such as
displacements) at finite number of points. There are two broad
categories:
(i) ing
Weighted residual methods such as Galerkin method,
Collocation method, Least squares method, etc.
.ne
(ii) Variational method (Rayleigh-Ritz method, FEM). FEM is an
improvement of Rayleigh-Ritz method by choosing a
variational runction valid over a small element and not on the
entire component, which will be discussed in detail later. These
methods also use the principle of minimum potential energy.
t
(iii) Principle of minimum potential energy: Among all possible
kinematically admissible displacement fields (satistying
compatibility and boundary conditions) of a conservative system,
the one corresponding to stable equilibrium state has minimum
potential energy. For a component in static equilibrium, this
principle helps in the evaluation of unknown displacements of
deformable solids (continuum structures).
Some of these methods are explained here briefly to understand the historical
growth of analysis techniques.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 5

1.3 WEIGHTED RESIDUAL METHODS

Most structural problems end up with differential equations. Closed form
solutions are not feasible in many of these problems. Different approaches are
suggested to obtain approximate solutions. One such category is the weighted
residual technique. Here, an approximate solution, in the form y =l:Nj.C j for
i = 1 to n where Cj are the unknown coefficients or weights (constants) and N j
are functions of the independent variable satisfying the given kinematic

ww
boundary conditions, is used in the differential equation. Difference between the
two sides of the equation with known terms, on one side (usually functions of
the applied loads), and unknown terms, on the other side (functions of constants

w
Cj), is called the residual, R. This residual value may vary from point to point in

.Ea
the component, depending on the particular approximate solution. Different
methods are proposed based on how the residual is used in obtaining the best

syE
(approximate) solution. Three such popular methods are presented here.
(a) Galerkin Method
It is one of the weighted residual techniques. In this method, solution is

ngi
obtained by equating the integral of the product of function Nj and
residual R over the entire component to zero, for each Nj. Thus, the on'

nee
constants in the approximate solution are evaluated from the on'
conditions jNj.R.dx=O for i = I to n. The resulting solution may

rin
match with the exact solution at some points of the component and may
differ at other points. The number of terms N j used for approximating
the solution is arbitrary and depends on the accuracy desired. This
method is illustrated through the following examples of beams in g.n
bending.

Example 1.1
et
Calculate the maximum deflection in a simply supported beam, subjected to
concentrated load 'P' at the center of the beam. (Refer Fig. 1.3)

R J =P/2

t:=L/2~ ~I
FIGURE 1.3

Downloaded From : www.EasyEngineering.net

6 FINITE ELEMENT ANALYSIS

Solution
y = 0 at x ::::; 0 and x = L are the kinematic boundary conditions of the beam. So,
the functions Ni are chosen from (x - at.(x - b)q, with different positive integer
values for p and q; and a ::::; 0 and b = L.
(i) Model-I (I-term approximation): The deflection is assumed as
y(x) =N.c
with the function, N = x(x - L),

ww which satisfies the end conditions y = 0 at x = 0 and y = 0 at x = L.

The load-deflection relation for the beam is given by

w.E EI(d ]=M dx 2

M = (P/2).x
2Y

:s
asy
where for O:S x Ll2
and M = (P/2).x - P.[x - (Ll2)] = (P/2).(L - x) for Ll2:S x:S L
Thus, taking y = x.(x - L).c,

En d2 y
-=2c
dx 2

gin
and tbe residual oftbe equation, R = EI (~; ) - M = EI . 2c - M
eer
Then, the unknown constant 'c' in the function 'N' is obtained from
L
"2 L

IN.R.dx+ IN.R.dx=O i ng.

o L
2

(two integrals needed, since expression for M changes at x = L )

2
net
!x.(x- Lll.[EI.2C-(~).X}x + px.(x- L)l.[EI.2C-m(L-X)fx =0
2
5PL
=>c=--
64 EI
5PL
Therefore, y=x(x-L). - -
64E1
L 5 pI} - PL3
At x= 2' y=Ymax= - - - o
256 EI
r
51.2 EI

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 7

_PL3
This approximate solution is close to the exact solution of - -
48EI

obtained by double integration of EI (~;) = M = (~ l.x, with

appropriate end conditions.
(ii) Model-2 (2-term approximation): The deflection is assumed as
y(x) = NI.cl + N 2.C2

ww with the functions NI = x(x - L) and N2 = x.(x - L)2

which satisfy the given end conditions.

w.E
Thus, taking y = x.(x - L).cl + x.(x - Lfc2,
2y
d 2)
dx = 2cI + 2.(3x - 2L).C2

a (

syE
and the residual of the equation,
2y

ngi
d 2 ) - M = EI.[2cl + 2.(3x - 2L).c21 - M
R = EI. ( dx

where M = (P/2).x for 0::; x ::; Ll2

and M = (PI2).x - P.[x - (Ll2)] = (P/2).(L -
nee
x)
Then, the unknown constants 'CI' and 'C2' in the functions 'N 1' are
for Ll2::; x::; L

obtained from
L

fN ,.R.dx = }[x.(x - L1I{ EI [2e, + 2.(3x - 2L )e,l- (

r~ }+.
ing
+ px.(x - L)J{ EI.[2e, + 2.(3x - 21)e,l- (~ }(L - xl}dX = 0
. net
2

and f N, .R.dx = }Ix.(x - L)' l.{ EI!2e, + 2~3 x - 2L}.c,I-( ~ }.}dX

l{
+ flx,(x - L)' EI!2e, + 2.(3x - 2L)e,I-( ~}<L - Xl}dx = 0
2
Simplifying these equations, we get
5PL 75PL
2cI - C2.L = - - and 5cI - 4~.L = --
16EI 192EI

Downloaded From : www.EasyEngineering.net

8 FINITE ELEMENT ANALYSIS

Solving these two simultaneous equations, we get

55PL 25P
c( = 192E1 and C2 = 96E1

Thus, we get
55PL 2 25P
y=x(x-L). - - +x(x-L).--
192EI 96EI
at x=L/2 y=y PL3
= PL\-55+25) or
and

ww , max 4x192EI 25.6EI

Note: The bending moment M is a function of x. The exact solution of y
2

w.E should be a minimum of 3rd order ft. :+ion so that d ; = M is a function

ofx.
dx EI

asy
(b) Collocation Method
In this method, also called as the point collocation method, the residual is

E
equated to zero at 'n' select points of the component other than those at

ngi
which the displacement value is specified, where 'n' is the number of
unknown coefficients in the assumed displacement field, i.e., R( {c} ,Xi) = 0
for i = 1, .. n. It is also possible to apply collocation method on some select

nee
surfaces or volumes. In that case, the method is called sub ·domain
collocation method.
i.e.,
or
JR({c},x).dSj = 0
JR({c},x). dV k = 0
for j = 1, .. m
for k = 1, .. m rin
g .ne
These methods also result in 'n' algebraic simultaneous equation in 'n'
unknown coefficients, which can be easily evaluated.
The simpler of the two for manual calculation, point collocation method,
is explained better through the following example.

Example 1.2
t
Calculate the maximum deflection in a simply supported beam, subjected to
concentrated load 'P' at the center of the beam. (Refer Fig. 1.4)

RI =P/2

~U24 ~I
FIGURE 1.4

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 9

Solution
Y = 0 at x = 0 and x = L are the kinematic boundary conditions of the beam. So,
the functions Ni are chosen from (x - a)p.(x - b )q, with different positive integer
values for p and q; and a = 0 and b = L.
(i) Model-l (I-term approximation): The deflection is assumed as
y(x) = N.c
with the function N = x(x - L),

ww
which satisfies the end conditions y = 0 at x = 0 and y = 0 at x = L
The load-deflection relation for the beam is given by

w.E 2y
EI (d 2 ]
dx
=M

where
and a
.
M = (P/2).x

syE
M = (P/2).x - P.[x - (Ll2)] = (P/2).(L - x)

d 2y
for 0 ::; x ::; Ll2
for Ll2::; x::; L

Thus, takmg y = x.(x - L).c, -2

dx
n= 2c

~ gi ~
and the residual of the equation, R
nee
EI ( ::; ]- M EI ( ::; ]- ( }x ~
Then, the unknown constant 'c' in the function 'N' is obtained by
choosing the value of residual at some point, say x = Ll2, as zero. rin
i.e., R(C'X)=EI.2C-(~JX=O at x= ~ ~ c= :~ g .ne
Therefore,

L
PL
y = x(x - L).-
8EI
PL 3
t
At x='2' Y=Ymax =- 32EI
(ii) Model-2 (2-term approximation) : The deflection is assumed as
y(x) = N).c) + N 2 .C2
with the functions N) = x(x - L) and N2 = x. (x - Li,
which satisfy the given end conditions.
Thus, taking Y = x. (x - L ).c) + x. (x - Li.C2,
d2y
-2 = 2c) + 2.(3x - 2L}c 2
dx

Downloaded From : www.EasyEngineering.net

:1.0 FINITE EI.EMENT ANALYSIS

and the residual ofthe equation,

R= EI.(~)- M dx-
=: EI.[2cI +;.(3x-- 2L).C2J - M
\\ here M = (P/2).x for 0 ~ x ~ Ll2
and M = (P/2).x - P.Lx - (Ll2)] = (P/2).(L - x) for Ll2 ~ x~L
Then, the unknown constants 'c\' and 'C2' in the functions 'N,' are

ww ohtained from
R( {c} ,x) = EI.[2c\ + 2.(3x - 2L).c21 - (P/2).x = 0 at x = Ll4

w.E and

or
R( {c} ,x) = EI.[2c\ + 2.(3x - 2L).C2] - (P/2).(L - x)}
PL
4c I -5L.c 2 = - and
PL
4c I + L.c 2 = -
= 0 at x = 3L14

asy ~ cl
4EI

=-PL- and C2 = 0
4EI

E ~ Ymax
16EI

PlY
=- 64EI at x
ngi L
=2"

n
Choosing some other collocation points, say x
eer =L
3
and x = 2L ,
3

CI -L.c 2 = - -
12EI
PL
and c\ +0=--
12EI
PL
ing
=> c\
PL
=- -
12EI
and C2 = 0
.ne
At x
solution
L
=2"' Ymax
-PlY
= 48E1 ' which matches exactly with closed fonn t
(e) Least Squares Method
In this method, integral of the residual over the entire component is
. . . d'
mInImIZe . I.e., - aI = 0 fior I. = I , .. n,
ac,
where 1= fiR({a},x)]2.dx

This method also results in 'n' algebraic simultaneous equation in 'n'

unknown coefficients, which can be easily evaluated.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION

Example 1.3
Calculate the maximum deflection in a simply supported beam, subjected to
concentrated load 'P' at the center of the beam. (R~fer Fig. 1.5)
Solution
,
Again, y = 0 at x = 0 and y = 0 at x = L are the kinematic boundary conditions
of the beam. So, the functions Ni are chosen from (x - a)p.(x - b)q, with different
positive integer values for p and q.

ww R J = P/2

w.E l:=L/2~ ~I

asy FIGURE 1.5

l-term approximation: The deflection is again assumed as y(x) = N.C,
with the function N
En
= x(x - L),

gi
which satisfies the end conditions y = 0 at x = 0 and y = 0 at x = L
The load-deflection relation for the beam is given by

EI (ddx2y ) = M nee
rin
2

where M = (P/2).x for 0 ~ x ~ Ll2

and M = (P/2).x - P.[x - (Ll2)] = (P/2).(L - x)
2
for Ll2 ~ x ~ L
g.n
Thus, taking y = x.(x -

and the residual of the equation, R

L).c, d ;
dx
= 2c

~ EI (::;- ) ~ M ~ Ef . 2c ~ M
e t
Then, 1= fiR({c},X)]2 .dx and the constant 'c' in the function y(x) is obtained from
L

~=~
8a &1
J[[R({c},x)f -(P).X]dX+~ Y[R({C},X)]2 -(P).(L-X)]dX
0 2 &1 lL 2
l

PL -prJ L
~ c=-
8EI
and Ymax = 32EI at x ="2

Downloaded From : www.EasyEngineering.net

12 FINITE ELEMENT ANALYSIS

1.4 VARIATIONAL METHOD OR RAYLEIGH - RITZ METHOD

This method involves choosing a displacement field over the entire component,
usually in the form of a polynomial function, and evaluating unknown
coefficients of the polynomial for minimum potential energy. It gives an
approximate solution. Practical application of this method is explained here
through three different examples, involving
(a) uniform bar with concentrated load,

ww (b) bar of varying cross section·with concentrated load,

w.E
Example :1..4

asy
Calculate the displacement at node 2 of a fixed beam shown In Fig. 1.6,
SUbjected to an axial load 'P' at node 2.

En 2 ~ P
.1
I: gin U2 "14 Ll2
J

~x.u

eer
Solution
FIGURE 1.6
ing
Method - :I.
.ne
The total potential energy for the linear elastic one-dimensional rod with built-
in ends, when body forces are neglected, is

1 JEA (d)2
1t="2 d: dX-PU J
t
Let us assume u = a1 + a2x + a3x2 as the polynomial function for the
displacement field.
Kinematically admissible displacement field must satisfy the natural
boundary conditions
u = 0 at x = 0 which implies a1 = 0
and u = 0 at x = L which implies a2 = - a3 L

At X= ~ , u1 =a2(~)+a3(~J =-a3 ~

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION

Therefore,

1t=-
1 L
fEA -
(dU)2 dx-Pu J
20 dx

1 fEA{a
=- L (
2 + 2a 3 xYdx- P -a 3 -
L2)
20 4
1 L L2
=-EAa~ f{2x-LYdx+Pa 3 -
20 4

ww 2
=EAa 3 -+Pa 3 -
6
U
4
L2

w.E
For stable equilibrium, -
an:
aa 3
= 0 gives
-3P
a =--
3 4EAL

Displacement at node 2, asy

En
gi
It differs from the exact solution by a factor of
4 nee
~ . Exact solution is obtained
when a piece-wise polynomial interpolation is used in the assumption of
displacement field, u.
rin
Stress in the bar, 0' = E( :~ ) = E{a 2 + a 3x) = E{x - L)a3
g.n
= -3PE{x-L)
4EAL

= + ( ~ ) (:) at x = 0
3P{L-x)
4AL

and ~
- ( ) (:) and x = L
e t
or ± (%)(2:)
Due to the assumption of a quadratic displacement field over the system,
stress is found to vary along the length of the bar. However, stress is expected
to be constant (tensile from J to 2 and compressive from 2 to 3). Hence, the
solution is not exact.

Downloaded From : www.EasyEngineering.net

FINITE ELEMENT ANAI.YSIS

Method - 2
In order to compare the accuracy of the solution obtained by Rayleigh-Ritz
method, the beam is analysed considering it to be a system of two springs in
series as shown in Fig. 1.7 and using the stiffness of the axially loaded bar in
the potential energy function.

ww FIGURE 1. 7

w.EThe stiffness of each spring is obtained from

K =!= {a.A} 2AE

asy u [E{~)] L

E ngi
For equilibrium of this I-DOF system, nee
arc
-=2K.u 2 - P=O rin
or
8u 2

u -----
2 -
P
2K - 4AE
PL g .ne
Stress in the beam is given by,
t
The displacement at 2 by Rayleigh-Ritz method differs from the exact
solution by a factor of i,
4
while the maximum stress in the beam differs by a

factor of i.
2
The stresses obtained by this approximate method are thus on the

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 15

conservative side. Exact solution is obtained when a piece-wise polynomial

interpolation is used in the assumption of displacement field, u. The results are
plotted in Fig. 1.8.

Rayleigh-Ritz

,,"
,,- \--
/' "
/
ww Rayleigh-Ritz
Displacement variation

w .Ea
.../
.........

+ve
............
............
Exact solution

syE
......
............
............ -ve
......
Stress variation
ngi ............
......

FIGURE 1.8
nee
Method - 3
If the assumed displacement field is confined to a single element or segment of rin
the component, it is possible to choose a more accurate and convenient
polynomial. This is done in finite element method (FEM). Since total potential g.n
energy of each element is positive, minimum potential energy theory for the
entire component implies minimum potential energy for each element. Stiffness
matrix for each element is obtained by using this principle and these matrices
et
for all the elements are assembled together and solved for the unknown
displacements after applying boundary conditions. A more detailed presentation
of FEM is provided in chapter 4.
Applying this procedure in the present example, let the displacement field in
each element of the 2-element component be represented by u = ao + a).x. With
this assumed displacement field, stiffness matrix of each axial loaded element of
length (L/2) is obtained as

[K]=( 2~E)[ ~ 1 ~ 1] and {P} = [K] {u}

Downloaded From : www.EasyEngineering.net

16 FINITE ELEMENT ANALYSIS

The assembled stiffness matrix for the component with two elements is then
obtained by placing the coefficients of the stiffness matrix in the appropriate
locations as

I
e~E) [ -I 1 1:1 -I
-1 0 J{U ~
u, } {PI} P,
o 1 1 u3 P3

Applying boundary conditions UI = 0 and U3 = 0, we get

ww 2AE)2U 2 =P2 =P
( L
~ u 2 = PL
4AE

w.EThe potential energy approach and Rayleigh-Ritz method are now of only
academic interest. FEM is a better generalisation of these methods and extends
beyond discrete structures.

a syE
Examples of Rayleiglt-Ritz metltod, witlt variable stress in tlte members
These examples are referred again in higher order I-D truss elements, since they
involve stress or strain varying along the length of the bar.

Example :1..5
ngi
nee
Calculate displacement at node 2 of a tapered bar, shown in Fig 1.9, with area of
cross-section Al at node 1 and A2 at node 2 SUbjected to an axial tensile

rin
load 'P'.

---'p
g .ne
--+x.u
L ---+I
FIGURE 1.9
t
Solution

Different approximations are made for the displacement field and comparison
made, in order to understand the significance of the most reasonable
assumption.
(a) Since the bar is identified by 2 points, let us choose a first order
polynomial (with 2 unknown coefficients) to represent the displacement
field. Variation of A along the length of the bar adds additional
computation. Let A(x) = A) + (A2 - A).xlL

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 17

Let u = al + a2.x At x = 0, u = al = 0
du
Then, ,. -=a
dx 2
and u)-= -
a).L

Therefore,

1t=-
I JEA(X
L ~d)2
~ dx-Pu 2
20 dx

-AJ.~]a~
ww = !
20
JE[AI +(A2

=~E[ AIL+(A2 -AJ ~Ja~

L
dx - P.a 2·L

-
w.E=~E[(AI ~Ja~ P.a 2·L

asy +A 2} -P.a 2·L

En
For stable equilibrium, On = E[(AI + A2)' LJa 2 - P.L = 0
8a 2 2

gi
from which a, can be evalualed as, a, = [ (p
E AI +A2
2 nee
)]

Then,
P.L
u, =a,.L = [E.(A'2+ A2 )] rin
and
", =cr, ={:~)= a, =[(A, ; A,l]
g .ne
For the specific data of AI = 40
E

mm 2, 2
A2 = 20 mm and L = 200 mm, we
t
obtain,
6.667 P
u2 = and 0'1 = 0'2 = 0.0333 P
E
(b) Choosing displacement field by a first order polynomial gave constant
strain (first derivative) and hence constant stress. Since a tapered bar is
expected to have a variable stress, it is implied that the displacement field
should ,be expressed by a minimum of 2nd order polynomial. Therefore,
the solution is repeated with
At x = 0, u = al = 0

Downloaded From : www.EasyEngineering.net

18 FINITE ELEMENT ANALYSIS

2 du 2
Then, u=a2.x+a3'x; -=a 2 +2a 3.x and u2=a2.L+a).L
dx
1L
Therefore, 1[=- JEA(X). ~
(d )2 dx-Pu
2
20 dx

= ~ JE[ AI +(A2 -AI)' ~Ja2 +2a3x)2dx-P(a2.L+a3.L2)

o ,-

an
ww For stable equilibrium,
aa2
=0

6P
E
w.E
~ 3 a2 (3AI - A2) + 2a3 L (SAl - 2A 2) =

and
an
-=0 ~
3P
a2(SAI-2A2)+a3L(7AI-3A2)=-
E
asy
aa3
For the specific data of AI = 40 mm 2, A2 = 20 mm 2 and L = 200 mm, we
obtain,

E ngi= 0.0339 P
and
n
0'2 = E(a2 + 2a3L) = 0.03518 P

eer
(c) This problem can also be solved by assuming a 2nd order displacement

ing
function, satisi)ring a linearly varying stress along the length but with 2
unknown coefficients as
u = al + a2.x2 At x = 0, u = al = 0
2 du 2 .ne
t
Then, U = a2.x ; - =2a 2.x and U2 = a2.L
dx

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 19

expression for a2 as
PL
a -
2 - [E (AI; 3A
-;=-----~

2) ]

For the same set of data for A" A2 and L, we get

ww 2
u 2 =a 2·L = -
6P
E

w .Ea
0"1 =E.E1=E.(dU)
dx x;o
= E.(2a 2xt;o =0

syE
0"2 = E.(2a 2x)X;L = 2Ea 2L = 0.06 P

Displacement,

ngi
Displacement Stress at 1, Stress at 2,
Sr. No.
polynomial U2 0"1 0"2

1 u = al + a2'x 6.667 PIE 0.0333 P 0.0333 P

2
3
u = al + a2.x + a].x 2
u = al + a2.x2
6.652 PIE
6.0 PIE
nee
0.0339 P
0.0
0.03518 P
0.06 P

rin
4 OJ = PIA, (Exact solution) 0.025 P 0.05 P

These three assumed displacement fields gave different approximate

solutions. These are plotted graphically here, for a better understanding of the
differences. Exact solution depends on how closely the assumed displacement g.n
field matches with the actual displacement field.

End I Stress End 2

The most appropriate displacement field should necessarily include

constant term, linear term and then other higher order terms.

Downloaded From : www.EasyEngineering.net

20 FINITE ELEMENT ANALYSIS

Example .1.6
Calculate the displacement at node 2 of a vertical bar, shown in Fig. 1.10, due to
its self-weight. Let the weight be w N/m of length.

Solution
Since the load is distributed, varying linearly from zero at the free end to
maximum at the fixed end, it implies that the stress also varies linearly from the
free end to fixed end. As shown in the last example, therefore, a quadratic

ww displacement is the most appropriate. However, work potential needs to be

calculated through integration of product of varying load and corresponding
displacement, along the length.

w.E 2
\
\ Self-weight

a syE
x \

V
\\
Distribution

ngi
\
\
\

FIGURE 1.10

(a) Let u = a\ + a2.x + a3.x2. At x = 0, u\ = a\ = 0

nee
du
-=a 2 +2a 3 ·x
rin
dx
Since applied load is zero at the free end,

Strain at x = L, (dU) = a 2 + 2a 3.L = 0 => a 2 = - 2a 3L

g .ne
Then,
dx 2

u= a3.(x2-2Lx); and du =2a 3.(x-L)

dx
t
Let P = - w(L - x) acting along -ve x-direction
Therefore,

1t=-
1L
fEA (d)2
~ dx- fPdl L

20 dx 0

1L
=- f EA.[2a 3.(x-L)f dx -
20
fi- w (L-x)l2aAx-L)dx
L

= 2EA.ai·e 2w.a 3 ·e
3 3

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 21.

For stable equilibrium, an = 0 => a = ~

aa 3 3 2EA
2 -w.L2
Atx=L, u 2 =-a 3.L = - - -
2EA

Stress, 0" = E. du = E.[2a 3.(x - L)]= (w ).(x - L)

dx A

W~L )
ww At x = 0, 0"( = -(

And at x = L, 0"2 = 0
compressive

w.E
Example 1.7

asy
Calculate the displacement at node 2 of a vertical bar supported at both ends,
shown in Fig. 1.11, du~ to its self-weight. Let the weight be w N/m of length.

Solution
E ngi
As explained in the last example, a quadratic displacement is the most
appropriate to represent linearly varying stress along the bar.
(a) Let u = a( + a2.x + a3.x2 nee
At x = 0, u( "= al = 0

rin
2
At x = L, U2 = a2.L + a3.L = 0 => a2 = -a3L
Then, u = a3.(x 2- Lx)

and du =a 3.(2x - L)
dx
Let P = - w(L - x) acting along -ve x-direction
g .ne
\
t
x \ \ Self-weight
\~DiStribution

\
\
FIGURE 1.11

Downloaded From : www.EasyEngineering.net

22 FINITE ELEMENT ANALYSIS

Therefore,
1L (d)2
1[=- JEA ~ dx- JPdu
L

20 dx 0

1L L
=- JEA.[a 3 .(2x - L)]2dx + JW(L- x).a 3 ·(2x - L}dx
20 0

1 2e
=-EA.a 3 ·--2w.a 3 · -
e
ww 233
For stable equilibrium,
On =0 ~ a 3 = 2w
w.E 8a 3
L
EA
L2 L2

a At

Stress,
X=-

syE
2'

dx
u= -a 3 .-=-w.--
4

A
2EA

(J = E. du = E.[aA2x - L)]=(2W·).(2X - L)

At
and at
x = 0,
x = L, ngi
(J\ = -(2w.L/A)
0"2 = (2w.LlA)
compressive
tensile

1.5 PRINCIPLE OF MINIMUM POTENTIAL ENERGY nee

rin
The total potential energy of an elastic body (1[) is defined as the sum of total
strain energy (U) and work potential (W).
i.e., 1[=U+ W,

=(~)J(JEdV
g .ne
where

and
U

W=- JuTFdV- fuTTdS- ~>I PI

v s
t
Here, 'F' is the distributed body force, 'T' is the distributed surface force
and 'P.' are the concentrated loads applied at points i = I, .. n. One or more of
them may be acting on the component at any instant.
For a bar with axial load, if stress 0" and strain E are assumed uniform
throughout the bar,

U =(~)(JE v =(~)(JE A L =(~)((J A)(E L)=(~) F8 =(~)k 8 2 ..... (~1)

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 23

The work potential, W = - fq T f dV - fq T T ds - L u/ Pi .....(1.2)

for the body force, surface traction and point loads, respectively.
Application of this method is demonstrated through the following simple
examples. Since FEM is an extension of this method, more examples are
included in this category.

Example 1.8

ww
Calculate the nodal displacements m a system of four springs shown m
Fig. 1.12

w.E k,
F"q,

asy 2
k) k4

E ngi
FIGURE 1.12 Example of a 5-noded spring system

Solution n eer
The total potential energy is given by

1t = ( -1 klo l2 + -1 k 20 22 + -1 k 30 32 + -1 k 40 2) (
4 + - Flql - F3q3
) ing
2 2 2 2
where, qh q2, q3 are the three unknown nodal displacements.
.ne
At the fixed points
q4 = q5 = 0
Extensions of the four springs are given by
01=ql-q2; 02=q2
t
03 = q3 - q2; 04 = -q3
For equilibrium of this 3-DOF system,

aan
q,
=0 for i = 1 2 3
' ,
or

Downloaded From : www.EasyEngineering.net

24 FINITE ELEMENT ANALYSIS

These three equilibrium equations can be rewritten and expressed in matrix

form as

ww
w .Ea
Considering free body diagrams of each node separately, represented by the
following figures,

syE k,

ngi 2

nee
the equilibrium equations are k]o] = F]

rin
k202 - k]o] -k303 = 0
k3 03- kt04 = F3
g.n
These equations, expressed in terms of nodal displacements q, are similar to
the equations obtained earlier by the potential energy approach.

Example 1.9
et
Determine the displacements of nodes of the spring system (Fig. 1.13).

40 N/mm 30 N/mm

~: VN 2 ~N----+t----13VN ~ 4 60 N

~ VN vV:;----J~ 50 N/mm
60N/mm

FIGURE 1.13 Example of a 4-noded spring system

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 25

Solution
Total potential energy of the system is given by

1
1[= ( -klol2+-k
1 0 2+-k
1 0 2+-k
33
1 0 2) + (-F q3 -F q4 )
4 4 3
2 2 4
2 2 2 2
where q2, q3 , q4 are the three unknown nodal displacements.
At the fixed points

ww ql = q5 = 0
Extensions of the four springs are given by,

w 01 = q2 - q I

.Ea
; 02 = q3 - q2 ; 03 = q4 - q3 ; 04 = q3 - q5
For equilibrium of this 3-00F system,

07t =-6- for i = 2,3,4

8q,
syE
ngi
07t
or - = k lq2 -k 2(q3 -qJ=O ..... (a)
8q2
07t
- = k 2(q3 -q2)-k 3(q4 -qJ+k 4q3 -F3 =0
8q3 nee .....(b)

07t
- = k 3(q4 -q3)-F4 =0
rin .....(c)

g.n
8q4
These equilibrium equations can be expressed in matrix form as

[kJ +k,
-k2
0
-k2
k2 + k3 + k4
-k3
-~31{::}=F}
k3 q4 F4
et
[40+30
-~o]{::}=Fo}
-30
or -30 30+50+60
0 -50 50 q4 60
Substituting

F4
q4 =-+q3 = (60)
- +q3 =1.2+q3 from eq. (c)
k3 50

Downloaded From : www.EasyEngineering.net

26 FINITE ELEMENT ANALYSIS

and k 2q3 _ 30q3 _ 3q3 from eq. (a)

q2=k 1 +k 2 -(30+40)- 7
in eq. (b), we get

k2 [q3 - 3~3 ]-k3 [1.2+q3]+k 4.q3 -F3 =0

which gives, q3 = 2.0741 mm

ww
and then, q2=0.8889mm; q4=3.2741 mm

WHYFEM?

w.E
The Rayleigh-Ritz method and potential energy approach are now of only
academic interest. For a big problem, it is difficult to deal with a polynomial

asy
having as many coefficients as the number of OOF. FEM is a better
generalization of these methods and extends beyond the discrete structures.

E
Rayleigh-Ritz method of choosing a polynomial for displacement field and
evaluating the coefficients for minimum potential energy is used in FEM, at the

ngi
individual element level to obtain element stiffness matrix (representing load-
displacement relations) and assembled to analyse the structure.

1.6 ORIGIN OF FEM n eer

The subject was developed during 2nd half of 20 th century by the contribution of

ing
many researchers. It is not possible to give chronological summary of their
contributions here. Starting with application of force matrix method for swept

R.Courant, MJ.Turner, R.W.Clough, .ne

wings by S. Levy in 1947, significant contributions by J.H.Argyris,
H.L.Langhaar, R.J.Melosh,
J.S.Przemieniecki, O.C.Zienkiewicz, J.L.Tocher, H.C.Martin, T.H.H.Pian,
R.H.Gallaghar, J.T.Oden, C.A.Felippa, E.L.Wilson, K.J.Bathe, R.O.Cook etc ...
lead to the development of the method, various elements, numerical solution
techniques, software development and new application areas.
t
Individual member method of analysis, being over-conservative, provides a
design with bigger and heavier members than actually necessary. This method
was followed in civil structures where weight is not a major constraint. Analysis
of the complete structure was necessitated by the need for a better -estimation of
stresses in the design of airplanes with minimum factor of safetyiand, hence,
minimum weight), during World War-II. Finite element method, popular as
FEM, was developed initially as Matrix method of structural analysis for
discrete structures like trusses and frames.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 27

FEM is also extended later for continuum structures to get better estimation
of stresses and deflections even in components of variable cross-section as well
as with non-homogeneous and non-isotropic materials, allowing for optimum
design of complicated components. While matrix method was limited to a few
discrete structures whose load-displacement relationships are derived from basic
strength of materials approach, FEM was a generalisation of the method on the
basis of variational principles and energy theorems and is applicable to all types
of structures - discrete as well as continuum. It is based on conventional theory

ww
of elasticity (equilibrium of forces and Compatibility of displacements) and
variational principles.

w.E
In FEM, the entire structure is analysed without using assumptions about the
degree of fixity at the joints of members and hence better estimation of stresses

asy
in the members was possible. This method generates a large set of simultaneous
equations, representing load-displacement relationships. Matrix notation is
ideally suited for computerising various relations in this method. Development

E ngi
of numerical methods and availability of computers, therefore, helped growth of
matrix method. Sound knowledge of strength of materials, theory of elasticity
and matrix algebra are essential pre-requisites for understanding this subject.

1..7 PRINCIPLE OF FEM nee

rin
In FEM, actual component is replaced by a simplified model, identified by a
finite number ~f elements connected at common points called nodes, with an
g
assumed behaviour or response of each element to the set of applied loads, and
evaluating the unknown field variable (displacement, temperature) at these .ne
finite number of points.

Example 1.10
t
The first use of this physical concept of representing a given domain as a
collection of discrete parts is recorded in the evaluation of 1t from superscribed
and inscribed polygons (Refer Fig. 1.14) for measuring circumference of a
circle, thus approaching correct value from a higher value or a lower value
(Upper boundsiLower bounds) and improving accuracy as the number of sides
of polygon increased (convergence). Value of 1t was obtained as 3.16 or ]0112
by 1500 BC and as 3.1415926 by 480 AD, using this approach.

Downloaded From : www.EasyEngineering.net

28 FINITE ELEMENT ANALYSIS

ww
w.E Case (a) Inscribed polygon Case (b) Superscribed polygon

asy
FIGURE 1.14 Approximation of a circle by an inscribed and a superscribed polygon

Perimeter of a circle of diameter 10 cm = nO = 31.4 cm

E ngi
Case-A : The circle of radius 'r' is now approximated by an inscribed
regular polygon of side's'. Then, using simple trigonometric concepts,

nee
the length of side's' of any regular n-sided polygon can now be obtained
as s = 2 r sin (360/2n). Actual measurements of sides of regular or irregular

rin
polygon inscribed in the circle were carried out in those days, in the absence of
trigonometric formulae, to find out the perimeter.
With a 4-sided regular polygon,
With a 8-sided regular polygon, g
perimeter = 4 s = 28.284
perimeter = 8 s = 30.615
With a 16-sided regular polygon, perimeter = 16 s = 31.215 .ne
approaching correct value from a lower value, as the number of sides of the
inscribed polygon theoretically increases to infinity. t
Case-B : The same circle is now approximated by a superscribed polygon of
side's', given by
s = 2 r tan (360/2n)
Then, With a 4-sided regular polygon, perimeter = 4 s = 40
With a 8-sided regular polygon, perimeter = 8 s = 33.137
With a 16-sided regular polygon, perimeter = 16 s = 31.826
approaching correct value from a higher value, as the number of sides of the
circumscribed polygon theoretically increases to infinity.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 29

A better estimate of the value of 1t (ratio of circftmference to diameter) was

found by taking average perimeter of inscribed and superscribed polygons,
approaching correct value as the number of sides increases.
Thus with a 4-sided regular polygon, perimeter = (40+28.284) / 2 = 34.142
With a 8-sided regular polygon, perimeter = (33.137+30.615) /2 = 31.876
With a 16-sided regular polygon, perimeter = (31.826+31.215) / 2 = 31.52C

Example 1.11

ww
In order to understand the principle of FEM, let us consider one more example,
for which closed form solutions are available in every book of 'Strength of

w
materials'. A common application for mechanical and civil engineers is the
calculation of tip deflection of a cantilever beam AB of length 'L' and subjected

.Ea
to uniformly distributed load 'p'. For this simple case, closed form solution is
obtained by integrating twice the differential equation.

syE
2
EJd y =M
dx 2
and applying boundary conditions

ngi
y = 0 and dy = 0 at x = 0 (fixed end, A),

_
dx
pL4
we get, at x - L, Ymax = - - nee
8EI

rin
This distributed load can be approximated as concentrated loads
(Ph P2, ••• PN) acting on 'N' number of small elements, which together form the
total cantilever beam. Each of these concentrated loads is the total value of the
distributed load over the length of each element (PI = P2 = ... = PN = P L / N),
g.n
acting at its mid-point, as shown in Fig 1.15. Assuming that the tip deflection
(at B) is small, the combined effect of all such loads can be obtained by linear
superposition of the effects of each one of them acting independently. We will
et
again make use of closed form solutions for the tip deflection values of a
cantilever beam subjected to concentr1:lted loads at some intermediate points.
1\ P
~- (
B B

Case 1 : Cantilever with Case 2 : Cantilever with

distributed load many conc. loads.

FIGURE 1.15 Cantilever beam with distributed load approximated by many

concentrated loads

Downloaded From : www.EasyEngineering.net

30 FINITE ELEMENT ANALYSIS

Tip deflection of the cantilever when subjected to concentrated load PJ at a

distance LJ from the fixed end is given by

YB = (Y)J + (dY) .(L - LJ)

dx J

Closed form solutions for (y)J and (dy/dx)J can be obtained by integrating the
beam deflection equation with appropriate boundary conditions, as
_p Lj
YJ - J 3 EI

ww dY) _ P L~
( dx J - J 2 EI

w.EDeflection at B, YB, due to the combined effect of all the concentrated loads
along the length of the cantilever can now be obtained, by linear superposition,
as

asy YB = [(Y), + (dy/dx), (L - L,)] + [(Y)2 + (dy/dx)2 (L - L 2)] + ... +

[(Y)N + (dy/dx)N (L - L N)]

E
The results obtained with different number of elements are given in the table

ngi
below, for the cantilever of length 200 cm, distributed load of 50 N/cm and
EI = 109 N cm 2 •

S.No.
1
No. of elements
3
n eer
Tip displacement, YB (cm)
9.815
2
3
4
5
9.896
9.933 ing
4 6 9.954
.ne
t
5 8 9.974
6 to 9.983
7 15 9.993
8 20 9.996

The exact value obtained for the cantilever with uniformly distributed load,
from the closed form solution, is YB = 10.0 cm. It can be seen, even in this
simple case, that the tip deflection value approaches true solution from a lower
value as the number of elements increases. In other words, the tip deflection
value even with a small number of elements gives an approximate solution.
This method in this form is not useful for engineering analysis as the
approximate solution is lower than the exact value and, in the absence of
error estimate, the solution is not practically usefuL

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 3:1

FEM approach, !lascd on minimum potential energy theorem, convergel' to

the correct soluti(Jn frtlm a hig/rer value as the numhcr of elements in the
model increases. While the number of elements used in a model is selected by
the engineer, based on the required accuracy of solution as well as the
availability of computer with sufficient memory, FEM has become popular as it
ensures usefulness of the results obtained (on a more conservative side) even
with lesser number of elements.
Finite Element Analysis (FEA) based 6n FEM is a simulation, not reality,
applied to the mathematical model. Even very accurate FEA may not be good

ww
enough, if the mathematical model is inappropriate or inadequate. A
mathematical model is an idealisation in which geometry, material properties,

w.E
loads and/or boundary conditions are simplified hased on the analyst's
understanding of what features are important or unimportant in obtaining the
results required. The error in solution can result from three different sources.

asy
Modelling error - associated with the approximations made to the real
problem.

E
Dlscretrsation error - associated with type, size and shape of finite elements
used to represent the mathematical model; can be reduced by modifYing mesh.

ngi
Numerical error - based on the algorithm used and the finite precision of
numbers used to represent data in the computer; most softwares use double
precision for reducing numerical error.
n eer
It is entirely possible for an unprepared software user to misunderstand the
problem, prepare the wrong mathematical model, discretise it inappropriately,

ing
fail to check computed output and yet accept nonsensical results. FEA is a
solution technique that removes many limitations of classical solution

.ne
techniques; but does not bypass the underlying theory or the need to devise a
satisfactory model. Thus, the accuracy of FEA depends on the knowledge of

t
the analyst in modelling the problem correctly.

1.8 CLASSIFICATION OF FEM

The basic problem in any engineering design is to evaluate displacements,

stresses and strains in any given structure under different loads and boundary
conditions. Several approaches of Finite Element Analysis have been developed
to meet the needs of specific applications. The common methods are:
Displacement method - Here the structure is subjected to applied loads and/or
specified displacements. The primary unknowns are displacements, obtained by
inversion of the stiffness matrix, and the derived unknowns are stresses and
strains. Stiffness matrix for any element can be obtained by variational
principle, based on minimum potential energy of any stable structure and,
hence, this is the most commonly used method.

Downloaded From : www.EasyEngineering.net

32 FINITE ELEMENT ANALYSIS

Force method - Here the structure is subjected to applied loads and/or

specified displacements. The primary unknowns are member forces, obtained
by inversion of the flexibility matrix, and the derived unknowns are stresses and
strains. Calculation of flexibility matrix is possible only for discrete structural
elements (such as trusses, beams and piping) and hence, this method is limited
in the early analyses of discrete structures and in piping analysis
Mixed method - Here the structure is subjected to applied loads and/or
specified displacements. The ·method deals with large stiffness coefficients as

ww well as very small flexibility coefficients in the same matrix. Analysis by this
method leads to numerical errors and is not possible except in some very special

w.E
cases.
Hybrid method - Here the structure is subjected to applied loads and stress
boundary conditions. This deals with special cases, such as airplane dour frame

asy
which should be designed for stress-free boundary, so that the door can be
opened during flight, in cases of emergencies.

E
Displacement method is the most common method and is suitable for solving

ngi
most of the engineering problems. The discussion in the remaining chapters is
confined to displacement method.

1.9 TYPES OF ANALYSES nee

rin
Mechanical engineers deal with two basic types of analyses for discrete and

g
continuum structures, excluding other applicati.on areas like fluid flow,
electromagnetics. FEM helps in modelling the component once and perform
both the types of analysis using the same model. .ne
(a) Thermal analysis - Deals with steady state or transient heat transfer by
conduction and convection, both being linear operations while radiation
is a non-linear operation, and estimation of temperature distribution in the
t
component. This result can form one load condition for the structural
analysis.
(b) Structural analysis - Deals with estimation of stresses and
displacements in discrete as well as continuum structures under various
types of loads such as gravity, wind, pressure and temperature. Dynamic
loads may also be considered.

Downloaded From : www.EasyEngineering.net

CHAPTER 1 INTRODUCTION 33

1.:1.0 SUMMARY

• Finite Element Method, popularly known as FEM, involves

analysis of the entire structure, instead of separately considering
individual elements with simplified or assumed end conditions. It
thus helps in· a more accumte estimate of the stresses in the
members, facilitating optimum design.
• FEM involves idealizing the given component into a finite

ww number of small elements, connected at nodes. FEM is an

extension of Rayleigh-Ritz method, eliminating the difficulty of
dealing with a large polynomial representing a suitable

w.E displacement field valid over the entire structure. Over each finite
element, the physical process is approximated by functions of

asy
desired type and algebraic equations, which relate physical
quantities at these nodes and are developed using variational
approach. Assembling these element relationships in the ·proper

•
E
way is assumed to approximately represent relationships of

ngi
physical quantities of the entire structure.
FEM is based on minimum potential energy theorem. It

n
approaches true solution from a higher value, as the number of

eer
elements increases. Thus, it gives a conservative solution even
with a small number of elements, representing a crude
idealisation.
ing
OBJECTIVE QUESTIONS
.ne
1. The solution by FEM is
(a) always exact
(c) sometimes exact
(b) mostly approximate
(d) never exact
t
2. Discrete analysis covers
(a) all 2-D trusses & frames
(b) all 3-D trusses & frames
(c) all 2-D and 3-D trusses & frames
(d) no tfUsses; only frames

Downloaded From : www.EasyEngineering.net

34 FINITE ELEMENT ANALYSIS

3. FEM is a generalization of
(a) Rayleigh-Ritz method (b) Weighted residual method
(c) Finite difference method (d) Finite volume method
4. Variational principle is the basis for
(a) Displacement method (b) Weighted residual method
(c) Finite difference method (d) Finite volume method
5. Displacement method is based on minimum

ww (a) potential energy

(b) strain energy

w.E (c) complementary strain energy

(d) work done

a
6.

7.
Hybrid method is best suited for problems with prescribed

syE
(a) displacements (b) forces (c) stresses
Primary variable in FEM structural analysis is
(d) temperature

(a) displacement
ngi
(b) force (c) stress (d) strain
8.

nee
Stress boundary conditions can be prescribed in
(a) displacement method (b) hybrid method

9.
(c) force method
Prescribed foads catt40rm input data in rin
{d) mixed method

(a) displacement method

(c) force method
(b) hybrid method
(d) mixed method g .ne
10. Stiffness matrix approach is used in
(a) displacement method
(c) force method
(b) stress method
(d) mixed method
t
11. Displacement method of FEM for structural analysis gives
(a) stiffness matrix (b) flexibility matrix
(c) conductance matrix (d) mixed matrix
12. Flexibility matrix approach is used in
(a) displacement method (b) stress method
(c) force method (d) mixed method

Downloaded From : www.EasyEngineering.net

CHAPTER 2
MATRIX OPERATIONS

ww
w.E
asy
FEM deals with a large number of linear algebraic equations, which can be
conveniently expressed in matrix form. Matrices are also amenable to computer
programming. Knowledge of matrix algebra is essential to understand and solve

E
problems in FEM. A brief review of matrix algebra is given below for a good
understanding of the remaining text.
ngi
A matrix is a group of m x n numbers (scalars or vectors) arranged in

n
III number of rows and n number of columns. It is denoted by [A] or [A]mxn

eer
IIldicating matrix of 'm' rows and 'n' columns or matrix of order m x n. Each
dc:ment of the matrix is identified by a'J' located in the i'h row andj'h column.
all a 12 a 13 a l4 a ln
ing
.ne
a 21 a 22 a 23 a 24 a 2n
[A]= a 31 a 32 a 33 a 34 a 3n ••.••(2.1)

amI a m2 a m3 a m4 a mn t
2.2- TYPES OF MATRICES

Based on the number of rows and columns as well as on the nature of elements,
some matrices are distinctly identified as follows. The symbol ''1/' indicates
'for all'.
(a) Square matrix is a matrix of same number of rows and columns and is
usually indicated as [Aln x n or square matrix of order n i.e., matrix with
m=n.
The elements from left top to right bottom .i.e., all .... an n fonn the leading
diagonal.

Downloaded From : www.EasyEngineering.net

36 FIN IT EEL E MEN TAN A I. Y SIS

(b) Row matrix is a matrix of only one row and is usually indicated as {A} T
i.e., matrix with m = I.
(c) Column matrL~ (or vector) is a matrix of only one column and is usually
indicated as {A} i.e., matrix with n = I.
(d) Bamled matrix i... a square matrix whose off-diagonal elements beyond half
bandwidth from the diagonal element are all zeroes i.e., a'l = 0 V j > i + h
and V j < i - h,

ww where h + I is half bandwidth and 2h + I is the bandwidth

For example, a 5 x 5 matrix with half bandwidth of2 is given by

w .Ea [A] =
all
a 21
a 31
a l2
a 22
a 32
a 13
a 23
a 33
0
a 24
a 34 a 35
0
0
•••.• (2.2)

syE 0
0
a 42
0
a 43
a 53
a 44
a 54
a 45
a 55

I.e., alJ = 0
ngi
(e) Diagonal matrix is a banded matrix of bandwidth equal to 1
Vi :t:j or Banded matrix with h ~ 0 .•...(2.3)

nee
For example, a diagonal matrix of order 5 is given by

all 0 0 0 0

[A] =
0
0
a 22
0
0
a 33
0
0 rin 0
0 ...•.(2.4)
0
0
0
0
0
0
a 44
0 a 55
0
g.n
(f) Identity matrix is a diagonal matrix with each diagonal element equal to
one and is usually indicated by [I] for any square matrix of order n.
et
i.e., all = I Vi = j and alj = 0 Vi :t: j ..•••(2.5)
(g) Transpose of matrix, written as [A]T, is the matrix with each element all of
matrix [A] written in l'
row and ilh column or interchanging its rows and
columns.
i.e., alJ of matrix [A] = all of matrix [A]T

It can be seen that ....•(2.6)

(h) Symmetric matrix, defined only for square matrices has elements
symmetric about its leading diagonal i.e., all = all
It can be seen that [Af = [A] ..... (2.7)

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 37

(i) Skew-symmetric matrix, also defined for square matrices only, has
elements equal m magnitude but opposite in sign, about the leading
diagonal,
i.e., alj = - ajl and elements on leading diagonal all = 0
It can be seen that [A]T = - [A] •...•(2.8)

2.2 MATRIX ALGEBRA

ww
Common mathematical operations on matrices are
(a) Addition of matrices A and B (defined only when A and B are of same

w.E
order) is addition of corresponding elements of both the matrices. If C is the
resulting matrix,

asy [C] = [A] + [B] or

(b) Multiplication of a matrix A by a scalar quantity's' is multiplication of
••••• (2.9)

i.e.,
E
each element of matrix [A] by the scalar s
s[A] = [s alj]

nG)gi [ 1 .....(2.10)
Same holds good for division of a matrix A by a scalar's', since division by

[~t
's' is equivalentto multiplication by
nee
i.e., as"

ri
(c) Matrix multiplication is defined only when number of columns in the first
matrix A equals number of rows in the second matrix B and is, usually, not
commutative. Element Clj is the sum of the products of each element of
i'h row of matrix A with the corresponding element of the jlh column of
ng.
matrix B.
i.e.,
where
[A]m x0 [B]o xp = [C]m xp
alj bjk = Cik V i = 1, m; j = 1, n; k = 1, P ....•(2.11)
net
It can be seen that for any matrix [A] and identity matrix [I], both of order
n x n,
LAUI] = [IUA] = [A] •••••(2.12)
(d) Transpose ofproduct of matrices
([AUBUC])T = [Cf [Bf [A]T ..... (2.13)
(e) Differentiation of a matrix, each element of which is a function of x, is
differentiation of each element w.r.t. x

i.e.,
d[B(x)] = [dbij(X)] ..... (2.14)
dx dx

Downloaded From : www.EasyEngineering.net

38 FINITE ELEMENT ANAL YSIS

(t) Integration of a matrix, each element of which is a function of X, is integration

of each element w.r.t. x,

i.e., ..... (2.15)

Matrix algebra differs from algebra of real numbers in some respects

For example, matrix multiplication is, in general, not commutative even if
both the matrices are of the same order and square,
I.e., [A][8]:;l: [8][A]

ww Matrix algebra also differs from vector algebra in some respects.

Vector product is defined as a dot product (giving a scalar result) or a cross

w.E
product (giving a vector, perpendicular to both the vectors, as the result)
whereas matrix multiplication is defined in only one way.
Trace of a matrix, defined for any square matrix of order m, is a scalar quantity

a
and is equal to the sum of elements of its leading diagonal.

I.e., syE
Tr[A]=a" +a 22 + ... +a mm = La"
m
.....(2.16)

ngi 1='
Minor Mlj of a square matrix [A] is the determinant (defined in Section 2.3) of

nee
sub matrix of [A] obtained by deleting elements of ith row and jth column from
[A].
Cofactor C IJ ofa square matrix [A] = ( - l t J Mlj
Matrix [C] with elements C IJ is called Cofactor matrix
rin
g
Adjoint of matrix [A] is defined as [C]T and is written as Adj [A] or [AIJ ]
Matrix approach is followed in FEM for convenience of representation.
Conventionally, stress components, strain components, displacement .ne
components etc., are called stress vector, strain vector, displacement vector etc.;
but they actually represent column matrices. Each component of these column
matrices is a vector quantity representing the corresponding quantity in a
particular direction. Hence, matrix algebra, and not vector algebra, is applicable
t
in their case.
Quadratic form - If [A] is an (n x n) matrix and {x} is a (n x I) vector, then
the scalar product {x} T[A]{x} is called a quadratic form (having terms like X,2,
X2 2, ... and X,X2, X,X3, ... ).

Example 2.1

Express q =x j -6X2 +3x~ +5XjX2 in the matrix form !XTQX+CTX

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 39

Solution
From the data, it can be concluded that vector x has two elements XI and X2.
Therefore, matrix Q is of order 2 x 2 and C must be a vector of order 2.

1 T
-x T
Qx+C l [ XI X2 l[qll
x =-
2 2 q21

ww
w.E
Comparing these coefficients with those of the given expression
XI- 6X2 + 3Xl2 + 5XIX2
we get
asy
CI = I; C2= -6; qll = 6;

+3x~ +5X IX2 =~[XI

q22 = 0; qJ2 + q21 = 10 or qJ2 = q21 = 5

:.XI -6X2
E 2
x21[6 5]{XI}+[1

ngi5 0 x2
-61{xl}
x2

2.3 DETERMINANT

n eer
It is a scalar quantity defined only for square matrices and is written as A or
t
I I

ing
det[ A]n x n. It is defined as the sum of the products (-1 j a.J M.j where aiJ are the
elements along anyone row or column and M.j are the corresponding minors.
For example, if[A] is a square matrix of order 2,

.ne
or
expanding by 1st row

expanding by 1st column

t
If [A] is a square matrix of order 3,

all a l2 a l3
a 21 a 22 a 23 = (-1)1+1 all All + (_1)1+2 al2 AI2 + (_1)1+3 al3 AI3
a 31 a 32 an

_ expanding by 1st row

Downloaded From : www.EasyEngineering.net

40 FINITE ELEMENT ANAl YSIS

-- (1)1+1
- all a 22 a 23 I +(- 1)1+2 an a 21
a 32 a 33 a 31

= all (an.a33 - a23.a32) - al2 (a2\.a33 - a23.a31) + au (a2\.an - an.a31)

It can also be expressed as

all a l2 a 13
an = (-1)1+1 all All + (-li+ 1 a21 A21 + (-I )311 a31 A31 expanding

ww a 21
a 31
a 22
a 32 a 33

w by 1sl column

.Ea
(-l)I+l all an
an
a23I+(_1)2+la21 a l2
a 33 a 32
aI31+(_1)31Ia31 a\2
a 33 an
a131
an

=
and so on.
syE
all (an.a33 - an.an) - a21 (a\2.a33 - al3.a32) + a31 (al2.a23 - al3.an)

(a)
It can be seen that,
ngi
Det ( [A] [8Hc] ) = Det [A] . Det [8] . Det [C] ..... (2.17)
(b)
nee
If[L] is a lower triangular matrix with 1,,= 1 and III = 0 Vi> j ,
Det [L]
=1
rin
= IIIC II + 112C12 + I 13C 13 = IIIC II = III .( In. /" - 0./32 ) = /11' /22, 133

(c)
(d)
In the same way, Det [I]
If [A] = [LHD] [L]T,
= 1
g.n
where [L] is a lower triangular matrix with 1,,=1 and I'J
and [0] is a diagonal matrix
= 0 Vi> j
et
then,
Det [A] = Det [L] . Det [0] . Det [L]T = Det [0] .....(2.18)
(e) The determinant of a matrix is not affected by row or column
modifications.
A matrix whose determinant is zero is defined as a singular matrix

Example 2.2

Given that the area of a triangle with corners at (X"YI), (X2,Y2) and (X3,Y3) can
be written in the form,

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 41

I
I Xl
Area = ~ Det 1 x 2
I X3

determine the area of the triangle with corners at (1,1), (4,2) and (2,4)

Solution:

ww
Area = Y2 I 4 2
240
=~ 0 3
3
with row modifications,

and
R2 = R2 - Rl
R3 = R3 - Rl

w 1
= -
2 .Ea
(I x (3 x 3 - 1 xl) - 0 x (I x 3 - 1 xl) + 0 x (1 x I - 3 xl» = 4

2.4 INVERSION OF A MATRIX

syE
[Ar l = Adj [A] / Det [A] = [Cf / A ngi
It is defined only for a square and non-singular matrix and is defined as
I I .....(2.19)
nee
It can be seen that product of any square matrix and its inverse is a unity
matrix and the unity matrix being symmetric, the product is commutative.
i.e., [A]"I[A] = [A][Ar l = [I]
rin ..... (2.20)
A necessary and sufficient condition for /Arl to exist is that there be no
non-zero vector {x} such that [A] {x} = {O}. The converse is also true. If g.n
there exists a non-zero vector {x} such that [A] {x} = {O}, then [A] does not
have an inverse.

2.5 METHODS OF SOLUTION OF SIMULTANEOUS EQUATIONS

et
Finite element method gives rise to a set of n independent equations m n
unknowns, expressed in matrix form as [K] {x} = {P}, where [K] is the stiffness
matrix, {x} is the displacement vector and {P} is the applied load vector.
This system is called a homogeneous system if all the elements of {P} are
zero. This system has a trivial solution Xl = X2 = X3 .•.•••• = Xn = O.
It is called a non-homogeneous system if at least one element of {P} is non-zero.
The system is said to be consistent if [K] is a square, non-singular matrix.
Most engineering problems generate a consistent system of equations and
have a unique solution for the displacements {x}. F

Downloaded From : www.EasyEngineering.net

42 FINITE ELEMENT ANALYSIS

These unknowns {x} can be evaluated by two basic approaches:

(a) Calculating inverse of the stiffness matrix [K] and multiplying by the
load vector i.e., {x} = [Kr 1 {P}
or
(b) Solving directly the system of equations [K] {x} = {P}.
These methods are broadly classified as:

ww (i) direct methods which give exact solution while requiring more
computer memory, and
(ii) iterative methods which give approximate solutions with

w.E minimum computer memory requirement.

Some of the numerical methods, based on these approaches, are.

asy
presented here.

2.5.1
E
By INVERSION OF THE COEFFICIENT MATRIX

ngi
(a) Method of Co/actors - It involves calculation of n x n cofactors and
determinant for a n x n matrix [A]. It is a costly process in terms of number

nee
of calculations and is not commonly used with large sized matrices in finite
element analysis.

Example 2.3
rin
g
Find the inverse of a square matrix [A] by the method of cofactors, if
.ne
[AJ~ r~ ~ :1 t
Solution
3
Mil = 11 = 3 x 4 - 1 x 2 = 10 .
2 4 '

MJ2=
4
~I = 4 x 4 - 1 x 1 =15

M13=
4
~I = 4 x 2 - 3 x 1 = 5

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 43

Similarly,

~I= 1
3 2 2
M21 = 41 = 4' M22 = 41 =4 . M23=
2 4 ' 4 ' 1

~1=-14;M33=~ ~I =-6
3 2
M31 = 41 =-9' M32 =
3 1 ' 4

IAI = all M II - al2 MI2 + al3 MI3 = 2 x 10 - 3 x 15 + 4 x 5 = 20 - 45 + 20 =-5

ww IA I IAI
Mil
[A]-I = Adj[A] = _1 -M
12
-M21
M
22
M3lj
-M
32
=-(.!.)
5
r-1510 -4 -9j
4 14

w M 13

.Ea
r - M 23 M 33

(b) Gauss Jordan method of inverting a square matrix [A] involves adding an
5 - 1 - 6

syE
identity matrix [I] of the same order as [A] and carrying out a sequence of
row operations like

ngi
R/ I)= Rl/all ; R2(1) = R2 - a21 RI(I) ; ....... R.,(I) = Rn - ani R/ I)
R2(2) -- R2(I)/a 22,. R I(2) = R I(I) - a 12 R2(2) ,.
. .. ..... Rn(2) = Rn(I) - an2 R2(2)

nee
R3(3) -- R3(2)/a33,• RI(3) = R I(2) - a 13 R3(3) .,....... Rn(3) = Rn(2) - an3 R3(3) etc ..

rin
on this combination such that [A] is transformed to [I]. This will result in
the identity matrix [I] initially appended to [A] getting transformed into
[Ar
l

g.n
Example 2.4

Find inverse of matrix [A] by Gauss Jordan method, if

et
[A]=r~ ~ ~l
1 2 4J
Solution

I]=[~
3 4
[A: 3 1
. 1 2 4

Downloaded From : www.EasyEngineering.net

44 FINITE ELEMENT ANALYSIS

Row operations R,(l) = R/all; R2(1) = R2 - a2l RI(I) and R3(1) = R3 - a31 RI(I)
will give
3
2 o 0
2 2
= 0 -3 -7 -2 1 0
1 1
o 2 o
2 2
Row operations R}2) = R2(1)/an; R I(2) = RI(I) -a12 R2(2) and R3(2) = R3(1) - an
R2(2) will give

ww 0 --
3
2
--
1
2
1
-
2
0

w.E 0
7
-
3
5
2
-
3
5
--
1
3
1
0

asy
0 0

Row operations R3 (3) = R3(2)/a33

6 6 6
R I(3) -- R I(2) - a 13 R3(3) and R2(3) -- R2(2)
- a23 R3(3) WIOil gIve
°

E 0 0 -2 ngi 4 9

nee
5 5
4 14
0 0 3 - - == [I: A -I]
5 5
0 0 -1
1
-
5
6
5 rin
Thus,
g .ne
[A ]-1 ==
-2

-1
3
4
5
4
5
1
9
5
14
--
5
6
=-G) -I~ l 10
-:-4
4 14
-1 -6
-9j t
5 5
which is same 'as the inverted matrix of [A] obtained by the method of
cofactorso

2.5.2 DIRECT METHODS

Many engineering problems in FEM will result in a set of simultaneous

equations represented by [K] {x} = {P} where [K] is the stiffness matrix of the
entire structure and {x} is the vector of nodal displacements due to the applied

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 45

loads {Pl. In many situations, a component may have to be analysed for

multiple loads, represented by [Pl. Corresponding to each column of [P],
representing one set of nodal loads, a particular solution vector {x} exists.
Matrix inversion approach is much faster in such cases than the direct solution
techniques presented below.
(a) Cramer's Rule gives direct solution to a set of simultaneous equations
[A]{x} = {B}.
In a system of n equations in n unknowns, this method involves in
calculating n + I determinants of n x n matrices. It is a very costly method

ww for large sized problems and is, therefore, not popular.

x =~. x =IA21. x )Anl

w.E
IAI' 2 IAI' ...... n IAI

where [AI] is obtained by replacing column i of matrix [A] with the elements
of {B}.

Example 2.5 asy

E
Solve the following system of equations
x+y+z=II ngi
2x-6y- z= 0
3x +4y+ 2z= 0 n eer
Solution
ing
The coefficient matrix is
.ne
I I 11
[A]= 2 -6 -I
r3 4 2 t
and the vector of constants is {B} = [11 0 0 ]T
Then,

and
I
II I 11
[A 1]= 0 -6 -I ;
o 4 2
1 111
-6 0

IAI= I [(-6) x 2-(-1) x 4] -I [(2) x 2-(-1) x 3] +1 [2 x 4-(-6) x 3]

4 0

= II

Downloaded From : www.EasyEngineering.net

46 FINITE ELEMENT ANALYSIS

IAII=II [(-6)x2-{-I)x4]-0[1 x2-4x 1]+0[1 x (-1)-(-6) x 1]=-88

IA21=-11 [2 x 2-(-1) x 3] +0 [I x 2-3 x 1] -0 [I x (-1)-2 x I] = -77
IA31=11 [2x4-(-6)x3]-0[1 x4-3x 1]+0[1 x(-6)-2x 1]= 286

Therefore, x=~=- 88 =-8

1A 1 II

y _I-A21_
- -77
-- - 7
1A 1 II

ww and z = 1 A3 1 = 286 = 26
1A 1 II

w.E
(b) Gauss Jordan method for solving a set of non-homogeneous equations
involves adding the vector of constants {B} to the right of square matrix

asy
[A] of the coefficients and carrying out a sequence of row operations
equivalent to multiplication with [Ar l like
R I(1)-- Ria
I II . R2(I) -- R 2 - a21 R I(I)., ....... R0(I) -- R0 - a 01 R I(I)

E
,
R2 (2) = R 2 22,. R I(2) = R I(I) _ a 12 R2(2) ,.
(I)/a . . . . . . .. R0(2) = R0(I) _ a 02 R 2(2)
R3 (3) = R (2)/a
3 33, I . R (3) = R
ngi
I(2) _ a R
13 3
(3) . R (3) = R0(2) _ a 03 R 3(3) etc ..
, . . • ••• 0

on this combination such that [A] is transformed to [I]. This will result in

I.e. [ A : B ] -7 [I: x ] n
the vector of constants {8} getting transformed into the solution vector {x}

eer
ing
It involves up to 2n 3 multiplications and is a costly method for large matrices
3
Example 2.6
Find the solution to the folIowing set of non-homogeneous equations by
.ne
calculating inverse of matrix [A] by Gauss Jordan method.
x+y+z=11
2x-6y-z= 0
3x +4y+ 2z=O
t
Solution
The set of equations can be expressed in the form [A] {x} = {B}
where,

[Al~l~ -: -J {x} ~ {~} and {B) ~ f~l}

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 47

I~]
1
-6 -1
4 2
Row operations RJ(J)= RJ/aJJ ; R2(J) = R2 - a2] RJ(J) and R3(J) = R3 - a3J RJ(J)
will give

1 1 1: II]
= 0 -8 -3 : -22

ww ro 1 -1 : -33
Row operations R2(2) = R2(1)/a22 ; R/2) = RJ(J) -a]2 R2(2) and R/ 2) = R3(J) -a32

w.E
R2 (2) will give

0
5 33

a 0

0 --
8

syE
3
8
11
--
-
4
11

143
4

ngi
0
8 4
Row operations R3(3) = R3(2)/a33 ; R J{3l = R J(2) - aJ3 R3(3) and R2(3) = R2(2)-a23
R3(3) will give

=r~o ~ ~ . -8] nee

0
: -7 =[I:x]
: 26 rin
Thus,
g .ne
which is same as the solution vector {x} obtained by the method of'
t
cofactors.
(c) Gauss elimination method is a direct method for solving [K]{x} = {P}
where [K] is a square matrix of order n. The method involves reduction of
the coefficient matrix to upper triangular form and back substitution.

Example 2.7
Solve the following system of simultaneous equations by Gauss elimination
method

Downloaded From : www.EasyEngineering.net

48 FINITE ELEMENT ANALYSIS

x+y+z=ll
2x- 6y-z = 0
3x +4y + 2z= 0

Solution
The given equations can be expressed in matrix form as

ww The elements of [K] are reduced to upper diagonal form by the following

w.E
operations, performing on one column at a time.
Column-1

asy gIve
and
En
Column-2 [~gi
R,= RrR, (k"Ik,,) give
nee
rin
Back substitution

Row-3
8
z=( _1:3)( -1 1)=26 g.n
Row-2
_ -22+3.z _
y- (-8) --
7 e t
Row-l x=ll-y-z =-8
1
which is same as the solution vector {x} obtained by the earlier methods.
(d) LU factorisation method is based on expressing square matrix [K] as a
product of two triangular matrices Land U.
i.e., [K] = [L] [U],
where [L] is a lower triangular matrix
and [U] is an upper triangular matrix

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 49

This method involves n3/3 + n2 operations and thus is an economical method

for large size problems. There are three variations of this method.
(i) Doolittle method wherein diagonal elements of lower triangular matrix
L are equal to I
Step-1 {P} = [K]{x} = [L][U]{x} = [L]{v} .....(2.21)

kll kl2 k13 ... kin 0 0 ... 0 Ull u l2 u13 ... u ln

k21 k22 k23 ... k2n 121 0 ... 0 0 u 22 u 23 ... u 2n
=
ww k31 k32 k33 ... k3n 131 132 1... 0 0 0 U33 ... u 3n

kn3 ... k nn 0 0

w.E
knl kn2 Inl in2 In3 ••• 1 0 ... unn

The elements of matrices [L] and [U] are obtained using the following
equations of matrix multiplication
kll = I . Ull asy ~ Ull
kl2 = 1 . Ul2
kin = 1 . Ul n
k21 = 121 • Ull
E ~

~
Ul2
Uln
=> hi ngi
k22 = 121 • Ul2 + 1 . U22
k23 = 121 . U13 + 1. U23
~ U22

~ U23 nee
k2n = 12l . Uln + 1 . U2n
and so on
~ U2n

rin
Step-2: Intermediate solution vector {v} is calculated from {P}
0 0 ... 0
=
g
[L]{v}
.ne
t
VI
Pll rI
P2 121 1 0 ... 0 v2
. :;i;e

~:J ~ l:::
132 1... 0 V3

102 In3···1 Vo
from which, elements of {v} can be obtained as
VI = PI
V2 = P2 -/21 . VI
V3 = P3 - hi . VI -/32 . V2 and so on

Downloaded From : www.EasyEngineering.net

SO FINITE ELEMENT ANALYSIS

Step-3 : Actual solution vector {x} is calculated from {v} = [U]{ x}

VI UII u l2 u l3 Uln XI
V2 0 un U23 u 2n x2
I.e., V3 0 0 u 33 u 3n X3

vn 0 0 0 Unn Xn
from which, elements of { x } can be obtained as

ww Xn =_11_
v

w.E X n_1 -
_ (V n_1 -Un_I'n Xn)
Un_I' n-I

asy
(v n-2 - lI n- 2, n-I Xn_1 - Un- 2' n Xn)
X I1 _ 2 = and so on
lIn-2' n-2

Example 2.8
E ngi
Solve the following system of simultaneous equations by L-U factorisation
method
x+y+z=ll
nee
rin
2x-6y-z = 0
3x +4y+2z= 0

Solution
The coefficient matrix is g .ne
t
The elements of [L] and [U] are obtained from
kll = III . UII and III = 1 ~ Ull = 1
kl2 = III . Ul2 and III = 1 ~ UI2 = 1
kl3 = Ill. Ul3 and III = 1 ~ U13= 1
k2l = 12l . Ull ~ 12l = 2
kn = 12l . Ul2 + In . Un and In = 1 ~ Un =-8
k23 = 12l . Ul3 + 122 , U23 and In = I ~ U23 =-3

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 51.

/3\ = 3
k32= hI • u\2 + 132 . U22 ~ h2 = -118
k33= /31 • U\3 + h2 • U23 + 133 • U33 and 133 =I ~ U33 = -11/8

Thus {P} = IK] {X} = [L] [U] [X]

ww
w Let {P} = [L]{v}

.Ea
Then, elements of vector {v} are obtained from
VI = PI = 11
V2 =P2- hI. VI
V3 = P 3 -/31 • VI syE - = -22
-/32 . V2 = 143/4

z=-
V3
=26ngi
Actual solution vector {x} is calculated from {v} = [U]{x}

U 33

y=(V 2 -U 23 Z}
=-7 nee
u 22

x= (V I -U I2 y- ul3 z) rin
UII
=-8

g.n
(ii)

(iii)
Crout's method is another factorisation method wherein diagonal
elements of upper triangular matrix U are equal to 1. The remaining
procedure is identical to Doolittle method.
Cho/esky method is also a direct method for solving [K]{x} = {P},
et
where [KJ is symmetric and positive definite.
i.e. for any vector {x}, {x} T [A] {x} > 0
Here, {P}= [K]{x} = [L][L]T {.x} where [LI is a lower triangular matrix
This method is also a factorisation method except that upper triangular
matrix [U] is replaced by the transpose of lower triangular matrix [L].
Hence, the diagonal elements of [L] will not be equal to I. Since one
matrix [L] only need to be stored, it requires less memory and is
especially useful in large-size problems.

Downloaded From : www.EasyEngineering.net

52 FINITE ELEl'AENT ANALYSIS

k1l kl2 k13 ... kIn 111 0 0 ... 0 111 121 131 •.• /nl

k21 k22 k23 ... k2n 121 122 0 ... 0 0 122 132 ••• / n2

Step-1 k31 k32 k33 ... k3n 131 132 133 ••• 0 0 0 133 •.• / n3

knl kn2 kn3 ... k nn Inl In2 I n3 ···/nm 0 0 o... /nn

ww The elements of matrix [L] are obtained using the following equations of
matrix multiplication

w.E kll = 111 . 111

k 12 =/II./2l

asy
kIn = 111 . Inl
k21 = 121 • III is same as k12 = III . 12l since [K] IS symmetric and IS
redundant
En
k22 = 121 • 12l + 122

gi
• 122

k23 = 12l . hI + lz2

k2n = 12l . lIn + In . 1n2
• 132

nee
and so on
rin
Step-2: Intermediate solution vector {v} is calculated from {P} = [L]{v}
PI III 0 0 ... 0 VI g.n
i.e.,
P2
P3 =
121

131
122

132
0 ... 0
133 ••• 0
v2
V3 e t
Pn Inl In2 In3 ··.lnn vn
from which, elements of {v} can be obtained as

vl =PI-
111
(P2 - /2l .vl )
v2 =
122

(P3 -/31 ·V I -132· v 2)

V3 = and so on
133

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 63

Step-3: Actual solution vector {x: I~ l:alculated from {v} = [U]{x}

VI III 121 III In 1 XI

v2 0 122 112 'n2 X2

I.e v] 0 0 In]
'13 X3

0 0 0 Inn

ww
Vn Xn

from which, elements of { 0 } can be obtained as

w.E X
v
=_n
n Inn

(v n-I -In '

asy
n-I X n )
X -
n-I - I
n-I, n-I

X
n-2 -
En
(V n - 2 - /n -I' n-2 X n _1
-~~--~-=~~~~~~~
I
n-2, n-2
-In 'n-2, X n)
and so on

gi
(iv) Gauss Jordan method is also similar to Gauss elimination method except
that back substitution operation is avoided by transforming coefficient
matrix [K] into a unity (diagonal) matrix, instead of an upper triangular nee
matrix. This increases number of operations and hence is not economical.
However, this method is popular for inversion of matrix [K] which can
rin
then be used with multiple right side vectors. In such applications, this
method is more economical. g.n
2.5.3 ITERATIVE METHODS
There are many different approaches, but two popular methods are briefly
explained here.
e t
(a) Jacobi iteration or Method of simultaneous corrections: this method In
also, each row is first divided by the corresponding diagonal element to
make diagonal elements equal to 1.
Then, {x}(m+ I) = {P} + ( [I] - [K]) {x}(m)

This form is used for successive improvement of the solution vector from
an initial approximation {x}(O). No component of {x}(m) is replaced with
new ones, until all components are computed. All components are
simultaneously changed at the beginning of each step. For better
understanding, matrix notation is avoided in the example.

Downloaded From : www.EasyEngineering.net

54 FINITE ELEMENT ANALYSIS

Example 2.9

Solve the following simultaneous equations by Jacobi iteration method.

20x + y - 2z = 17
3x + 20y - z = -18
2x - 3y + 20z = 25

Solution

ww These equations are rewritten as

x' = (17 - y,-I + 2Z'-I) / 20
yi = (-18 _ 3X,-1 + Zi-I) / 20

w.E z' = (25 - 2X,-1 + 3y'-I) / 20

The values obtained after each step are tabulated below. Starting assumption
refers to i = 0
i-> 0 asy 1 2 3 4 5 6
x
Y
z
0
0
0 E0.85
-0.9
1.25
1.02
-0.965
1.1515
ngi
1.0134 1.0009 1.0000 1.0000
-0.9954 -1.0018 -1.0002 -1.0000
1.0032 0.9993 0.9996 1.0000

nee
Exact solution is x = I, Y = -1 and z = I
(b) Gauss Siedel iteration or Method of successive corrections : In this
method, each row is first divided by the corresponding diagonal element to

rin
make diagonal elements equal to I. Coefficient matrix is assumed as sum of
three matrices, [I], [L] and [U] where [I] is the unity matrix, [L] is a lower

triangular matrix with diagonal elements equal to o.

Then,
g
triangular matrix with diagonal elements equal to 0 and [U] is an upper

.ne
[K] {x}= ([I] + [L] + [U]){x} = {P}or{x}={P}- [L] {x} - [U] {x}
This form is used for successive improvement of the solution vector from
an initial approximation {x} (0) with
{x}(m+ I) = {P} _ [L] {x}(m) _ [U] {x}(m)
t
replacing old terms of solution vector {x} (m) with new ones, as soon as they
are computed.
For better understanding, matrix notation is avoided in the example.
Example 2.10

Solve the following simultaneous equations by Gauss Siedel iteration method.

20x + y - 2z = 17
3x+20y-z=-I8
2x - 3y + 20z = 25

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 55

Solution
These equations are rewritten as
x' = (17 - y'-I + 2Z'-') / 20
y' = (-18 - 3x'+ Z'-I) / 20
z' = (25 - 2x' + 3y') / 20
Note the change in the superscripts (iteration no), compared to the Jacobi
iteration method.

ww The values obtained after each step are tabulated below. Starting assumption
refers to i =0

w i->
x
.Ea
0
0
0
I
0.85
-1.0275
2
1.0025
-0.9998
3
1.0000
-1.0000

syE
Y
z 0 1.0109 0.9998 1.0000
Thus, successive correction (Gauss Siedel iteration) method converges much

ngi
faster than simul aneous correction (Jacobi's iteration) method.

2.6 EIGEN VALVES AND EIGEN VECTORS

nee
For every square matrix [A], there exist Aand {u} such that [A] {u} = A{u} or
([A] - A[ID {u} = {O}
rin ••••.(2.22)
For a non-trivial solution, {u} =t- {O},
I[A] - A[I] I = 0 is called the characteristic equation g.n
..... (2.23)
A, s are called eigen values. Eigen values may be real or complex. Most of
the engineering problems will have real eigen values. The system of n
independent equations represented by ([A] - A[ID {u} = {O} will have n eigen
et
values, where n is the order of the square matrix [A]. Some of them may be
repeated. {u,} associated with each A, is called an eigen vector and is calculated
from the system of equations ([A] - A.,[ID {u,} = {O}.
[A] is a positive definite matrix if all its A, are +ve
or if {x} T[A]{ x} > 0 for any non-zero vector {x}
[A] is a positive semi-definite matrix if all its A, are ~ zero
[A] is an indefinite matrix if its A, are -ve, zero or +ve
[A] is singular ifand only ifone or more of its eigen values are zero ..... (2.24)

Downloaded From : www.EasyEngineering.net

56 FINITE ELEMENT ANALYSIS

Eigen values and eigenvectors of an (n x n) matrix can be obtained by one of

the two approaches .
• Solving nth order characteristic equation for the n eigen values and solving
the !'ystem of n simultaneous equations with each eigen value ifor the
corresponding eigenvector .
• Solving the system by iterative method through successive

ww approximations of the eigen values and corresponding eigenvectors

Since eigen values in a dynamic system represent the natural frequencies and

w.E
only the first few eigen values representing the dominant modes of vibrations
are usually required, the latter approach is faster and is more commonly used.
Both these techniques are explained here through simple examples.

asy
Example 2.11

E
Determine the eigen values for the equation of motion given below

ngi
n eer
where ~ and cp, are eigen values and eigen vectors respectively

ing
.ne
Solution

Characteristic equation is given by

t
or Detf2-14'A., 2-4'A., 1 1=0
I I 2-4'A.,

(2 - 4~) [ (2 - 4'A.,i - I] - [ (2 - 4~) - I] + [I - (2 - 4~) ]

= (2 - 4~)3 - (2 - 4'A.,) + 2 =0
or -64 'A.,3 + 96 'A.,2 - 36 ~ + 4 = -4('A. - 1)(4'A. - 1)2 = 0

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OPERATIONS 57

This cubic equation has three real roots given by 1..=1 ~ ~

'4'4

Example 2.12

Detennine the eigen values and eigen vectors of the equation

HA] -A[I]} {q} = 0 where 1..= eigen value and q = eigen vector, if

wwA~l: ~ i1
w.E
Solution

a syE
Characteristic equation is given by
2

ngi 3-1..
2
1 =0
2-1..

nee
(2 - A) [(3 - A) (2 - A.) - (I) (2)] - (2) [(I) (2 - A.) - (1) (1)]
+(1) [(1).(2)-(3-A.)(1)]=0
A.3 -n2 rin
i.e.,
or
+ 11 A. - 5 = 0
(A.-5)(A.2 -2A.+ 1)=(A.-5)(A.-I)(A.-l)=0 g.n
Therefore, eigen values ofthe matrix are 5, I and 1
Eigen vectors corresponding to each of these eigen values are calculated by
substituting the eigen value in the three simultaneous homogeneous equations.
et
Corresponding to A. = 5, (2 -5)xl + 2X2 + x) = 0
XI + (3 - 5)X2 + X3 =0

n
XI + 2X2 + (2 - 5)X3 = 0

which gives {x} ~

This eigenvector indicates rigid body mode of vibration.

Downloaded From : www.EasyEngineering.net

58 FINITE ELEMENT ANALYSIS

Corresponding to A = I, (2 - I )XI + 2X2 + X3 = 0

XI + (3 - I )X2 + X3 = 0
XI + 2X2 + (2 - I)X3 =0
nd rd
Since the 2 and 3 equations are identical, we have effectively two
independent equations in three unknowns. Therefore, a unique solution does not
exist. These equations are satisfied by

ww
w.EIn many practical applications involving many degrees of freedom, only the

asy
fundamental mode is dominant or important. Iterative methods such as
Rayleigh's Power method, House holder's tri-diagonalisation method are very
commonly used for computing few dominant eigen values and eigen vectors.

En
Rayleigh's Power method: This iterative method is used for calculating
fundamental or largest eigen value
[A] {x}
gin
= A {x} is rewritten for iterations as [A] {X}(i-I) = Ai {X}(I)

Example 2.13
eer
ing
Calculate largest eigen value for the matrix [A] by Rayleigh's Power method

2 -12-10]
where [A]=
[ o -1 2
-1
.ne
Solution
t
Let the initial assumption be {x}(O) = [1 0 O]T

Then, [A](X},OI =H -1
2
-1
-!]H={-!}=+H=A'11 {x}'])

=[-~ -~1{- ~.5} =r-~} =2.5 {- ~.8} =A

-1
[A](x}(]) 2 (2) {x}{2l,
-1 2 0 0.5 0.2

Downloaded From : www.EasyEngineering.net

CHAPTER 2 MATRIX OpLRA I'IONS 59

Eigen values and elgen vectors obtained after successive iterations are
tabulated below.

;-> 0 J 2 3 4 5 6 7
0.87 0.80 0.76 0.74
{x} 0 -0.5 -0.8 -1 -I -1 -1 -I
0 0 0.2 0.43 0.54 0.61 0.65 0.67
1 2 2.5 2.8 3.43 3.41 3.41 3.41

ww
2.7 MATRIX INVERSION THROUGH CHARACTERISTIC EQUATION

w
Cayley-Hamilton theorem states that that every square matrix satisfies its own
characteristic equation. Let characteristic equation of square matrix [A] be
given by
.Ea .
DCA) = An + CIl- I An-I + Cn-2 An-2 + Cn-3 An-3 + ...... + C I A+ Co = 0
Then, according to this theorem,
syE
An + Cn- I An-I + Cn-2 An-2 + Cn-3 An-3 + ...... + C I A + Co = 0
Multiplying throughout by
ngi
A-I , All-I +C n-I AI1 - 2 +C n-2 An-3 +C n-3 An4 + . . . +C I I +C 0 A-I = 0
or
nee
A-I =-[ An-I +C n- I An-2 +C n-2 An3 +C n 3 An-4 + ... +C} I] / Co

Example 2.14
rin
Obtain inverse of the following matrix using Cayley-Hamilton theorem

g.n
A=r~ : ;j et
Solution
Characteristic equation is given by

Det
1- A.
2
2
4-A.
3
5
j=0
r 3 56-A.
i.e., A3- 11A.2 -4A+1=0 or A3-1IA2 -4A+I= 0
Multiplying by A-I,

Downloaded From : www.EasyEngineering.net

60 FINITE ELEMENT ANALYSIS

we get A2-11A-4I+A-'= OorA-'= -[A2-IIA-41]

23 23
A2=[A][A]= 2 4 51[12
1 4 51 = [1425311
25 45 56
[
3 5 6 3 5 6 3I 56 70

Then,
31
4[~ ~1
25 2 0
[14
A-'=-[A 2
-11A-41]=- 25 45 [I
561 +.11 2 4
;1 +
1
31 56 70 3 5 0

ww
w.E+~
-3
3
-1 -!1
Note : Reader is advised to verity the solutions obtained by these various

asy
methods with the standard check [A] [Ar' = [I], since numerical errors are
frequently encountered while practicing these methods.

2.8 SUMMARY En
• gin
A matrix is a group of (m x n) numbers (scalars or vectors) arranged

eer
in 'm' number of rows and 'n' number of columns. FEM deals with
solution to a large number of simultaneous equations, which can be
expressed more conveniently in matrix form as [K] {x} = {P}, where

ing
[K] is the stiffness matrix of the component which is usually square,
symmetric and non-singular.
• This system of equations can be solved for {x} when [K] is a square,
non-singular matrix, in the form {x} = [Kr' {Pl. Direct methods such .ne
•
as Gauss elimination method or factorisation method and iteration
method like Gauss Siedel method are commonly used techniques for
the solution of the system of equations.
For every square matrix [A], there exist f... and {u} such that
t
[K]{ u} = f...{ u}.f... is called the eigen value and {u} is the
corresponding eigen vector. Stiffness matrix is usually positive
definite i.e., {u}T[K] {u} > 0 for all non-zero {u}. Eigen values
represent the natural frequencies and eigen vectors represent natural
modes of a dynamic system. Iterative methods such as Rayleigh's
Power method, House holder's tri-diagonalisation method are very
commonly used for computing the first few dominant eigen values
and eigenvectors.

Downloaded From : www.EasyEngineering.net

CHAPTER 3
THEORY OF ELASTICITY

ww
w.E
A brief review of theory of elasticity, with specific reference to applications of

asy
FEM, is presented here for a clear understanding of the subsequent chapters. For
a more detailed explanation, the reader may refer to other standard books on this
subject.

En
Every physical component is a three-dimensional solid. However, based on
the relative dimensions along three coordinate directions and nature of applied

gi
loads / boundary conditions, they are classified as 1-0, 2-0 or 3-0 components.

nee
This idealisation helps in analysing the component quickly and at lower cost.

3.1. DEGREES OF FREEDOM

rin
The direction in which a point in a structure is free to move is defined as its
degree of freedom (OOF). In general, any point in a component can move along
an arbitrary direction in space and rotate about an arbitrary direction, depending g.n
upon the loads applied on the component. Since specifying this arbitrary
direction through angles is tedious, the movement and rotation at any point are
identified by their components in the chosen coordinate system. Thus, in
Cartesian coordinate system, a point can at best have translation identified by its
e t
three components along the three coordinate directions X, Y and Z; and rotation
identified by its three components about the three coordinate directions X, Y
and Z. Oepending on the way a member is assembled in a structure, some or all
of its OOF at a point may be fixed or free. Truss members are designed for only
axial loads and rotational OOFs are not relevant whereas beam is designed for
bending loads which result in displacement at each point normal to the axis. Its
derivative or slope is independently constrained and hence is considered as an
independent OOF. Similarly, in plates subjected to in-plane loads, rotational
OOFs are not significant and need not be considered as independent OOFs

Downloaded From : www.EasyEngineering.net

62 FINITE ELEMENT ANALYSIS

whereas in plates and shells subjected to bending loads, rotational OOF are
significant and need to be treated as independent OOFs.

3.2 RIGID BODY MOTION

It is the motion of entire component or part under the influence of external

applied loads. Such a rigid body motion, with no relative deformation between
any two points in the component, cannot induce stresses or strains in a
component. Since design or analysis of a component involves calculation of

ww stresses and strains in a component, any static part with 'n' degrees of freedom
can not be solved unless it is restrained from moving as a rigid body by
constraining it at least at one point along each OOF. The number of OOFs to be

w.E
constrained depends on the type of component. For example, a truss has to be
constrained for rigid body motion along X and Y axes; a plane frame (in X-Y
plane) has to be constrained along X and Y axes as well as for rotation about Z

asy
axis; a thick shell has to be constrained along X, Y and Z axes as well as for
rotation about X, Y and Z axes. In the conventional analysis by closed form

En
solution to the differential equation, the rigid body motion is constrained by
using relevant boundary conditions for evaluating constants of integration.

3.3
gi
DISCRETE STRUCTURES

nee
Structures such as trusses and frames, which have many identifiable members,

rin
connected only at their end points or nodes are called discrete structures. Each
member of the structure is considered as a one-dimensional (I-D) element along

g.n
its length, identified by its end point coordinates. Their lateral dimensions are
reflected in element properties like area of cross section in trusses and moment
of inertia and depth of section in beams.

3.4 CONTINUUM STRUCTURES e

Structures such as plates, thin shells, thick shells, solids, which do not have
t
distinctly identifiable members, can be modeled by an arbitrary number of
elements of different shapes viz., triangles and quadrilaterals in 2-D structures
and tetrahedron and brick elements in 3-D structures. These are called
continuum structures. In these structures, adjacent elements have a common
boundary surface (or line, if stress variation across thickness is neglected as in
the case of plates). The finite element model represents true situation only when
displacements and their significant derivates of adjacent elements are same
along their common boundary.

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 63

3.5 MATERIAL PROPERTIES

FEM is ideally suited for analysing structures with varying material properties.
The variation of properties such as Modulus of elasticity (E), Modulus of
rigidity (G) and coefficient of linear thermal expansion (a) may be:
(a) constant (linear stress-strain relationship) or variable (non-linear stress-
strain relationship) over the range of load.
(b) same in all directions (isotropic) or vary in different directions
(anisotropic or orthotropic).

ww (c) constant over the temperature range or vary with temperature,

particularly when the temperature range over the component is large.

w.EThese variations may be inherent in the material or induced by the

manufacturing processes like rolling, casting. Treatment of non-homogeneous
material, with varying properties at different locations of a component, is

asy
difficult and is also very unusual. In most cases, variation of material properties
with the direction and with temperature may not be significant and hence

E
neglected. So, an isotropic, homogeneous material with constant (temperature-

ngi
independent) properties is most often used in the analysis of a component.

3.6
nee
LINEAR ANALYSIS

It is based on linear stress-strain relationship (Hooke'S law) and is usually

rin
permitted when stress at any point in the component is below the elastic limit or
yield stress. In this analysis, linear superposition of results obtained for

g
individual loads on a component is valid in order to obtain stresses due to any
combination of these loads acting simultaneously. In some designs, it is
necessary to check for many combinations of loads such as pressure and thermal
.ne
at different times of a start-up transient of a steam turbine. In such a case,
analysing for unit pressure; multiplying the stress results with the pressure
corresponding to that particular time of the transient and adding to the stresses
due to temperature distribution will be economical.
t
3.7 NON-LINEAR ANALYSIS

In many cases, the mathematical formulations are based on small deflection

theory. A component with large deflections due to loads, such as aircraft wing,
comes under the category of 'geometric non-linearity': In some aerospace
applications, where the component is designed for single use, stress level above
yield point, where stress-strain relationship is non-linear, may be permitted. In
some other cases involving non-metallic components, material may exhibit non-
linear stress-strain behaviour in the operating load range. These two cases come

Downloaded From : www.EasyEngineering.net

64 FINITE ELEMENT ANALYSIS

under the category of 'material non-linearity'. In both these cases, analysis

carried out by applying the load in small steps.
(a) Geometric non-linearity : In these problems, geometry of the
component is redefined after every load step by adding the
displacements at various nodes to the nodal coordinates for defining the
true geometry to be used for the next load step.
(b) Material non-linearity : In these problems, total load on the
component is applied in small steps and non-linear stress-strain

ww
relationship in the material, usually represented by the value of Young's
modulii or Modulii of elasticity (normal stress/normal strain) Ex, Ey
and Ez in different directions, is considered as linear in each load step.

w These values are suitably modified after each load step, till the entire

.Ea
load range is covered. Here, normal stress and normal strain can be
tensile (+ve) or compressive (-ve) and E has the same units as stress,
since strain is non-dimensional.

syE0"3 ------------
E4

ngi
nee
rin
FIG U R E 3.1 Stress-strain diagram of a non-linear material
g.n
3.6 STIFFNESS AND FLEXIBILITY

Loads and displacements in an element are related through stiffness and

et
flexibility coefficients. Stiffness coefficient (K) is the force required to produce
unit displacement, while Flexibility coefficient (F) is the displacement produced
by a unit force. They are usually defined for 1-0 elements such as truss
elements or spring elements by

P=K.u where, K = AE is the stiffness coefficient

or u F.P where, F = ~ is the flexibility coefficient

Downloaded From : www.EasyEngineering.net

CHAPTER 3 TIIEOR Y OF ELASTICITY 65

Since a = : = EE = E( ~) where, u IS the change in length and,

therefore, P =
(LAEJ.u
However, a more general definition of stiffness coefficient k'J is the force
required at node 'i' to produce unit displacement at node 'j'. i.e., P, = klJ uJ"
An important feature of the stiffness coefficient is that it has different units

ww
with reference to different loads and different displacements. For example,
units of k'J relating load at 'i' to displacement at 'j' in a truss element are in

w.E
'N/mm' whereas units of k'J relating moment at 'i' to slope at 'j' in a beam
element are in 'N mm'.

asy
Stiffness coefficients k'J connecting loads at 'n' number of nodes with
displacements at these 'n' number of nodes thus form a square matrix of order
n x n, represented by [K]. In structures having linear force-deflection

E ngi
relationship, the flexibility and stiffness coefficients have the property k'J = kJ'
and f,J = ~,. This is called Maxwell's reciprocity relationship. This makes the
flexibility and stiffness matrices symmetric.

3.9 PRINCIPLE OF MINIMUM POTElIolTIAL ENERGY nee

rin
Every component subjected to external applied loads reaches stable equilibrium,
when its potential energy or the difference between work done by external
g
forces and internal strain energy due to stresses developed is zero. It can also be
.ne
t
expressed as - During any arbitrary kinematically consistent virtual
displacement from the equilibrium state, satisfying constraints prescribed for the
body, potential energy equals tozero or the work done by the external forces
equals the increment in strain energy
i.e.,

3.10 STRESS AND STRAIN AT A POINT

Stress is the internal reaction in a component subjected to external forces. Thus,

stress exists only when external force is applied on a component and varies
from point to point. Stress at any point in a component is defined as a tensor
with three components on each of the six faces of an infinitesimal cube around
that point, as shown in the Figure 3.2.

Downloaded From : www.EasyEngineering.net

66 FINITE ELEMENT ANALYSIS

Z az

LyZ I LXY ax

ww /:L=~~----
---------- - - . X
+ X face

w.E
,,"'ay
" + Y face
"
""
Y

asy FIGURE 3.2 Stress at a point

En
Stress components are identified by two subscripts, the first subscript
representing the direction of outward normal of the plane or face of the cube

gi
and the second subscript representing the direction of stress on that plane. Thus

nee
O'xx (or O'x) represents normal stress on the plane whose outward normal is
along +ve X-axis (identified as + X face) while O'xy (or'tXy) and O'xz (or 'txz)

rin
represent shear stresses along Y and Z directions on the same plane. Stresses are
similarly defined on the other five faces of the cube.
For the equilibrium of this small element along X, Y and Z directions, it is

g.n
seen that O'xx, O'xy and O'xz on the +X face are equal to O'xx, O'xy and O'xz on the-X
face. Similarly on the Y and Z faces. So, out of the 18 stress components on the
6 faces of the cube, only the following 9 stress components are considered
independent. Thes~ are represented as
'tyz
e t
Also, for the equilibrium of the element for moment!> about X, Y and Z
directions, 'txy = 'tyx, 'txi = 'tzx and 'tyz = 'tzy being the complementary shear
stresses on two perpendicular faces. Thus, only three normal stress components
O'x, O'y, O'z and three shear stress components 't xy, 't yz and 'tzx are identified at each
point in a component. In the following chapters, these components are written
as a stress vector {E} for convenience of matrix operations.
Normal stress on any plane whose normal is N (with components N x, Ny and
N z along cartesian x, y and z directions) can be obtained from the stress

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 67

components available in cartesian coordinate system by the following

relationship.
an = T x N x + T y Ny + Tz N z
where, Tx = ax N x + 'tXy Ny + 'txz N z
T y = 'tXy N x + ay Ny + 'tyZ N z
Tz = 'txz N x + 'tyZ Ny + az N z

ww In a similar way, strain at a point is defined as a tensor of 18 components,

out of which six components consisting of normal strains Ex, Ey and Ez and shear
strains "(xy, "(yz and "(ZX are considered independent. Normal strain along a

w.E
direction is defined as the change in length per unit length along that direction
while shear strain is defined as the change in the included angle as shown in

asy
Fig. 3.3. In the following chapters, these components are written as a strain
vector {E} for convenience of matrix operations.

En
In the case of 2-D and 3-D elements, general relations between
displacements and strains and between strains and stresses, as obtained in the

gi
theory of elasticity, are used for calculating element stiffness matrices. These
relations are given later with the following notation.

nee
Notation: u, v, ware the displacements along x, y and z directions

rin
Ex, Ey, Ez are the normal strains
"(xy, "(yz, "(ZX are the shear strains
ax,O"y, o"z are the normal stresses
't xy, 't yz, tzx are the shear stresses g.n
y e t
T
dy Ov
ox
1 1+4--- dx ---".1
x

FIGURE 3.3 Shear strain at a point

Downloaded From : www.EasyEngineering.net

68 FINITE EI.I:MI:N I ANAI.YSIS

3.11 PRINCIPAL STRESSES

Depending on the type of structure and the load applied, three normal stresses
and three shear stresses can exist at any point in the structure. These stress
components are calculated with reference to the coordinate system used.
However, there exists a plane along which shear stress is zero and the
corresponding normal stresses (maximum and minimum) are called principal
normal stresses or principal stresses. Along some direction, inclined to the two
principal stresses, there exists maximum or principal 'shear stress. These are of

ww interest to any designer as the component has to be designed to limit these

stresses to the allowable limits of the material. They can be calculated, for a
two-dimensional stress state, from

w.E {ax +aJ +~{ax -ayf +4,2

asy 2 2

E
. ngi -
(al-()2)
--
~(()x -()yY +4.2

nee
~----'------
max- 2 2

From uni-axial tensile test, where ax = 0 and, = 0,

we get and 'max

rin ay
=2 =0.5 a max

3.12 MOHR'S CIRCLE FOR REPRESENTATION OF 2-D STRESSES g .ne

The principal stresses can also be obtained by graphical method using Mohr's
circle. Here, normal stresses are represented on X-axis and shear stresses on
Y-axis. Principal or maximum normal stress is inclined to the given stress state
t
by an angle e, while maximum shear stress is the maximum ordinate of the
circle and exists on a plane inclined at (45 - at
from the given stress state or at
45° from the Principal stresses.
Procedure: Plot OA and OB to represent stresses ax and a y along X-axis.
Plot AC and BD, to represent shear stress 'xy, parallel to Y-axis. With CD as
diameter, construct a circle intersecting X-axis at F and G. Then OF and OG
indicate maximum and minimum principal stresses, inclined at an angle e (half
of angle between EC and EA) with line CD representing the given stress state.
Length EH, radius of the circle, indicates maximum shear stress.

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 69

It can be observed that sum of normal stresses on any plane is constant and is
called I st stress invariant IJ
i.e., for 2-D stress case

Shear
stress, '(

ww o
F

Nonnal

w.E ..--cry--~I
stress, cr

asy
En
FIGURE 3.4 Mohr's circle for 2-D stress representation

gi
Two special cases of Mohr's circle are of special importance:

nee
(a) For uni-axial tensile test, from which material properties are usually
evaluated, the load is applied along one axis (usually Y-axis) and other

rin
components of stress are all zero. Then,
a

g.n
0"\ = O"y; 0"2 = 0 and 't max =_Y as can be seen from the Figure 3.5.
2

Shear
stress, T
e t
---+--------L-------+----t> Nonnal
stress, cr

FIGURE 3.5 Mohr's circle for uni-axial tensile test

Downloaded From : www.EasyEngineering.net

70 FINITE ELEMENT ANALYSIS

(b) In the earlier cases, it is seen that normal stress corresponding to

maximum shear stress is not zero. If the normal stress, associated with
maximum shear stress, is zero, then it is called pure shear state. It is
possible when a component is subjected to torsion only. In this case of
pure shear, on a plane inclined at 45° to the plane of maximum shear
stress, radius of the circle is
01 = --<J2 = 'tmax as can be seen from Fig. 3.6.

ww Shear stress, 1:

i
w.E t max

asy
---+--------t-------+-""--+Nonnal

E ngi
stress, cr

nee
FIGURE 3.6 Mohr's circle for pure shear rin
g
This can also be understood by considel ing stresses actitlg at a point,
identified by a small cube around it, when the component is subjected to .ne
torsion. Torsion is represented by a couple formed by two equal and opposite
shear forces acting on opposite faces of the cube, as shown in Fig. 3.7 (a).
An equal and opposite couple is automatically formed, if the component is 'in
t
static equilibrium, as shown in Fig. 3.7 (b). The shear forces resulting from this
couple are called complementary shear forces (stresses). If two different free
bodies of half this cube about its diagonals are considered, resultant of the two
shear forces on adjacent surfaces will result in tensile stress on one diagonal
(BD) and a compressive stress on the other diagonal (AC), as shown in
Fig. 3.7 (c).

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 7:1

D CD+---- CD+---- C

iCJi iOi Ai ~i
1\
(a)
B A ----+
(b)
B A ----+ II
(c)

FIGURE 3.7 Representation of shear, complementary shear and principal

ww normal stresses

If'q' is the shear stress on the face of the cube and's' is the side ofthe cube, then

w.E Resultant force on the diagonal, R = 2.q.S2.COS 9.

R 2.q.s2.cos B

aStress on this diagonal,

syE
a--- -q
- A - 2.s2.cos () -

Thus, both these stresses (maximum or principal nonnal stresses) are equal

ngi
in magnitude to the shear stress on the surfaces and are inclined at 451 0 to the
shear stresses, as already seen in the corresponding Mohr's circle.

nee
Maximllm shear ...tress theory, the most conservative and commonly used
theory of failure, suggests that a component fails when the maximum shear
stress at any point in a component exceeds the allowable maximum shear stress
value of the material.
rin
3.1.3 VONMISES STRESS g .ne
VonMises stress or equivalent stress is related to the three principal stresses at
any point. It is used in Maximum distortion energy theory (which states that a
component fails only due to distortion in shape and is independent of
t
volumetric expansion or contraction). Equivalent stress is given by
_ [(a\ -(2)2 +(a2 -(3)2 +(a3 -al)2l <
a eq - 2 - ay

It is also represented in lenns of I st and 2nd stress invariants (I I and h) as

aeq =~If -31 2

Downloaded From : www.EasyEngineering.net

72 FINITE ELEMENT ANALYSIS

In 2-D plane stress case, Oz= 0 and 03 = 0

II = Ox + Oy = 01 + 02

12 =O'xO'y -'t;y =0'10'2

In 2-D plane strain case, Oz = v (ax + Oy)

II = ax + Oy + Oz = 01 + 02 + 03

and

ww Comparing distortion energy of specimen in un i-axial tensile test,

where 0,

w
01 = Oy ; 02 = 03 =

.Ea
with pure shear state where 't max = 01 = 02 ,

we get, "max ::; .J3 =.0 577

0" y
O"y

syE
This stress value is used in the distortion energy theory of failure and is
very popular.

3.14 THEORY OF ELASTICITY

ngi
nee
The definitions of stress, strain and the relationships between displacement,
strain and stress are explained in Strength of materials with special reference to

rin
I-D structures. Before analysing continuum structures, a more general
understanding of these concepts is essential. In the following, a brief discussion
of these concepts is presented. Reader is advised to go through any book on
theory of elasticity for a more detailed presentation of these concepts.
g.n
(a) Poisson's ratio
Normal strain (tensile or compressive) due to an applied load along the
direction of load is called longitudinal strain. In most engineering
materials, increase of size in one direction is associated with reduction of
et
size in the other two directions, to minimise change in the volume of the
component. Thus tensile longitudinal strain is associated with
compressive lateral strains and vice versa. Ratio of lateral strain to
longitudinal strain is found to be a constant for each material and is called
Poisson's ratio (usually represented by v or 11m). With the usual notation
of tensile strains as +ve and compressive strains as -ve, this ratio is
always negative. However, this ratio is given a positive value and the
negative effect is taken care of in the corresponding equations. Its value
ranges theoretically from 0 to 1, while it varies from 1/3 to 114 for most
of the engineering materials. v = Y2 indicates a perfectly plastic material.
Poisson's ratio, v = Lateral strain I Longitudinal strain

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 73

(b) Rigidity modulus, Bulk modulus

Shear modulus or Rigidity modulus, G = shear stress / shear strain
and Bulk modulus, K = Nonnal stress / Volumetric strain
where,
. . ov
vo I umetnc stram =- : : : Ex + Ey + E z
V
(neglectin phigher order small terms)
(ax -yay -vaJ ~ay -va z -vax) (a z -vox -yay)

ww =----=----+
E E
+-----
E

w.E =(a x +a y +aJ (I-:V)

when same load is acting along x, y and z directions

asy
=a x (1- 2 v) when load is acting along x direction only
E

E ngi
Rigidity modulus and bulk modulus also have the same units as stress
(N/m 2 or Pa)

nee
Bulk modulus has very limited applications in structural analysis. For
any given material, modulus of elasticity and modulus of rigidity reduce
at higher temperatures.
These material constants (modulii) are mutually related by the
following expression rin
(c)
E=2G(1 +v)=3K(I-2v)
Strain-Displacement Relations
g .ne
Strains can also be expressed as functions of displacement components ar
a point in the three Cartesian coordinate directions. If u, v and w (all
functions oflocation of point, represented by its X, Y, Z coordinates) imd
represent the displacement components along X, Y and Z directions, then
t
au
=-
Ov
=-
Ow
=-
E
Xax E
yiJy E
z Oz

au Ov Ov Ow
Yxy = iJy + ax Yyz= Oz + iJy .•••• (3.1)

(d) Thermal stress

Thermal strains do not induce any stresses unless thermal expansion
is constrained.

Downloaded From : www.EasyEngineering.net

74 FINITE ELEMENT ANALYSIS

For example stress in a uniform bar subjected to temperature rise by

~T, is dependent on the end condition as shown below. Let a be the
coefficient of linear thermal expansion. Total elongation of a bar of
length L due to increase in its temperature by ~T is L oc ~T. Then, stress
in the bar depends on the constraint (boundary condition) for its
expansion, as shown Fig. 3.8. In each case,

stress a = E.E =(~) x Restrained part of expansion.

ww
w .Ea Case (a): Unconstrained
~L -+I aI
Case (b) : Partially
constrained
Case (c) : Fully
constrained

Stress, 0=0syE o
(a < L.a.AT)

E(L.a.AT-a) o=E.a.AT

..... ngi
FIGURE 3.8
L

nee
'

(e) Stress-strain Relations or Constitutive equations (from generalised

Hooke's law)
• For linearly elastic and isotropic material
rin
E
x
= (ax -vay -vaJ.
E ' g.n
E
y
= (-vax +ay -va z ).
E '
tyz
'Y yz =G- .....(3.2) et
E
z
= (-vax -yay
E
+az )
.
' 'Yzx
t
= ;;

where G = (E ) is the shear modulus or rigidity modulus

21+v
Sum of the three equations gives,

{ax +ay +a z )
Ex +Ey +E z = (1-2V\ E

These equations can also be written in terms of stresses as functions of

strains.

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 75

E[(I-v}Ex +VEy +VE ] Eyxy

O=--'--,..-:..---=r-;------'-__ z ~
xy = 2(1 + v}
x (1 +v)(1-2v) 't

E[wx +(l-v}E y +VE z ] E yyz

O=---=;--~,..-:..~;----=- 't =--;-~,
y (1 +v)(I-2v) yz 2(1 + v}

E[v[x +VEy +(l-v}E z ] E

o =--~--~--~--- 't
ZX
= 21+v'ZX
( ) 'V ••••• (3.3)
z (1 + v)(1 - 2v)

ww These equations are expressed more conveniently in matrix notation as

{o} = [D] {E} •••••(3.4)

w.E
In general, {o} = [D] ({e} - {eo})
where, {Eo} = [nAT, nAT, nAT, 0, 0,
strain vector
of
•••••(3.5)
is the initial or stress-free

and
asy AT is the change in temperature of the component

E
since thermal expansion produces only normal strain (with no shear

ngi
strain and no Poisson's effect) and thermal strains do not induce any
stresses unless thermal expansion is constrained

n
The 3-D stress-strain relations are simplified below for l-D and 2-D cases.
1-D Case:

4--------_,"",f. .
P ..
,t

M_--+~ P eer x
FIGURE~
ing
2-D cases:
(}'=Ee
.ne
•••••(3.6)

(i) Plane stress case, represent) by a thin plate in X-V plane, plane
subjected to in-plane loads ~ng X-and/or V-direction, and no load
(and, hence, no stress) along., normal to the plane (in Z-direction).
t
(a) Subjected to no load~, hence, no stress) along
i.e., «(}'z = 0, Ez :;t:'0) •••••(3.7)

(I~V)1
1 v
[D]=~ v 1 •••••(3.8)
\1-v }[
o 0
.....(3.9)

Downloaded From : www.EasyEngineering.net

76 FINITE ELEMENT ANALYSIS

Py
y

ww
w.E
asy on this surface

E ngi
FIGURE 3.10

(ii) Plane strain case, represented by a thin plate in X-Y plane, which is

nee
constrained along the normal to the plane in Z-direction (i.e., no strain
along the normal). It is an subjected to in-plane loads along X-and/or
V-directions, and no load (and, hence, no stress) along the normal to the
which is constrained along the normal to the plane, has no strain along
the normal. It is an approximation of a 3-dimensional solid of vary large rin
dimension along Z compared to its dimensions along X and Y and
loaded in X-Y plane. Example: A hydro dam between two hills, which
can be considered as a set of slices or plates in the flow direction (X).
g .ne
Each slice in X-Y plane is modeled by plane strain elements.
Ez = t
I
Here, (O"z -::;:. 0, 0) •.... (3.10)

[0] = ( )(
1+ v 1- 2v
) [1 ~0 v 1 ~ ~
0
v
(1-22V)
••••• (3.11)

and {EO} = (1 + v) [ai\T ai\T oy ••••• (3.12)

(-vO"x -vO" +o"z)
This is obtained by using Ez = y =0 to represent O"z
E
in terms of o"x and O"y as O"z = v(O"x - O"y) and substituting for o"z in the
relations for Ex and Ey •

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 77

ww
w.E
asy
E FIGURE 3.11

ngi
Another example of plane strain is a segment of an axi-symmetric solid

n
which is self constrained in the circumferential or hoop direction.
• Orthotropic materials
3-D case eer
Eq. 3.2 is modified to account for variation of E and v along material axis
(1,2,3) as ing
- v 21 -V3\
0 0 0 .ne
1>,

1>2
E\
- v 12
E1
- V13
E2
1
E2
- V 23
E3
- V 32

E3
1
0 0 0 0",

0"2
t
0 0 0
1>3 E1 E2 E3 0"3
••••• (3.13)
1'2 't 12
0 0 0 0 0
123
G 12 't23

131 0 0 0 0 0 't 3 ,
G 23
0 0 0 0 0
G 31
where, E\ V2\ = E2 V12 E2 V32 = E3 V23 and E3 VI3 = E1 V31

Downloaded From : www.EasyEngineering.net

78 FINITE ELEMENT ANALYSIS

2-D Plane stress case

Eq. (3.8) is modified to account for variation of E and v along material
axes (1-2) as
E\ v 2 \E\
[0]= ( 1 ) v12 E2 E2 ..... (3.14)
1-v\2v 2 \ [
o 0

ww where, V12E2 = V2\ E\

When an orthotropic plate is loaded parallel to its material axes, it
results only in normal strains. If the material axe~ (1, 2) are oriented at an

w.E angle S w.r.t. global (x,y) axes, then

[0] = [T]T[O] [T] ...•.(3.15)

a where

sy:~:::E 1
[T]=f :~:::
2sinScosS

ngi -2sinScosS
2 sinS cosS - 2 sinS cosS cos 2s - sin 2S
..... (3.16)

(f) Compatibility equations

nee
While in 1-0 elements, only one stress component and one strain

rin
component are used in the strain energy calculation, stresses and strains
in 2-D and 3-D elements have more components due to the effect of

g
Poisson's ratio even for a simple loading. While six strain components
can be obtained from the three displacement components by partial
.ne
t
differentiation, the reverse requirement of calculating three displacement
components is possible only when the six strain components are inter-
related through the following three additional conditions, called
compatibility equations.
a2 Ex a2 Ey _ a2yxy
- - + -2- - - -
ay2 ax axay .
a2
__ 2E
Ey + __ a 2
z= _Y
a
yZ
_
&2 ay2 ay&

.•...(3.17)

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 79

(g) Equilibrium equations

Stress at a point in a component is described by the stress tensor - one
normal stress component and two shear stress components on each of the
six faces of a cube around that point. For equilibrium of this cube, these
eighteen stress components should satisfy the following equilibrium
conditions, where Fx, Fy and Fz are the forces acting on the cube along X,
Y, and Z axes.

ww + a't
aCJ x mxz _F .
xy _
ax ay---;:;;:- x' 't xy ='tyx

w .Ea
•.•••(3.18)

syE
ngi
A Jew problems are included here, based on the equations presented above.

Example 3.1

If a displacement field is described by u = -5x2- 6xy + 3y2;

determine E", Ey and 'Yxy at point x = 1 and y = -1
nee V = 4x2+ 9xy + 41

Solution rin
au
Ex = ax =-10x-6y+O=-10 x 1-(-6)(-1)=-4 g.n
E =-
Yay
av = 0 + 9x + 8y = 9 x 1 + 8(-1) = + I et
au av
'Y xy = -ay + -ax = (0 - 6x + 6y) + (8x + 9y + 0) = 2x + 15y

=2 xl + 15 (-1)=-13

Example 3.2

If a displacement field is described by u = - 4x2- 12xy + 31;

v = - 3x2+ 4xy + 21 determine Ex, Ey and 'Yxy at point x = 1 and y = -1

Downloaded From : www.EasyEngineering.net

80 FINITE ELEMENT ANAI.YSIS

Solution
au
Ex = ax =-8x-12y+O=-8 x 1-12 x (-1)=+4

av
Ey=ay =O+4x+4y=4 x 1+4x(-1)=O

au av
Yxy = ay + ax =(O-12x+6y)+(6x+4y+O)

ww = - 6x + lOy = -6 x I + I 0(-1) = - 16

w.E
Example 3.3
If a displacement field is described by u = 4x2- 4xy + 6l; v = _2X2 - 8xy + 3l

asy
determine Ex> Ey and Yxy at point x = -1 and y =-1

Solution

E
au
Ex = ax =8x-4y+O=8(-1)-4(-1)=-4

av
ngi
Ey= ay

au av
n
=O-8x+6y=-8 x (-1)+6 x (-1)=+2

eer
Yxy = ay + ax =(O-4x+ 12y)+(4x-8y+O)=4y=4(-1)=-4
ing
Example 3.4
.ne
= 4x2+ 6xy - 8y2
If a displacement field is described by u = 3x2 - 2xy + 6y2;
determine Ex, Ey and Yxy at point x = -1 al1d y = 1

Solution
V

t
au
Ex = ax =6x-2y+O=6(-1)-2 xl =-8

av
Ey=ay =O+6x-16y=6(-1)-16 x 1=-22

au av
Yxy = ay + ax =(O-2x+ 12y)+(8x+6y-O)=6x+ 18y

=6(-1)+18 xl = 12

Downloaded From : www.EasyEngineering.net

CHAPTER 3 THEORY OF ELASTICITY 81.

Example 3.5

Ifu = 2X2+ 3y, v = 3y + l find the normal and shear strains

Solution

Normal strains are

E
au
= - = 4x + 0 = 4x
Xax

ww
w.E
Shear strain is
au av
yxy = ay + ax = (0 + 3) + (0 + 0) = 3
asy
Example 3.6
E ngi
A long rod is subjected to loading and a temperature in.crease of 30°C. The total
strain at a point is measured to be 1.2 10-5 • IfE = 200 GPa and

n
x

Q. = 12 x 10-61 DC, determine the stress at the point.

eer
Solution

a = E (ETotal - ETbermal) = E (ETotal - a .AT) ing

= 200 X 109 (1.2 x 10-5 -12 X
.ne
10-6 x 30) = 200 x 109 ( - 34.8 x 10-5)

Example 3.7
= - 69.6 x 106 N/m 2or - 69.6 N/mm2
t
Consider the rod shown in Fig. 3.12 'Yhere the strain at any point x is given by
2
Ex= 1 + 2x . Find the tip displacement o.

i:;.r--------'~
11+4---- L ~I

FIGURE 3.12

Downloaded From : www.EasyEngineering.net

82 FINITE ELEMENT ANALYSIS

Solution
L

0L = fExdx = f{1 + 2X2 ~X

ww Example 3.8
In a plane strain problem, if o"x = 150 N/mm2, O"y = -100 N/mm2,
E = 200 KN/mm 2 and v = OJ, determine the value of stress o"z

w Solution
.Ea
syE
In a plane strain problem, strain Ez =
(-vO" -VO" +0" )
x
E
y z =0

100) = 15 N/mm 2
Therefore, o"z = VO"" + v O"y = OJ
ngi
x 150 + OJ x (-

Note that the value ofE is not required, but given only to mislead students.

3.1.5 SUMMARY
nee
•
rin
Based on the relative dimensions, a component may be idealised by
1-0, 2-D or 3-D elements. At every point in a component, stress and
strain are expressed by a normal component and two shear
g.n
components on each of the six faces of a small cube around that point.
Out of 18 such vector components, from equilibrium considerations,
only three normal components and three shear components in any
orthogonal coordinate system, are independent and are represented by
et
a column matrix of order (n x 1). Combined effect of these various
components will be different on planes, inclined to the coordinate
axes. Principal stresses are the maximum values on any such plane.
These values are significant while designing ~ component and can be
also obtained by graphical method, called Mohr's circle.
• The flexibility of movement of any point in the component is
identified by degrees of freedom (OaF). With reference to the
orthogonal coordinate system, the every point in a component can
have 1-6 OaFs, covering 3 translations along the three axes and
3 rotations about the three axes.

Downloaded From : www.EasyEngineering.net

CHAPTER 3 TIIEOR Y OF ELASTICITY 83

• Components, which can be modeled by a combination of

1-0 elements, are also called discrete structures. These include truss
element for axial loads (I OOF/node, along its axis), beam element
for bending loads (2 OOFs/node in each of its two planes of bending,
deflection and slope normal to axis) and torsion element for torque
load (I OOF/node, rotation about its axis). These four modes of
deformation are mutually independent and are called uncoupled OOF.

ww
•
A generalised beam element has 6 OOFs/node and is subjected to all
these loads.
Different elements of a discrete structure are joined at their end points

w.Eonly and hence only modulus of elasticity of the material is relevant

(Poisson's ratio is not relevant in their analysis). Normal stress and

asy
normal strain are related by modulus of elasticity (E) while shear
stress and shear strain are related by modulus of rigidity (G).
• Components, which are modeled by 2-D or 3-D elements, are called

E
continuum structures. These elements have surface contact along the

ngi
common boundary and hence Poisson's ratio is also relevant.
Isotropic and homogeneous material is commonly used, within its

nee
linear elastic range. Stress strain relationship is a matrix of order 3 x 3
for 2-D elements and 6 x 6 for 3-D elements. 2-D elements m~y be

rin
used in plane stress condition or plane strain condition, depending on
the particular component.
•
plane stress, 2-D plane strain or 3-D thick solid) or along with g
2-D and 3-D elements may be used with translational OOFs only (2-D

rotational degrees of freedom (2-D plate bending, 2-D thin shell and .ne
3-D thick shell) depending on the nature of applied loads.
t

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
E ngi
"This page is Intentionally Left Blank"

nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 4
DISCRETE (I-D)
ELEMENTS
ww
w.E
asy
A discrete structure is assembled from a number of 'easily identifiable 1-0

E
elements like spars, beams. Nodes are chosen at the junctions of two or more

ngi
discrete members, at junctions of two different materials, at points of change of
cross section or at points of load application. In the 1-0 element, axial

nee
dimension is very large compared to the cross section and load is assumed to act
uniformly over the entire cross-section. So, the displacement is taken as a

rin
function of x, along the axis of the member. Stress and strain are also uniform
over the entire cross section. The solution obtained in most of these cases, is
exact.

4.1. DEGREES OF FREEDOM OF DIFFERENT ELEMENTS

g .ne
Based on the relative dimensions of the element, the individual elements can be
broadly classified as 1-0, 2-0 and 3-0 elements. The load-displacement
relationships of these elements depend on the nature of loads (axial/in-plane
t
loads, torsion or bending loads) and are calculated using variational principle.
Some such elements and their degrees of freedom at each node in element
(or local) coordinate system are given below.
AxiallIn-plane loads Bending (Normal loads and/or moments)
1-0 Spar or Truss (I OOF /node) Beam (2 OOF/node for bending in I-plane)
2-0 Plane stress/Plane strain/ Plate bending (3 OOF/node)
Axisymmetric (2 OOF/node) Thin shell (6 OOF/node)
3-0 Solid (3 DOF/node) Thick shell (6 OOF/node)

Downloaded From : www.EasyEngineering.net

86 FINITE ELEMENT ANALYSIS

4.2 CALCULATION OF STIFFNESS MATRIX BY DIRECT METHOD

1-0 elements are broadly classified based on the load applied on them as spar or
truss element for axial load, torsion element for torque load and beam element
for bending in one or two planes through neutral axis/plane. Stiffiless matrix of
each of these elements can be derived based on the well known relations in
strength of materials.
(a) Truss element

ww 2 P, U

w.E
E,A,L

0' = PIA = E E = E u I L

a or

syEP= (A ElL) u =Ku .••••(4.1)

is the familiar relation for axial load carrying element fixed at one end
and load applied at the free end with displacement 'u' at the free
end 2.
ngi
In general, if loads PI and P2 are applied at the two ends of an element

nee
resulting in displacements
proportional to (U2 - UI)
UI and U2
rin
at these two ends, stress is

Then, -PI =P2 =( :E)(u -u2 l) g .ne

•••••(4.2)

In matrix notation, {:J ~:J

=[K t
where .....(4.3)

(b) Torsion element

~1--------2 ~~

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 87

From strength of materials, T = G8

J L

or T = ( ~) 8= K8
is the familiar relation for torsion load carrying element fixed at one end
and torque applied at the free end with rotation '8' at the free end 2.

ww For a general element on which torque loads TI and T2 are applied at

the two ends and corresponding rotations are 8 1 and 8 2, torque is

w proportional to (8 2 - 8,).

Then,
.Ea
syE
In matrix notation,
ngi
{~:}=[K]{::}
where, nee ••••• (4.4)

rin
g.n
(c) Beam element

From simple beam theory, forces and moments required at the two ends
ofa beam in X-Yplane to give,

(i) VI = 1 and V2 = 8 1 = 8 2 = 0 and

(ii) 8 1 = 1 and VI = V2 = 8 2 = 0 are given in the figures with upward
et
forces, counterclockwise moments, and corresponding
displacements and rotations as +ve.
6EI
131 12E1
IT
12EI
IT 2 ~ 6EI
------------------------ }!j!
L

Downloaded From : www.EasyEngineering.net

88 FINITE ELEMENT ANAL YSIS

Similar values can be obtained for the conditions

(i) V2 = 1 and VI = 9 1 = 9 2 = 0 and
(ii) 92 = 1 and VI = V2 = 9 1 = 0
Arranging these values in matrix form,

PI VI
MI 91
=[K]
ww P2
M2
v2
92

w.E [K]= EL~

12 6L -12
6L 4L2 -6L 2L2
6L

where,

asy -12 -6L

6L
12 -6L
2L2 -6L 4e

4.3
E ngi
CALCULATION OF STIFFNESS MATRIX BY
VARIATIONAL PRINCIPLE

nee
Stiffuess matrix of each element is calculated, using the principle of minimum

rin
potential energy which states that "Every component, subjected to some
external loads, reaches a stable equilibrium condition when its potential energy

g
is minimum". So, the problem lies in identifYing the set of displacements at
various points in the component which ensures that the potential energy of the
component is minimum. .ne
This is analogous to the problem in variational calculus of finding a
stationary value y(x) such that the functional (function of functions)

(~ d)
t
1= J Fr,y,
X2

Xl
d~ dx ...•.(4.6)

is rendered stationary. Integral I is stationary when its first variation vanishes

I.e." 01=0 ..... (4.7)
There are many different approaches such as Euler-Lagrange method, for
solving such problems. Interested readers are advised to refer to the relevant
books in mathematics for a more detailed presentation of the variational
calculus.

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 89

In Rayleigh-Ritz method, which is used in FEM, a mathematical expression

in the form of a power series in x is assumed for the unknown function y(x).
Then, eq. (4.7) becomes

~=O fori=O,I, ... n .....(4.8)

8a j

where ao, al>" .an are the coefficients ofthe assumed power series.

ww Finite element method is based on the variational principle where the

functional I is the potential energy of the system with nodal displacements as

w.E
the independent variables y(x) and strains as the functions dy of the
dx
independent vari~bles. This method leads to an approximate solution. Potential
energy is an extensive property i.e. the energy of the entire component is the

asy
sum of the energy of its individual sub regions (or elements). Hence eq.(4.7) can
be written as

E
01 = ~ ole = 0

ngi
Since the number and size of the elements are arbitrary, this relation is

n
satisfied only when
ole = 0

eer•••••(4.9)
In using this method, y(x) must be kinematically admissible i.e., y(x) must

ing
be selected so that it satisfies the displacement boundary conditions prescribed
for the problem. Choice of a function for the entire component satisfYing this
condition becomes difficult for complex problems. Finite element method
overcomes this difficulty by relating the primary unknown function to the
individual element, rather than to the total problem. Hence, geometry of the .ne
overall component and the system boundary conditions are of no concern when
choosing the function.
For the individual element,
t
..... (4.10)
v

where, [Ke1 is the stiffness matrix of the element

Since ~ue} represents the arbitrary nodal values of displacements, they can
not be identically zero in a loaded component.
..... (4.11)

Downloaded From : www.EasyEngineering.net

90 FINITE ELEMENT ANALYSIS

In the displacement method, calculation of stiffness matrix for an element

starts with an assumed displacement function over the element in each degree of
freedom, usually in the form of a polynomial. By substituting nodal coordinate
values in these polynomials, the unknown constants in the polynomial can be
evaluated in terms of nodal displacements. This condition necessitates choosing
displacement polynomial with as many coefficients as the number of nodal
DOF for that element. This procedure is explained in more detail in the

ww remaining sections of this chapter.

Displacement function for the element

w.E
The function shall be continuous over the entire element with no singularities
and easily differentiable to obtain strains for calculation of potential energy. The

asy
polynomial should be symmetric in terms of the global coordinate axes, to
ensure geometric isotropy.

En
Strains in the element are obtained as derivatives of the displacement
polynomial, and are thus expressed in terms of the nodal displacements.

gi
Stresses are expressed in terms of strains, using the appropriate stress-strain

nee
relationship, given earlier. By equating work done by the external forces to the
change in internal strain energy of the element and applying variational
principle, load-displacement relationships of the element in terms of stiffness

rin
coefficients are obtained. They represent a system of simultaneous equations in
terms of nodal loads and nodal displacements.

g.n
By using suitable transformation matrix, this stiffness matrix derived in local

e
coordinate system of the element, is transformed to a global coordinate system
which is common to all the elements. The stiffness matrices of all elements are
then added together such that the stiffness coefficient at a common node is the
sum of the stiffuess coefficients at that node of all the elements joining at that
t
node. This assembled stiffness matrix is square, symmetric, singular and
positive definite. This method is described in detailed here for a truss element.
A truss element is subjected to axial load only and therefore has one degree
of freedom (axial displacement) per node. Variation of axial displacement u(x)
between the two end nodes is represented by a linear relationship in the form of
a polynomial with two constants.

Let u{x)=a J +a 2 x=[1 X~:J={f{X)}T {a} ..... (4.12)

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 91

Choosing node I as the origin of local coordinate system for this element
with X-axis along the axis of the element and substituting the values of x for the
two end points of the truss element, (x = 0 at node i and x = L at node j), nodal
displacement vector {ue} or [u. UJ]T can be written as

or {Ue} = [G) {a} and {a} = [Gr 1 {ue} ..... (4.13)

ww Solving for the coefficients {a} from eq. (4.13) and substituting in eq.(4.12),

w u(x) = {f(x)T} [Gr 1 {ue }

.~[IExtla ~J{:J
=[1-
syE
~ ~}uJ=[N]T {uJ ..... (4.14)

Strain, {E}= du = {f'{x}?

dx
n[G]-l {uJ=[B]{uJ
gin ..... (4.15)

[B]~ If' (xW [G]l ~ [0 [i rJ ~[-IIL eilL]e.....

where, I] l
rin (4.16)

and {BE} = [B) {Bue}; {BE}T = {Bue}T [Bf

g.n
et
Stress, {cr} = [0] {E} = [0] [8] {u e}.
Here [0] = E, since only one axial stress component is relevant for a 1-0
truss element
From eq. (4.10),

{BuJT {pJ- J{OEeV {cr}dv=O

v
T
{Buel {Pel- J{BuJT [Bf [O][8]{uJdv=0
v

{Bu,Y [{P.l- PBY [O][B]dV {u,}] ~ 0

{Bu e ? {{Pe }- [KJ {uJ} = 0

Downloaded From : www.EasyEngineering.net

92 FINITE ELEMENT ANALYSIS

Since {oue } cannot be zero,

{Pe} = [Ke] {ue} .....(4.:17)

where, [Kel= nBf [OHB]dv= JJftBf E[B]dxdydz

= AE nBf [B}dx
v

ww since, [B] is not a function ofy or z and Hdy dz = A

I] _ I -1]
w.E AE [
- dx--
L L -1 1
..(4.:18)

4.4a syE
TRANSFORMATION MATRIX

ngi
Stiffness matrix and load vector of any element are initially derived in the local
coordinate system, with its x-axis along the element, and can vary from one

nee
element to another. A global coordinate system is common to all the elements.
If different elements have different local coordinate systems, stiffiless

rin
coefficients relating nodal load vector and nodal displacement vector can not be
combined together unless directions of load and displacements of different

g
elements joining at a common node coincide i.e., sum of two vectors is equal to
their algebraic sum only when the vectors are collinear. If the local coordinate
system of an element is inclined to the global coordinate system at an angle e, .ne
then transformation of load vector, displacement vector and the stiffiless matrix
are to be carried out before they are assembled with other elements.
t
For a truss element, if Pi, Pj , Uj and Uj represent axial load and displacement
values in the local or element coordinate system at nodes i and j and [Ke] is the
2 x 2 stiffiless matrix of the element and (Px')j, (Py')j, Uj', vi', (Px')j' (Py')j' uj'
and vj' are the components of axial load and displacement along global x and y
axes at nodes i and j then [K' e], stiffiless matrix of the element in the global or
structure coordinate system, is derived below.
Pi = (P' x)j cos e + (P' y)j sin e Uj = U'j cos e of v'i sin e
PJ = (P' x)j cos e + (P' y)j sin e

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 93

These relations can be expressed in matrix fonn as

or .....(4.19)

ww y, v

w.E
asy
E
~~------~~~--------------+X,u
u;
V'I
ngi
1+-- ----+I

These can also be expressed as nee

{p;} = [Te]T {Pel and {u~} =[Te]T {ue}
rin
{Pel = [Kel {ue} in local coordinate system can now be written in global
coordinate system as g .ne
or
we get
[Tel {Pe' J= [KeJ,[Tel {u e' }

r
{Pe'} = [Te [Kel [Tel {u:} = [K~] {u~} t
cos 29 cos9sin9 - cos 2 9 - cos9sin9
cos9 sin9 2
sin 9 -cos9sin9 - sin 2 9
[Ke'J=:E ..... (4.20)
- cos 2 9 - cos9sin9 cos 2 9 cos9sin9
- cos9sin9 - sin 2 9 cos9sin9 sin 29

Downloaded From : www.EasyEngineering.net

94 FINITE ELEMENT ANALYSIS

where, L=~(X2 -xli +(Y2 _y\)2

This can also be written in the fonn
12 1m _/2 -1m
1m -1m _m 2
m2
[ Ke'J=~ _/2
.....(4.21)
-1m 12 1m
-1m _m 2 1m m2

ww Transformation of truss element stiffness matrix In 3-D space

For a truss element arbitrarily oriented in 3-D space, a similar transfonnation

w.E
matrix can also be derived and the stiffness matrix in 3-D space can be written
using
1=(x2 -xJ. m=(Y2-Y\) and n=(Z2- Z\)

a
where,
L'

L=~(X2 _X syE j
L L

)2 +(Y2 _y\)2 +(Z2 -z\Y

in the fonn
12 ngi
1m In _/2 -1m
-In

[ Ke'J=A~
1m
In
m
mn
2
mn
n2 nee 2
-1m _m -mn
-In -mn _n 2
.....(4.23)
-p -1m
-1m _m 2 -mn
-In 12
1m
1m
m2 rin
In
mn
-In -mn _n 2 In mn n2
g .ne
4.5 ASSEMBLING ELEMENT STIFFNESS MATRICES

In a truss having three elements connecting nodes 1-2, 2-3 and 3-1, let the
element stiffness matrices (each of 4 x 4) after transfonnation to global
t
. coordinate system be:
y

~------~~---+x
1 2

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (1-0) ELEMENTS 95

For element-l joining nodes 1 and 2, the local coordinate system coincides
with global coordinate system (l = I, m = 0) and, so, displacement components
v, and V2 are zero.
(pJ, (k ll ), 0 (k 13 ), 0 u,
(Py), 0 0 0 0 VI
Therefore,
(Px )2 = (k 31 )1 0 (k33 ), 0 u2
(PY)2 0 0 0 0 v2

ww For element-2 joining nodes 2 and 3, the member is inclined to the global
axes.

w.E
Therefore,
(pJ2 (k ll )2 (k 12)2 (kl3 )2 (k14 )2 u2
(PY)2
asy
(k 2')2
(px )3 = (k 3')2
(k 22)2
(k32 )2
(k23 )2
(k33 )2
(k24 )2
(k34 )2
v2
u3
(PY)3
En
(k 41 )2 (k42 )2 (k 43 )2 (k 44)2 V3

gi
For element-3 joining nodes I and 3, the local coordinate system is
perpendicular to the global coordinate system (l = 0, m = I) and so displacement
components UI and U3 are zero.
nee
o 0 o 0
o (k 22 )3 o (k 24 )3
o 0 o 0 rin
o (k 42 )3 o (k 44 )3 g.n
Then, the process of assembling element stiffness matrices involves
combining the nodal stiffness values of all the elements joining at every
common node so that the order of the assembled stiffness matrix equals the total
e t
number of degrees of freedom of the structure.
o (k 13 ), o o o
(k 22 )3 o o o (k 24 )3
o (k 33 ), + (k ll )2 (k 12 )2 (k 13 )2 (k'4)2
o (k 21 )2 (k 22 )2 (k 23 )2 (k 24 )2
o (k 31 )2 (k 32 )2 (k 33 )2 (k34 )2
(k 24 )3 (k 4,)2 (k 42 )2 (k 43 )2 (k 44 )2 + (k 44 )3
or {P} =[K] {u}

Downloaded From : www.EasyEngineering.net

96 FINITE ELEMENT ANALYSIS

which represents a set of n simultaneous equations, where n is the total number

of degrees of freedom in the structure. In the case of mechanical structures, the
assembled stiffness matrix is symmetric, singular and positive definite.

4.6 BOUNDARY CONDITIONS

The singularity of the matrix indicates possibility of rigid body movement of

the structure in different directions and hence the possibility of many solutions

ww for the unknown nodal displacements. Boundary conditions, in terms of fixed

degrees of freedom or known values of displacem,ents at some points of the
structure, are therefore applied. In some structures, where no part of the

w.E
structure is fixed, it is possible to apply different boundary conditions. Each
solution gives displacements at other points in the structure, with reference to

a
the chosen fixed points.

syE
(a) Elimination method
The columns and rows of the stiffness matrix, displacement vector and load

ngi
vector are rearranged so that the set of equations can be written as

nee .....(4.24)

rin
where q, is the set of unknown displacements and q2 is the set of specified
displacements.

g
From static equilibrium considerations, terms in the load vector {P2}
corresponding to the fixed or specified values of degrees of freedom {q2} .ne
represent reactions at those degrees of freedom to balance the applied loads.
Taking all known values to the left side, first set of these equations can be
rewritten as
t
{Pd - [K 12] {q2} = [K II ] {q, } ..... (4.25)
The reduced stiffi1ess matrix of the structure [ Kll1 is usually non-singular
and can be inverted so that unknown displacements {qd and the reactions
{R} can be evaluated from
{qd = [Kllr 1({ P,}- [K 12 ] {q2}) ...••(4.26)

{R} = {P2 }= [K21 K 22 1{::} =[K 21 ]{ql}+ [K 22 ]{q2} .....(4.27)

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 97

In case of specified zero displacements {q2}, eq.(4.24) reduces to

••••• (4.28)
This procedure is equivalent to deleting rows and columns corresponding to
the fixed degrees of freedom from the assembled stiffness matrix,
displacement vector and load vector. The reduced stiffness matrix [Krl is a
non-singular matrix.
The unknown displacements are now obtained by using a suitable matrix
inversion algorithm like Gauss elimination method or Gauss-Jordan method

wwor Cholesky method. Eq. (4.26) and (4.27) thus simplify to

{qd = [Kllr l {Pd or {ur } = [Krr l {Pr } ••••• (4.29)
and
w.E
{R} = {P2} = [K2d {qd .....(4.30)
In the displacement formulation, displacements are calculated in the global

asy
coordinate system for the entire structure while the stresses are calculated in
each element, in the local or element coordinate system, from the nodal

E
displacements of that element using
{ae} = [O]{E} = [0][ Be 1 fUel
ngi
In a structure with 'm' fixed degrees of freedom, assembling the complete

n
stiffuess matrix and then deleting some rows and columns will involve more

eer
computer memory as well as more time. It is therefore a common practice to
ignore the rows-and columns of element stiffness matrices corresponding to

ing
the fixed degrees of freedom during the assembly process, thus storing [Krl
of order (n-m) x (n-m) only. In that case, calculation of reaction values
corresponding to the fixed degrees of freedom requires storing appropriate
terms [K2d in a different matrix.
(b) Penalty approach .ne
In this technique, a quantity C is added to the diagonal term in the row
corresponding to the specified displacement (nth OOF) while (C x qn) is
added to the force term of the corresponding equation, where C is a large
t
stiffness value, usually max (kij ) x 104 • The assembled stiffness matrix is
then inverted by anyone of the conventional approaches. This method gives
a very small value of the order of 10- 4 for the displacement qn corresponding
to the fixed degree of freedom. This displacement value can be reduced
further by using a smaller value of C. This may give rise to numerical errors
during the inversion of the stiffness matrix. Reaction is then obtained from
Rn = C x qn. This approach was used in the first general purpose software
(SAP-IV) developed by Prof. Wilson and his associates at the University of
Southern California. This method is equivalent to adding a spring of very
large stiffness value in the direction of the fixed degree of freedom. Member

Downloaded From : www.EasyEngineering.net

98 FINITE ELEMENT ANALYSIS

force calculated for the spring element indicates reaction corresponding to

the fixed degree of freedom. The size of the matrix to be inverted is large in
this approach of order n x n, without deletion 'of rows and columns
corresponding to the fixed degrees of freedom and is, therefore, !!!ll usually
adopted now.
(c) Multi-point constraints
There are many situations in trusses where the end supports are on inclined
plane and do not coincide with the coordinate system used to describe the

ww truss. In such cases, the displacement and force components along the
coordinate axes have to be resolved along and perpendicular to the inclined
plane and necessary conditions specified on them.

w.E UT = u cos e + V sin e

PT = Px cos e + Py sin e

asy V N = --U sin e + V cos e = 0

orV=tan eU

E ngi
Some other types of multi-point constraints are those linking

n
displacement of one node with that of another. A few of them are shown

eer
below, with node 1 as the fixed point and node 2 as the point of load
application. If '0' is the gap, U3 = U2 - 0 can be substituted in the load-
displacement relations to reduce the number of unknowns by 1 and
corresponding columns of stiffness matrix are modified. Accordingly, order
ofthe stiffness matrix also reduces by 1. ing
-+~-+~
r--

I .ne
--.jo~ ---+I oj.-
--+

'--

---+l8j.-
I
I
t
3 I 3
2 11 21 3 2 11
Case - I Case - 2 Case - 3

4.7 BEAM ELEMENT STIFFNESS MATRIX BY VARIATIONAL ApPROACH

Let us consider bending of a typical beam of uniform cross section in a plane

perpendicular to its axis, due to load and moment applied at its two ends. From
strength of materials, it is well known that its deformed shape is a curve and can
not be represented by a linear function. Considering 2nd order polynomial

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 99

having three coefficients, with deflection alone as the nodal degree of freedom
fails to express the three coefficients in terms of the two nodal deflections. Also,
natural boundary conditions at the ends of the beam may include not only
deflection but also the slope. Hence, a cubic polynomial is generally used and
its four unknown coefficients are represented in terms of deflection and slope
(first derivative of deflection) at each end.
If X-Y is the plane of bending, P y and Mz are the loads applied while v and
e z are the deflection and slope in the plane of bending,

wwand
Let vex) = at + a2.X + a3.x2 + a4.x3 = [1 X x 2 x3 ] {a 1

e z = dv =[0 1 2x 3x 2]{a}
.....(4.37)

or w.E W dx

{:J={U}={f(X {a} ..... (4.38)

asy
Choosing node i as the origin of local coordinate system for this element

E
with X-axis along the axis of the element and substituting x = 0 at node i and

VI 1 0 0 0 ngi
X =IL at node j, we get the nodal displacement vector,

at
(eJi
Vj
0 0
L L2
0
e
a2
a3 nee
or rUe} = [G] {a} ..... (4.39)

(eJJ 0 1 2L 3L2 a4
rin
Solving for the coefficients {a} and substituting in eq. (4.38)
u= {f(x)}T {a} = {f(X)}T [Gr t {u } e g
..... (4.40)

.ne
t
From theory of bending,
2
strain, I> = cr =L = y. de = y(d : ) = [B]{U e } ••••• (4.41)
E R dx dx

d2 v
where, -2 =[0 0 2 6x]{a}=[0 0 2 6x][G]-t {uJ
dx
I 0 0 0
0 1 0 0
and so, [8]= y [0 0 2 6x] 2 2
-31L -2/L 3/L -IlL
3
2/e lIL2 _2/L - l/L2

= (yIL3)[6(2x - L) 2L(3x - 2L) -6(2x - L) 2L(3x - L)] ..... (4.42)

Downloaded From : www.EasyEngineering.net

:100 FINITE ELEMENT ANAL YSIS

.....(4.43)
where,

2L(3x-2L) -6(2x-L) 2L(3x-L)]dx

ww 12 6L -12 6L

w.E =0
E Iz 6L 4L2
-12 -6L
-6L 2L2
12 -6L
.....(4.44)

asy
6L 2L2 -6L 4L2

4.8
E
GENERAL BEAM ELEMENT

ngi
A beam in a space frame is generally subjected to axial load, torsion load and
bending loads in two planes, due to the combined effect of loads acting at

nee
different locations of the space frame and in different directions, as shown in the
figure.

rin
g .ne
t
• •..•.•.•...........•... ~ X

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS j,0i,

If a single beam AB in a space frame with concentrated loads Ph P2, P3 and

P4 acting on some members is considered, load PI contributes to axial load
giving rise to displacement u; load P2 contributes to bending in X-V plane
giving rise to deflection V and slope Sz, load P3 contributes to bending in
X-Z plane giving rise to deflection wand slope Sy and load P4 contributes to
torsion in AB giving rise to S2. Therefore, general stiffuess matrix for beam AB
in its local coordinate system should include response ofthe beam for these four
types of deformations, which are mutually independent. Such OOFs are called
uncoupled degrees of freedom i.e. torsion of a beam does not result in axial

ww
elongation or compression of the beam; deflection of the beam in X-Y plane
does not cause any d.isplacement of the beam in X-Z plane etc. Hence, stiffness

w.E
contribution of a beam in these 6 OOFs can be placed directly, without any
modifications, in the appropriate positions of the general stiffness matrix of
order 12 (2 nodes x 6 OOF/node) .
y

a v l ,DOF2

j-: syEn
Yl . DOF 5
u , DOF 1 9 x2, DOF 10
l
x ~~~~----------------~---r-+~,DOF7
9 xl , DOF 4
Z
gin
WI' DOF 3 9zl , DOF 6 9Z2 ' DOF 12

eer
Stiffness matrix with only axial load response (relating load Px and
displacement u) is
Pxl EA
ing
0 0 0 0 0 -EA 0 0 0 0 0 ul
Pyl
Pzl .ne
o 000 0 0 o 0 0 0 0 0
o 0 0 0 0 0 o 00000
VI

Mxl

Myl

Mzl
t
o 0 000 0 o 0 0 0 0 0
o 0 000 0 o 0 0 0 0 0
o 0 0 0 0 0 o 00000
Sxl

SYI

Szl
..... (4.45)
Px2 L -EA 0 0 0 0 0 EA 0 0 0 0 0
Py2 o 0 000 0 o 0 0 0 0 0
Pz2 o 000 0 0 o 0 0 0 0 0
Mx2 o 0 000 0 o 0 0 0 0 0
My2 o 000 0 0 o 0 0 0 0 0
Mz2 o 0 0 000 o 0 0 0 0 0

Downloaded From : www.EasyEngineering.net

102 FINITE ELEMENT ANALYSIS

Stiffness matrix with only torsion load response (relating load Mx and
displacement 9x) is

Pxl 0 0 0 0 0 0 0 0 0 0 0 0 ul
PYI 0 0 0 0 0 0 0 0 0 0 0 0 VI

Pzl 0 0 0 0 0 0 0 0 0 0 0 0 WI

0 0 0 GJ 0 0 0 0 0 -GJ 0 0 9 xI

ww
Mxl

My) 0 0 0 0 0 0 0 0 0 0 0 0 9 YI
0 0 0 0 0 0 0 0 0 0 0 0 9 z1

w.E
Mzl
Px2
=L 0 0 0 0 0 0 0 0 0 0 0 0 u2
.. (4.46)

Py2 0 0 0 0 0 0 0 0 0 0 0 0 v2

a
Pz2
Mx2
0 0 0

0 0 0
0

syE
0 0 0 0 0
0 0 0 -GJ 0 0 0 0 0
0 0 0 0 0 0
0
GJ
0
0 0
0 0
0 0
w2
9 x2
9 y2

ngi
MY2

Mz2 0 0 0 0 0 0 0 0 0 0 0 0 9 z2

nee
Stiffness matrix for bending in X-Y plane (relating Py and
displacements V and 9 z) is
M z with

rin
P XI

P yl

P zl
0
0
0
0
12EI,le
0
o
o
o
0
0
0
0
0
0
6EIJL
0
0 0
o
0
0
-12EI z /e
0
o
o
g
0 0
0 0
0 0 0
0
6EI,/L
0
.ne
UI

MXI

Myl

Mzl

P X2
0
0
'I 0
--
L 0
0
0
6EI z /L
0
0 0 0
0 0 00
0 0 0
0 0 0
0
0
4EI z
0
0
0
0
0
0
0
-6EI z /L
0
0 o 0
0 o 0
0 0 0
0 0 0
0
0
2EI,
0
8,1
8 YI
8,1
u2
t
P y2 o -12EI z /e 0 0 0 -6EI,/L 0 12EI z Ie 0 0 0 -6EI z /L v2
P,2 0 0 0 0 0 0 0 0 0 0 0 0 w2
Mx2 0 0 0 0 0 0 0 0 0 o 0 0 8 x2
MY2 0 0 o 0 0 0 0 0 000 0 8 y2
M,2 0 6EI z /L o 0 0 2EI, 0 -6El z /L 000 4EI z 8z2

.....(4.47)

Stiffness matrix for bending In X-Z plane (relating Pz and My with

displ.acements wand 9y) is

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (1-0) ELEMENTS 103

o 0 o o o 000 o o o
o 0 o o o 000 o o o
Pzl 0 0 12EIylL2 0 6El y /L 0 0 0 -12Ely/L2 0 6ElyIL 0 WI

Mxl 0 0 0 0 0 0 0 0 0 0 0 0 9. 1
Myl 0 0 6El y IL 0 4EIy 0 0 0 -6El y IL 0 2EIy 0 Oyl
M.. 0 0 0 0 0 0 0 0 0 0 0 0 9 ..
p. 2 L 0 0 0 0 0 0 0 0 0 0 0 0 u2
Py2 0 0 0 0 0 0 0 0 0 0 0 0 V2

Pz2 0 0 -12EI y /L2 0 -6EI y /L 0 0 0 12Ely/L2 0 -6EI y /L 0 w2

ww
M.2
My2
0
0
0
0
0
6El y /L
0 0
0 -12EIy/L2
0
0
0
0
0 0
0 -6EI y /L
0
0
0
4EIy
0
0
9. 2
9 y2

w.E
Mz2 0 0 0 0 0 0 0 0 0 0 0 0 9. 2
.••••(4.48)
The combined stiffness matrix of a general beam element thus includes

a
stiffness coefficients linking loads along the three coordinate axes and moments

syE
about the three axes at each end of the beam to the corresponding displacements
and rotations, and is obtained by a simple addition of the coefficients of the
above four matrices.
P",
Pyl
EA
0
o
12E1,1L'
o
o
o
o
0
0
o
6EI,1L
ngi -EA
0
0
-12EI,IL'
o
o
0
0
o
o 6E~/L 1~:

nee
Pzl 0 o 12EI,IL' o 6EI,1L o 0 0 -12EI,IL' 0 6EI,1L o w,
M", 0 o o GJ 0 o 0 0 o GJ o o Ox!
M" 0 o 6EI,1L o 4EI, o -6EI,1L 0 2EI, o 9)'1

M" =! 0 6EI,IL o o 0 4EI. 0 - 6E,IL o 0 o 2EI, 6"

P",
P"
P"
L -EA
0
0
o
- I2EI,IL'
o
o
o
-I2EI,IL'
o
o
o
0
0
-6EI,1L o
o
- 6EI,IL
EA
0
0
0
I2EI,IL'
o
o
o
I2EI,IL'
0
0

rin
o
o
0 -6EI,1L
o
-6EI;JL
o
"1
V2

w,
M",
M"
Mol
0
0
0
o
o
6EI,1L
o
6EI,1L
o
-GJ
o
o
0
2EI,
0
o
o
12EI,IL'
0
0
0
o
0
- 6EI,1L
o
-6EI,1L
o
GJ
0
0
o
4EI,
o
g o
4E1,

•• (4.49)
8",
8"
8"

.ne
4.9 PIPE ELEMENT

A pipe element is essentially a one-dimensional element subjected to:

t
(i) distributed load due to self weight and weight of fluid inside,
(ii) concentrated loads in the form of pipe fittings like valves,
(iii) axial loads due to change of direction of fluid flow in pipe bends,
T-joints etc. and due to restrained thermal expansion.
It is thus similar to a generalised beam element, except that stiffness of a
beam member is a function of a geometric properties (I, L) and material
property (E), whereas stiffness of a pipe element is a function of load (internal
pressure) also. Stiffness increases due to internal fluid at high pressure (also
called pressure stiffening).

Downloaded From : www.EasyEngineering.net

:1.04 FIN ITE ELEMENT ANAL YSIS

For these reasons, this is listed as a different element, in many general

purpose software even though the analysis is completely similar to that of a
three-dimensional frame (with general beam elements).

4.10 SUMMARY

• Finite Element Method (FEM) is based on mlDlmum potential

energy of the system, as applied to a model of the component
consisting of finite number of elements connected at common

ww nodes. In the displacement method, which is commonly used in the

solution of mechanical problems, load-displacement relations (or

w.E stiffness coefficients) are calculated for each element satisfYing the
condition that variation of potential energy (sum of work done and
strain energy) for any virtual displacement is zero.

a • In 1-0 elements, a polynomial expression is assumed for the

syE
variation of displacement along its axis. Axial load carrying truss
element is identified by its axial displacement (1 OOF/node) at its
two end points (nodes) and is modeled by a linear variation of

ngi
displacement along its axis. A torsion element is similarly identified
by the rotation about its axis (1 OOF/node) at its two end points

nee
(nodes) and is modeled by a linear variation of rotation along its
axis. Beam element is identified by deflection normal to the axis
and slope (derivative of displacement) at its two end points

rin
(2 OOF/node) and is modeled by a cubic polynomial for the
deflection, in each of its two planes of bending. Stiffness matrix of

•
g
a general beam element with 6 OOF per node is obtained by a

.ne
combination of the above, treating them as uncoupled responses.
Stiffness coefficients relate nodal displacement components with
nodal load components, both being vectors. These vector
components of different elements can be combined only when they
are all oriented in the same directions. Hence, element stiffness
matrix obtained in its local coordinate system, defined w.r.t. its
t
axis, is transformed into a common or global coordinate system,
using a transformation matrix (function of orientation of the
member w.r.t. global coordinate system) if different elements are
inclined to each other.
• The assembled stiffness matrix represents unconstrained system
(with rigid body modes included) and hence can not be inverted (or
has infinite solutions). Specified displacement boundary conditions
are applied to avoid rigid body modes and thus obtain a unique
solution for the given problem.

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DI"l RE I F (I-D) EI I'MI'NTS 1.05

• Onc· of thc standard techniques for the solution of simultaneous

equations \\ III give primary unknowns (nodal displacements) in
glooal coordinate system, since the nodes are common to many
elements. Using stress-displacement relationship of each element,
stresses in local coordinate system are ootained from the nodal
displacements.

OBJECTIVE QUESTIONS

ww
1. Transformation matrix
(a) is always same
for all elements
(b) is different

w.E
(c) depends on element axes
2. Transformation matrix relates
(d) depends on material
in element coordinate system with

asy
structural coordinate system
(a) displacements (b) stresses
(c) stiffness coefficients
En
(d) material properties
3. Primary variable in FEM structural analysis is
(a) displacement (b) force (c) stress
gin (d) strain

eer
4. A singular stiffness matrix means
(a) unstable structure
(b) one or more DOF are unrestrained
(c) wrong connectivity of elements ing
(d) wrong solution expected
.ne
t
5. One possible load in structural analysis is the specified
(a) nodal temperature (b) stress in an element
(c) heat flow (d) strain in an element
6. Assembled stiffness matrix after applying houndary conditions is NOT
(a) square (b) symmetric (c) handed (d) singular
7. Detenninant of assembled stiffness matrix before applying boundary
conditions is
(a) < 0 (b) = 0 (c) > 0 (d) depends on the problem
8. Determinant of assembled stiffness matrix after applying boundary
conditions is
(a) < 0 (b) = 0 (c) > 0 (d) depends on the problem

Downloaded From : www.EasyEngineering.net

2.06 FINITE ELEMENT ANAL YSIS

9. A pipe with internal pressure behaves _ _ a hollow pipe 0"[ same section
(a) with exactly same deflection as
(b) with lesser bending deflection than
(c) with more bending deflection than
(d) with a different type of deflection
10. Any point in a structure can have maximum of _ OOF

ww (a) 2 (b) 3
II. A 1-0 structural element is a
(c) 4 (d) 6

w.E(a) truss element (b) beam element (c) pipe element (d) all of them
12. Meshing for 1-0 elements is

asy
( a) essential (b) optional

E
( c) reduces input data

(a) truss elements ngi

(d) depends on other data
13. A structure with loads at joints only is usually modeled by _ __
(b) beam elements
(c) pipe elements
n (d) anyone ofthem
eer
14. A frame with nodal loads only is modeled as an assembly of truss elements,

(a) very small (b) very large ing

if resistance to rotational degree of freedom of joints is

(c) not related (d) depends on other data

.ne
15. A frame with nodalloarfs only is modeled as an assembly of beam elements,
if resistance to rotational degree of freedom of joints is
(a) very small (b) very large
t
(c) not related (d) depends on other data
16. A frame with distributed loads along members is modeled by an assembly
of elements
(a) truss (b) beam (c) pipe (d) anyone of them
17. A frame with welded joints can be approximated by truss elements
(a) always (b) sometimes
(c) never (d) depends on assumed flexibility of rotation

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (1-0) ELEMENTS 107

18. A structure assembled by multiple bolts/rivets at each joint is modeled by

truss elements
(a) always (b) sometimes
(c) never (d) depends on assumed flexibility of rotation
19. Stress across any 1-0 element is assumed to be constant
(a) true for beam elements (b) true for truss elements
(c) true for pipe element (d) true for all 1-0 elements

ww
20. A bar is mooeled as 1-0 element only if its
(a) area of cross section is small

w.E
(b) moment of inertia is small
(c) length is very large compared to cross section dimensions
(d) all the above
asy
21 . A truss element in space has a stiffness matrix of order
(a) 2 x 2 (b) 4 x
E
4 (c) 6 x 6
22. A spring element is similar to _ _ element
ngi
(d) I x 1

(a) truss (b) beam (c) pipe

23. A plane truss element has a stiffness matrix of order nee
(d) anyone of them

(a) 2 x 2 (b) 4 x 4 (c) 6 x 6

24. A pipe element differs from a beam element by inclusion of
(d) 1 x 1
rin
(a) cold cut
(c) anchors
(b) internal pressure stiffening
(d) sliding supports
g .ne
25. Stiffness matrix of a torsion element is of the same order as
(a) truss element
(c) pipe element
(b) beam element
(d) none of them
t
26. A spring of stiffness at the supports is used for calculating support
reactions in penalty approach
(a) very small (b) same as other connected members
(c) very large (d) sum of connected members
27. Penalty approach leads to displacements at supports
(a) zero (b) very small
(c) significant (d) depends on stiffness of connected members

Downloaded From : www.EasyEngineering.net

108 FINITE ELleMENT ANALYSIS

28. Penalty approach takes time

(a) more (b) less
(c) depends on other data (d) no change
29. Accuracy of sol ution __ with increase of number of beam elements

(a) improves (b) reduces

ww (c) no change (d) depends on other data

w.E
SOLVED PROBLEMS

Example 4.1

a syE
Determine the nodal displacements and element stresses by finite element
formulation for the following figure. Use P=300 k N; A,=0.5 m2; A2=1 m2;
E=200 GPa

ngi
300 kN
...... A I •••••••••••••• nee
:--+.................................
AI =0.5 m
2
A2 = 1 m
2

rin
E=200GPa

j.-lm
Solution
g .ne
The structure is mo~aled
2-3 and 3-4 as shown below
with 3 axial loaded elements connected by nodes 1-2,
t
------ 2 ----- 3 -------4

Stiffness matrices of elements I, 2 and 3 are given by

where, k,= A,E/L, = 0.5 x 200 x 10911.0 = 1.0 x 10"

and k2 = A2 E.':..2= 1 x 200 x 109/2.0 = 1.0 x 10"

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (1-0) ELEMENTS 109

Assembled stiffness matrix is obtained by adding corresponding terms as,

kl -kl 0 0 -I 0 0
-kl kl +k2 -kl 0 -I 2 -I 0
kl=[K]= =1.0x lOll
0 -kl kl +k2 -k2 0 -I 2 -I
0 0 -k2 k2 0 0 -I
•..•. (4.32)

ww
are
Corresponding assembled nodal load vector and nodal displacement vector

p= w.E
300,000
0

0
;q=
UI
u2
u3
•••••(4.33)

R
asy
u4

Thus,
[K] {q} = {P} E ngi
or

-1
-1 0
-1
0 ul 0
nee
rin
2 0 u2 300,000
1.0 X 1011 ••••• (4.34)
0 -1 2 -1 u3 0
0 0 -1 u4 R

After applying boundary condition, U4=0, the fourth row and fourth column g .ne
are removed resulting in

1.0xlOII -11 -12 -1 01 {Uu I}~ {300,000

2
0 } •••••(4.35)
t
ro -1 2 u3 0

Solving the above set of equations gives,

UI = 6 x lO-6 m; •••••(4.36)

Stress in element-I,

al = E [Bd {ql } = E [-IILI IIL I] {:J = 0 N/m 2

Downloaded From : www.EasyEngineering.net

110 FINITE ELEMENT ANALYSIS

Stress in element-2,

0"2 = E [B 2] {q2} = E [-I1L\ l/Ld {~:} = - 6 x 10 5

N/m
2

Stress in element-3,

0"3=E[B3] {q;}=E[-I/L2 I1L 2] {~;} =-3 x 10 5

N/m
2

ww Example 4.2

w.E
An axial load P=200x 103 N is applied on a bar as shown. Using the penalty
approach for handling boundary conditions, determine nodal displacements,
stress in each material and reaction forces.

asy ,
I -A 1 =2400mm-;E 1 =70 x
9'
]0 N/m-

I+-- 200 ~
E 300 -+I+-
ngi
200 --+I
2 - A2 - 600 I11Ill2; £2 - 200 x
(I
]0 N/m
2

Solution

-----2----- 3----- 4 nee

rin
Considering a 3-element truss model, stiffness matrices of elements 1,2 and 3
(connected by nodes 1,2; 2,3 and 3,4 respectively) are given by,

[K\]=[ k\ -k\]; [K 3]=[ k\ -k\]; [K 2]=[ k2

g .ne
-k 2] ..(4.37)

where
-k\ k\ -k\ k\

k\ = AJE\/L\ = 2400 x 70 x 1031200 = 84 x 104

-k2 k2
t
and k 2 =A2 E2 /L 2 =600 x 200 x 103/300=40 X 104
Assembled stiffness matrix is obtained by adding corresponding terms as,
k\ -k\ 0 0 84 -84 0 0
-k\ k\ +k\ -k\ 0 -84 84+40 -40 0
[K]= =10 4
0 -k j k\ +k2 -k2 0 -40 40+80 -84
0 0 -k2 k2 0 0 -84 84

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS

Corresponding assembled nodal load vector and nodal displacement vector

are
0 uj
200,000 u2
p= q=
0 u3
0 u4

For the penalty approach, C = max(k,) x 104 = 124 x 104

wwSince the bar is fixed at nodes I and 4, the equations are'then modified using
Cas,

w 0
200,000
.Ea
=10 4
84+ 124x 10 4
-84
-84
124
0
-40
0
0
0
0
syE
0
0
-40
0
124
-84
-84
84+ 124x 10 4

From 4th eqn.

or U4 = 6.7737 X 1O- 5 11 3 ngi
0 = 10 4 [-84 U3 + (84 + 124 x 104) U4]

From 3 rd eqn 0 = 104 [-40 U2 + 124 113 - 84 U4]

nee
substituting for 114 from the above,
U3 = 0.3226 U2 rin
2 nd eqn now becomes,
g.n
or
200,000 = 104 [-84 Uj + 124 U2 - 40 U3]

I st equation gives,
et
0= 104 [(84 + 124 x 104) UI -84 U2]
From these two equations,
Uj = 1.2195 x 10-5 mm; U2= 0.180034 mm
Substituting in 3 rd and 4th eqn.,
U3 = 0.058079 mm; U4 = 3.9341 x 10-6 mm
Reactions, RI = -CUI = (124 x 108 ).(1.2195 x 10-5) = -151.22 x 103 N

Downloaded From : www.EasyEngineering.net

112 FINITE ELEMENT ANALYSIS

Stresses in the elements,

0") = E)E) = E)B)q)_2

= 70 x 103 [2 1] {1.2195 x 10-

200 200 0.180034
5
}

= 63.01 N/mm 2
0"2 = E2E2 = E 2B 2q2-3

ww = 200 x 103 [-=l I]

300 300
{0.180034}
0.058079

w.E 0"3
= -81.3 N/mm 2
= E3E3 = E)B)q3-4

asy =70x10 3 [-1

- - 1]{ 0.058079 }
200 200 3.9341 x 10-6

Elimination method
E = -20.3 N/mm 2
ngi
nee
Since the bar is fixed at nodes 1 and 4, corresponding rows and columns of the
assembled stiffness matrix are deleted, resulting in {P}R = [K]R {uh

or {200,000} =10 4 [124 - 40] {u 2} rin

o -40 124 u 3

Solving these two simultaneous equations, we get g .ne

and
U2= 155/861 =0.180023mm
U3 = 501 861 = 0.058072 mm
Reactions can now be obtained by substituting the nodal displacements in
t
the deleted equations of the assembled stiffness matrix.
R) = 104 [(84 + 124 x 104) -84 0 0] [u) U2 U3 U4]T
=-84 x 104u2 =-84 x 104 (155/861) = 151219N
~ = 104 [0 0 -84 (84 + 124 x 104) ] [u) U2 U3 U4]T
= -84 x 104 U3 = -84 x 104 (50/861) = 48780 N
These reaction values are identical to hose obtained by the penalty approach

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 1.:1.3

Check: For force equilihrium of the structure,

Rl + ~ = Applied load P :::: 200 kN

This equation is satisfied with the results obtained
Note that results by penalty approach match very closely with those by
elimination approach.

Example 4.3

ww
Consider the truss element with the coordinates I (10,10) and 2 (50,40). If the

w
displacement vector is q=[ 15 10 21 43]T mm, then determine (i) the vector q'
(ii) stress in the element and (iii) stiffness matrix if E=70 GPa and A=200 mm2

Solution: .Ea
syE
(i) The nodal displacement vector in local coordinate system

I
{q'} = [ 0
mOO]
0 1 m {q}
ngi
nee
where 1= (x2-x\)/L and m=(YrYl)/L are the direction cosines of the element
Length of the element,
L=~(X2 -X 1)2 +(Y2 -Yl)2 =~(50-10)2 +(40-IOi =50mm rin
1=(50-10)=i. m=(40-1O)=~ g.n
et
50 5' 50 5
15

{q,}=[4~5 3~5 4~5 3~5] ~~ -{;10;~5}

(ii) Stress in the element, (J = E = E[ ~I

E ~] {q'}

=70000 [ -- 1 1
- ] { 90/5 }
, 50 50 213/5
= 34.44 x 103 N/mm2

Downloaded From : www.EasyEngineering.net

114 FINITE ELEMENT ANALYSIS

(iii) Stiffness matrix of the element,

12 1m _/2 -1m 16 12 -16 -12

[K]= AE 1m m2 -1m _m 2 200x 70,000 12 9 -12 -9

L _/2 -1m 12 1m 50x25 -16 -12 16 12
-1m _m 2 1m m2 -12 -9 12 9

Example' 4.4

ww Determine the stiffness matrix, stresses and reactions in the truss structure
shown below, assuming points 1 and 3 are fixed. Use E = 200 GPa and A =
1000 mm 2•

w .Ea
Solution
Stiffness matrix of any truss element is given by

[K]= AE
12
1msyE m2
1m
-1m _m 2
_/2 -1m

L _/2
-1m
-1m
_m 2
12
1m
1m
m2 ngi
nee
T rin
[i]
g.n
500mm Y
et
1 yI---+ .x

1-- 750mm

In the given problem, L] = 750 mm; L2 =~[7502 + 500 2] = 250 JU

For element-I, 1= (X2 -x]) =1 and
L]

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 115

o -1 0
K] _ AE 0 0 o 0
[ 1-
750 -1 0 1 0
o 0 o 0

AE = 266.67 x 10 3
750

ww Forelement-2, 1= (X3 -X 2 )=_3_ and

L2 J13

[K] _
w.E 6
9
4 -6 -4
6 -9 -6

asy
AE
2 - 250 x 13 x J13 -9 -6 9 6
-6 -4 6 4

AE
250x13J13
E
= 17.07 x 10 3
ngi
nee
The assembled stiffness matrix is given by appropriate addition of stiffness
coefficients of the two elements,
266.67 0 -266.67 0 0 rin0

[K]= 10
0 0 0
3 -266.67 0 266.67 -153.63
0
102.42
0
-153.63 -102.42
0
g .ne
0 0
0
0
-102.42
0
0
-153.63
-102.42
68.28
-102.42
-68.28
-102.42
153.63
102.42
-68.28
102.42
68.28
t
After applying boundary conditions that UI = VI = U3 = V3 = 0, the load-
displacement relationships reduce to {P}R = [K]R {U}R

0 }=103[-266.67+153.63 102.42]{U 2 }
{ - 50000 102.42 68.28 v 2

Solving these two simultaneous equations gives

Downloaded From : www.EasyEngineering.net

1.1.6 FINITE ELEMENT ANALYSIS

Displacements of element-l in local coordinate system are given by

o
{ql'}=[~ ~ ~ ~] 0.281~ ={0.281~}
-1.154
Stress in element-I, a 1 =EE 1 =E[-11L llL]{q\}

ww = 200 x 103 x 0.2813 1750 = 75 N/mm2

Displacements of element-2 in local coordinate system are given by

w.E{ ,}=[3/JIj 2/JIj 0 0]

0.28313
-1.154

aq2 0

syE
0 3JIj 2JIj

Stress in element-2, a 2 =EE 2 =E[ ~1 ~J{q/}

0
o

ngi
=200xl0 3 x (-0.406) =90.08N/mm2
250JIj
nee
Reactions at the two fixed ends are obtained from the equations of th~

rin
assembled stiffness matrix corresponding to the specified zero displacements
o
R 1- X
R 1_ y
=10 3
266.67
0
0 - 266.67
0 0
0
0
0
0
0
0 g .ne
0
0.2813

t
R 1- X 0 0 -153.63 -102.42 153.63 102.42 -1.154
R 1_ y 0 0 -102.42 - 68.28 102.42 68.28 0
0
-75014.3
o
N
74976.6
49984.4
The exact solution can be obtained from the equilibrium conditions as
follows -
The force in element-2 is such that its vertical component IS equal to the
applied load P. Horizontal component ofthis force is given by
P x (750/500) = 75000 N

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (1-0) ELEMENTS 117

R3- y + P = 0 or R3_y = 50,000 N

R3-X + R\_x = 0 or R\-x = - R3-X = 75,000 N
It can be seen that the approximate solution obtained by FEM is in close
agreement with the exact solution obtained from equilibrium consideration.

Example 4.5
Determine the nodal displacements, element stresses and support reactions in
the truss structure shown below, assuming points 1 and 3 are fixed. Use E =

ww
70 GPa and A = 200 mm 2.

Solution

w
Stiffness matrix of any truss element is given by

.Ea
12 1m _/2 -1m
m2 _m 2
[K]AE 1m
L _/2 -1m
_m 2
syE
-1m
12 1m
m2

ngi
-1m 1m

In the given problem, L\ = 500 mm ; L2 =~[4002 +300 2] =500mm

For element-I, 1= (X2 - XI) =1 and m=(Y2-Y\)=0nee

L\
1 0 -1 0
L\

rin
[Kl = AE 0 0
500 -1 0
0 0
0 g.n
AE =28xI0 3
o 0 0 0
et
500

21 P 12 kN

---f-- /'~~------------------------~I~
300 mmj :
Y I

1 i
:
!
I
-- ---~--------------------------------~-----------------
: I
;. 500 mm .:" .
400 mm ----+l "

Downloaded From : www.EasyEngineering.net

:1.18 FINITE ELEMENT ANAL YSIS

For element-2, 1= (X3 -x 2)=i and m=(Y3-Y2)= -3

L2 5 L2 5
16 -12 -16 12
[K] _ AE -12 9 12 -9
2 - 500x25 -16 12 16 -12
12 -9 -12 9

ww AE 28x 10 3
---=---
25x500 25

w The assembled stiffness matrix is given by appropriate addition of stiffness

.Ea
coefficients of the two elements,
0 -1 0 0 0

[K]=28xl0 3
0 0
syE 0 0 0 0
-1 0 1+(16/25) -12/25 -16/25 ]21125
0 0
0 0
-12/25
-16/25
9/25
12/25
ngi
]2/25 -9/25
16/25 -]2/25
0 0 12/25 -9/25 -12/25 9/25
nee
After applying boundary conditions that UI = VI = U3 = V3 = 0, the load-
displacement relationships reduce to {P}R = [K]R {U}R

{-12000O} =28xlO3[]+(16/25)
rin
-12/25
-12/25]{U 2}
9/25 v 2
g.n
Solving these two simultaneous equations gives
U2 = - 417 mm and V2 = - 4112] mm
Displacements of element-l in local coordinate system are given by
et
o

II 0 0 0]
, {ql}= 0 0 1 .0 -4/~ ={-4/~}
-41121

Stress in element-I, ° =E1 61 = E[ ~1 ~ ] {ql}

_ 70xl0
-
3
X
(-4) _ SON/mm
- --
2
500 7

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 1.1.9

Displacements of element-2 in local coordinate system are given by

-417
= [41 5 - 3 15 0 0] - 411 21 . = {I 5 121}
{ }
q2 0 0 4/5 -3/5 0 0
o
Stress in element-2, a 2 = E = E[ ~1 ~}q
E2 2}

ww =_ 70 X 10 x
3
(.!2)
=-100 N/mm 2

w .Ea
500 21
Reactions at the two fixed ends are obtained from the equations of the
assembled stiffness matrix corresponding to the specified zero displacements
o
R I-X
R I_y
0
0 0
syE
-1
o
0
0
0
0
0
0
0
-417
R 3- X
=28xl0 3
0 0 -16/25 12/25
ngi
16/25 -12/25 - 41/21
o
R 3- y 0 0 12125 -9/25 -12/25 9/25

nee o
16000
0
rin
g.n
= N
-16000
12000

Example 4.6
Estimate the displacement vector, stresses and reactions for the truss structure as
et
shown below.

Solution

Stiffness matrix of any truss element is given by

12 1m _12 -1m
1m -1m m2 _m 2
[K]= AE
L _/2 -1m 12 1m
-1m _m 2 1m m2

Downloaded From : www.EasyEngineering.net

120 FINITE ELEMENT ANALYSIS

Tmm ~
30n [2]
y

ww 1 Lx 200N

w.E 14-----
.. 400 mm-----~.I

asy In the given problem,

En L J =400 mm;

gi
and
1.2 = 300 mm;

L3 =~4002 + 300 2 =500mm nee

For element-I, 1= (X2 - x J) =1 and
LJ rin
m = (Y2 - YJ) =0
LJ

For element-2, 1=
(x 3 -x 2) = 0 and m=(Y3-yJ -1 g.n
0 0
I
L2
0 -1
0 0
0
L2

0
0 0 0
1 0 -1
0 et
[K] _ AE . [K] _ AE
J - 400 -I 0 1 0 ' 2 - 300 0 0 0 0
0 0 0 0 0 -1 0

For element-3, 1= (X3 - x J) = i and m=(Y3-YJ)= -3

L3 5 L3 5

16 -12 -16 12
-12 9 12 -9
[Kh-~
12500 -16 12 16 -12
12 -9 -12 9

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS

The assembled stiffness matrix is given by appropriate addition of stiffness

coefficients of the two elements,
114+ 16/25 -121125 -1/4 0 -161125 12 I 125
-121125 9/125 0 0 121125 -91125
-1/4 0 1I 4 0 0 0
[K]= AE
100 0 0 0 1/3 0 -1/3
-16/125 121125 0 0 16/25 -12/125

ww 121125 -91125
After applying boundary conditions that UI = VI
0 -1/3 -121125
= V2 =
113 + 91125
0, the load-

w.E
displacement relationships reduce to {Ph = [K]R {uh

rOCOS30 } [/4
16/~J{:J
0

a
- 500Sin 30
-200
= AE
100
0

sy
0
1I 3
0

fj
En fj
These three equations give u 2 =400x500--=100000-mm

- 500 x 300 75000

=--A-E- mm
2AE AE

gin
- 200 x 12500 - 250000
v2 = 2AE u3 = 16AE

Displacements of element-3 in local coordinate system are given by eer 16AE mm

{ ':.1 t =[4/5 -3/5 0 0] 0

o
i
{O} ng.
q3f 0

Stress in element-I,
0 il/5 -3/5 -250000/16AE = 200000/16AE
o net
EU 2 fj 250fj 2
(J 1 = E I:: I = - - = Ex 100000 ( )= N/mm
LI 400AE A

Stress in element-2,
_E _ E v 2 _ E x 75000 _ 250 N I 2
(J2 - 1::2 -"L; - (300AE) --;;: mm

Stress in element-3,
_E - E q 2 _ E x 200000 _ 400 N I 2
(J3 - 1::3 -~ - (500AE) --;;: mm

Downloaded From : www.EasyEngineering.net

122 FINITE ELEMENT ANALYSIS

Reactions are obtained from the equations corresponding to the fixed OOFs
in the assembled stiffness matrix as given below:
ul
VI

_(AE)[(!+~) 12 -1 -16 u2
R I-x 0 12 ]
- 100 4 125 125 4 125 125 v2
u3

ww ~ ~~[( ~too~H ~~:)( -)2::EOO)]

w.E =

asy
-250J] +200N
ul

=(~~)[ ~;: 1~5 En ;2~] ~:

R I_y 0 0
gin
/225

eer
= AE [(12/125)(-250000116 AE)] = -150 N
100 ing
.ne
-12
125 t
= AE
100
[(=!)
3
x (- 75000) + (~) (- 250000)] = 400 N
AE 125 16AE
Check: For equilibrium of the truss, from basic strength of materials,
R I -x = -500 cos 30 +200 N
and R I _y + R3-y = 500 sin 30 = 250 N
These two equations are satisfied by the results obtained

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 123

Example 4.7

For the three bar truss shown in figure below, determine the displacements of
node 'A' and the stress in element 3.

1+--450 ~I" 450 ~I" 350--.j

V! DT
ww
w.E A I__________________I x

asy18KN

E
A = 250 mm 2 ,• E = 200 GPa

ngi
For a truss element with direction cosines I and m, w.r.t. global X-axis,

nee
12 1m _/2 -1m
1m m2 -1m _m 2
[Ke'J= ALE
r= in
_12 -1m 12 1m
-1m _m 2 1m m2

where, Elem.No. Nodes

A B
L(mm)
750
I=X/L
-0.6
m (Y/L)
0.8
g .ne
2
3
AC
AD
1000
750
0.8
0.6
0.6
0.8
t
After assembling and applying boundary conditions, u = v = 0 at b, c and d
we get

O} [
3
10 )[ 0.36/3 + 0.64/4 + 0.36/3 - 0.48/3 + 0.48/4 + 0.48/3] {u I}
= 250x200x-
{ - 18000 250 - 0.48/3 + 0.48/4 + 0.48/3 0.64/3 + 0.36/4 + 0.64/3 v I

Solving these two equations, we get

UI = 0.06036 mm; VI = -0.2012 mm

Downloaded From : www.EasyEngineering.net

124 FINITE ELEMENT ANALYSIS

Stress in element 3,

03 =EE3 =E[Ih]{q'}A-C =E[ ~l ~}q'}A-C =(~)[-l -m I m]{q}A_C

0.06036

= 200 x 10
750
3
r- 0.6 _ 0.8 0.6 0.8] - 0.2012
0.0
0.0
2

ww Example 4.8
= 33.264 N/mm

w A concentrated load P = 50 kN is applied at the center of a fixed beam of length

.Ea
3m, depth 200 mm and width 120 mm. Calculate the deflection and slope at the
mid point. Assume E = 2 x 105 N/mm 2 •

syE
Solution
ngi
The finite element model consists of 2 beam. elements, as shown here, with

nee
nodes 1 and 3 at the two fixed supports and node 2 at the location where load P
is applied.
---2
ill rn rin 3

Stiffness matrices of elements I and 2 (connected by nodes 1 and 2

respectively, each with L = 1500 mm) are given by, g.n
~ 2 and 3

[K]- E I z
-e
[~~ 4~: ~~~ 2~;l- 2xl0 x 62_0-ilQQ3)[ ~~ 4~; ~~~
-12 -6L 12 -6L -
5

L3 -12 -6L 12
et
oL 2U -6L 4U 6L 2U -6L
Assembling the element stiffness matrices, we get
PI 12 6L -12 6L 0 0
4L2 -6L 2e 0 91
MI
-120x 2003 6L 0
P2
5
2 x lOx ------
12 -12 -6L 12+ 12 -6L+6L -12 6L W'I
w 2

M2 1500 3 6L 2L2 -6L+6L 4L2 +4e -6L 2e 92

P3 0 0 -12 -6L 12 -16L
W3 J
M3 0 0 6L 2L2 -6L 4L2 93

Downloaded From : www.EasyEngineering.net

CHAPTER 4 DISCRETE (I-D) ELEMENTS 125

After applying boundary conditions VI = V3 = 0 alJd (9z)1 -= (9z)3 = 0, the

equations reduce to
5(120 X 200 3 )
P2
{ M2
}= 2xl0 x
3
12 [12+12
1500 -6L+6L

The applied loads are P2 =,- 50000 Nand M2 = 0

- 50000 X 1500 3
1
ww Therefore, V2 =[ (
2 x 10 5 x _120 X 200
12
3 )
- x 24
=- 0.4395 mm

w.E -P L3 P(2 L)3

a
Check: From strength of materials approach, v 3

syE
=- -
=-
24 EI
or

0.4395 mm
--'-~'-
192 EI

ngi
and the deflection being symmetric, slope at the center (9 zh = o.

nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
E ngi
"This page is Intentionally Left Blank"
nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 5
CONTINUUM (2-D & 3-D)
ww ELEMENTS

w.E
5.1.
a syE
2-D ELEMENTS SUBJECTED TO IN-PLANE LOADS

ngi
When one of the cross sectional dimensions, width is significant compared to
the length of the member while the thickness is very small, it is considered as a

nee
2-D element. Displacement variation is therefore neglected across the thickness.
Let us consider the element in the X-Y plane while dimension in the Z-direction

rin
represents the thickness of the element. The load is assumed to be acting in the
plane of the element, along X-direction and/or Y-direction. Such a plane
element has two degrees of freedom per node, displacements along X and
Y directions.
g .ne
t
Y,v

1
/ II 11 1
p+t-----:..!..-- j/-+p P 1
1
p
1 ______________ _
1-
1-D element x,u
2-D element

If a concentrated load is applied at a point on the width of the plate, load can
not be considered as uniformly distributed over the width and hence the
displacement 'u' at any point is a function of its x and y coordinates. Load
along X-direction produces lateral strain and, hence, a displacement 'v' in the y-
direction (because of Poisson's effect). Thus, displacements u and v are
functions ofx and y coordinates of the point.

Downloaded From : www.EasyEngineering.net

128 FINITE ELEMENT ANALYSIS

In the case of discrete structures, with each member treated as a 1-0

element, nodes a~ chosen at junctions of two discrete members, junctions of
two different materials, at points of change of cross section or at points of load
application. However, in the case of continuum, which is modeled by 2-D or
3-D elements, there is no unique finite element model for analysis. Each
engineer may use a particular number of nodes and a particular orientation
of elements. Hence, the results obtained by different engineers may vary.
Mandatory safety codes for the design of pressure vessels are not therefore
based on the results of FEM. The results obtained by FEM have to be suitably

ww modified for compliance with mandatory safety codes. Varying number or type
of elements, but at a higher computational cost, may improve accuracy.
A judicial compromise has to be made between better accuracy of results and

w.E
computational cost. This aspect is further discussed under 'modelling techniques'

5.2

•
a
SIMPLEX, COMPLEX AND MULTIPLEX ELEMENTS

syE
Finite elements are classified into three categories.
Simplex elements are those obtained by joining n + 1 nodes in

ngi
n-dimensional space. Displacement functions of such elements consist of
only constant terms and linear terms, if nodal DOFs include only
translational modes.
Ex:
nee
2-noded truss (I-D) element, displacement represented by u = a, + a2x

rin
3-noded triangular (2-D) element, displacement represented by
u = a, + a2 x + a3 Y

•
u = a, + a2x + a3Y + a4z
g
4-noded tetrahedron (3-D) element, displacement represented by

.ne
Complex elements are those elements whose displacement function consists
of quadratic or higher order terms. Such elements naturally need additional
boundary nodes and, sometimes, internal nodes.
Ex: Quadratic models like 6-noded triangular element and
t
10-noded tetrahedron element
Cubic models like 10-noded !riangular element and
20-noded tetrahedron element
• Multiplex elements are those elements whose boundaries are parallel to the
coordinate axes and whose displacement function consists of higher order
terms.
Ex: 4-noded rectangle (2-D) element
8-noded hexahedron (3-D brick) element

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 129

y y

3
4 1---------. 4~---_i__r

ww ~--------~--_.x
2
~-----~----_+x

w.E
Calculation of stiffness matrix for a triangular element is first considered
since triangular elements are the simplest and can be used to define arbitrary
boundaries of a component more conveniently, by approximating curved

asy
boundary with a large number of elements having straight edges.

5.3
E
STIFFNESS MATRIX OF A CST ELEMENT

ngi
Let u(x, y) = a] + a2.x + a3.y and v(x,y) = a4 + as.x + ~.y be the displacements in

nee
the element. Displacement function f(x,y), representing u or v, can be
graphically represented by the following figure. In general, I-I', 2-2' and 3-3'
are not equal.

Hx,y) rin
g
3'

.ne
y
t
x

FIG U R E 5.1 Graphical representation of displacement function on a triangle

Substituting nodal coordinates, while the element 1-2-3 displaces to 1'-2'-3'

on application of load, we get nodal displacement vector as

Downloaded From : www.EasyEngineering.net

130 FINITE ELEMENT ANALY"I"

XI YI 0 0 0

r
rUI
U2 X2 Y2 0 0 0 a2
u) 1 x) Y) 0 0 0 a)
=
VI
V2
0
0
0
0
0
0
XI
X2
Yl
Y2
la, a5
v) 0 0 0 x) Y) a6

ww Y, v
3'

w.E
a syE
ngi
E : . - _ - - ! . ._ _ _ _ _ _ --+ X, u

{uJ = [G]{a} = [[A] [0]] {} nee

rin
[0] [A] a

1 [0 00 o~l
where, [A]= 1 xXI2
[1 x)
YI1
Y2

Y3
and [0]= 0
0 g
0
.ne
or { a}=[G]-I{u }=![Arl
e l[O]
[0]
[Arl
]{u} e t
{~} = [f(x,y)] {a} =[f(x,y)][G r {u l
e}

{:: }={: ~:
Yxy au/ay + fJv/fJx
} = [f'(x, y)][G ]-1

{E}=I~ ~ ~ ~ ~ ~l[G]-I{ue} = [8]{ue}

lo 0 0 1 .0

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 131

where, strain-displacement matrix,

0 Y3I 0 Y\2
[8]=_1_ [YD
0 x32 0 xI3 0 J
x:, where [J] is Jacobian
OetJ
x32 Y23 xI3 Y3I x2I Y\2

XI YI 0 XI -X3 YI -Y3 0 xJ3 YI3

and Oet J = 1 x 2 Y2 =0 x 2 -X3 Y2 -Y3 0 X23 Y23

ww 1 X3 Y3

= xJ3 Y23 - YI3 X23

X3 Y3 X3 Y3

w.E
It can be seen that area of the triangle, A = ~ Oet J = (~)
o xJ3
0 X23 Y23
YJ3

asy 1 X3 Y3
If nodes are numbered counter-clockwise, in right-handed coordinate system
Oet J is +ve.

E ngi
Stiffness matrix of the 3-node triangular element can now be obtained from
[KJv = f[8]T [O][8]dV = t f~B]T [O][B]dx dy
where t =
n
fdz is the thickness of the element and

[8] is a function of X and Y only. eer

Here, matrix' [0] corresponding to the particular deform::aion pattern of the
component (plane stress, plane strain or axi-symmetric) is to be used. An ing
explicit evaluation of the stiffness matrix is not generally feasible (except for a
few special cases). .ne
This triangular element, with 3 nodes and 2 OOF per node chooses linear
displacement functions for u and v and hence gives constant strain terms over
the entire element as seen from [f'(x, y)] or [B] and hence is popularly known
t
as 'Constant Strain Triangle (CST), element.

5.3.1 Stiffness Matrix ?fRight Angled Triangle

For the particular case of a right angled triangle with coordinates 1(0,0), 2(a, b)
and 3(0, b), let u = al + a2 x + a3 y and v = a4 + as x + a6 Y represent the
displacement field. Substituting nodal coordinates, vector of nodal
displacements can be written as

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
al
a2
E
Evaluating coefficients al to a6 in terms of nodal displacements,

0
0
l/a -lla ngi
0 0
0
0
0
0
0
ul
u2
a3
a4
-lib
0
0
0
1I b
0
0
I nee 0
0
0
0
U3

as 0
0
0
0
0
0 -lib
0 1/ a
0
-i/a
lib rin v2

g
a6 V3

or (a) = (Gr'(u,) =(a)=[[A] ,

[0] . [Ar l
[0] Jru )
e .ne
{.,&
Strain, {E} = au / Oy
}=riO 0 0 0 0 0 0 0 {a} = [B] {ue}
t
au/Oy+8v/8x 0 0 I 0 1 !J
=(.~ )r-~a -~J
b -b 0 0

where, [B] 0 0 -a 0
0 a 0 b
a b

[K] = f[B]T [0] [B] dv = t f f[Bf [0] [B] dx dy = tA [B]T [0] [B]
V 0 0

since elements of matrices [B] and [0] are not functions ofx or y

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS U3

For a plane stress case, stress-strain relationship is given by

1 v
[0] = E2 V 1 Let 13 = (1 - V)
(I-V) [ 0 0 2

Then, with element OOFs arranged in the sequence of [UI U2 U3 VI V2 V3]T

2
l3a

ww EtA
0
-l3a
2 _ b2 2
b +l3a
2
Symmetric

w.E
[K]= 2 2
a b (I_v 2
) 0
-l3ab
-vab
0
vab
l3ab 0
a 2

asy
l3ab vab
If the element OOFs are arranged in the sequence Of[UI
- ab(v + 13) _a 2
VI U2 V2 U3 V3]T,

l3a
0
E
the elements of stiffness matrix are rearranged as
2

a2 ngi
[K]= 2
EtA
2
a b (l-v 2
0 -vab
0
b2
0 13 b 2 nee
Symmetric

rin
) -l3ab
-l3a 2 vab _ b2 l3ab
2
b +l3a
2

_a 2

g
2 2 2
l3ab vab -l3b - ab(v + 13) a +l3b

5.4 CONVERGENCE CONDITIONS .ne

(To BE SATISFIED By THE DISPLACEMENT FUNCTION)

While choosing the function to represent u and v displacements at any point in

the element, care should be taken to ensure that the fol1owing conditions are satisfied.
t
(i) The function should be continuous and differentiable (to obtain strains)
within the element. This is automatically satisfied with polynomial
functions.
(ii) The displacement polynomial should include constant term,
representing rigid body displacement, which any unrestrained portion of a
component should experience when subjected to external loads.

Downloaded From : www.EasyEngineering.net

134 FINITE' ELEMENT ANALYSIS

(iii) The polynomial should include linear temls, which on differentiation

give constant strain terms, Constant strain is the logical condition as the
element size reduces to a point in the limit.
(iv) Compatibility of displacement and its derivative~. up to the required
order, must be satisfied across inter-element boundaries. Otherwise the
displacement solution may result in separated or overlapped inter-
element boundaries when the displacement patterns of deformed

ww elements with a common boundary are plotted separately (explained in

more detail in section 5.7).

w.E
(v) The polynomial shall satisfy geometric isotropy (terms symmetric in
terms of coordinate axes x, y and z), Otherwise, different users

asy analysing the same component may get different results by following
different node number sequence to define the elements (different local

E
coordinate systems).

ngi
Terms used in the polynomial, satisfying all the above conditions, are
represented by Pascal triangle given below, for a 2-D element.

n eer Constant term

ing
Linear terms

.ne
Quadratic terms

Cubic telms
t
I Quartic terms
I
I
I
I
!.--
I
Axis of symmetry
I
I

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 135

Similarly, the polynomial for a 3-D element is represented by the terms of

Pascal tetrahedron, as given below.

Constant term

Lmearterms

ww Quadratic terms

w.E , ------4---
2
Y

asy
'" X

2 "
X Z ' ..
2 "

E
XZ "

ngi
This can also be represented by the nodes of a hypercube, as given below.

n
Here, terms with other combinations are on the invisible s!des of the cube.

eer
ing
.ne
"
t
,
i-

Downloaded From : www.EasyEngineering.net

136 FINITE ELEMENT ANALYSIS

5.5 GEOMETRIC ISOTROPY

If all the terms of a particular order are included in the polynomial, it is called
the complete set. If the terms are symmetric w.r.t. x and y, it is called geometric
isotropy.
Based on the terms included in the polynomial, the function may be termed
as
complete and isotropic

ww complete and non-isotropic

w.E or
incomplete and isotropic
incomplete and non-isotropic

asy
As the number of terms in the polynomial depend on the number of OOF
and the--RUmber of nodes, use of complete set of terms of a particular order may
not be possible in all cases. But, isotropy can be maintained in all the cases and

E
is preferable so that user of a general-purpose program can start with any

ngi
particular node of his choice for defining the nodal sequences (which decide the
local coordinate systems) of different elements of the structure.

nee
For each element, local coordinate system is usually defined with node 1 as
the origin; X-axis along 1-2 and Y-axis perpendicular to X-axis in the plane of
nodes 1-2-3.

rin
The displacement function u(x) = al + a2.x + a3.y of a triangular element is
complete and isotropic while u(x) = al + a2.x + a3.y + a4.xy of a quadrilateral
element is incomplete but isotropic.
Higher order elements are broadly classified as -
g .ne
• Serendipity elements - TheBe are the elements having no internal nodes
Ex: 8-noded quadrilateral, 12-noded quadrilateral, etc. t
~.

~. \
\
.--.\
.--
\ •
\
.-.--.-- •

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 137

• Lagrange elements - These are the elements having internal nodes which
can be condensed out at the element level before assembling.
Ex : 9-noded quadrilateral, 16-noded quadrilateral, etc.

~. \
w \. \ \ . • •

w\.--
w.E .---. .-. --.-- •
\
•
•

asy
Polynomials used for some 2-D elements (subjected to in-plane loads),
satisfYing the convergence and isotropy conditions, are given below.

Element
En Order of
Displacement
No.of
nodes
Terms included
Polynomial
type
Triangle
(Fig.S.2 a) ginLinear 3 al + a2·X + a3·Y Complete &
Isotropic
Triangle
(Fig,S.2 b)
eer
Quadratic 6 al + a2 x + a3 y + a..X2 + a5 xy + a" y2 Complete
& Isotropic

ing
Triangle Cubic 9 al + a2.X + a3.y + a..X2 + a5.xy + a".y2 + Incomplete,
(Fig.S.2 c) a7.x2y + a..xi + a9 x2y2 Isotropic

Triangle'
(Fig.S.2 d)
Cubic
. IO
al + a2X + a,y + a.x 2 + a5xy + a"i
+ a7xly + a.xy~ + ~X3 (Notj>referred)
al + a2.X + a3.y + a..x 2+ a5.xy+ a".i +
a7.x3 + a..x2y + ~ xi + alO.y3
Incomplete
Non-Isotropic
Complete,
Isotropic
net
Quadrilateral Linear 4 al + a2.X + a, y + a..xy Incomplete,
(Fig.S.2 e) IsotropIc

al + a2X + a3Y + a.x 2 Incomplt:te,

(Not preferred) Non-lsotroJllc
Quadrilateral Quadratic 8 al + a2.X + a3.y + a..x 2+ a5. xy + a,,·i + Incomplete,
(Fig's.2 f) a7.x2y+ a•. xi Isotropic
Quadrilateral Cubic 12 al + a2.X + a3.y + a..x 2+ a5.xy+ a".y2 + Incomplete,
(Fig.S.2 g) a7.x3+ a..x2y+ ~.xy2+ alO.y3+ all.x 3y Isotropic
+ al2. xy3

Downloaded From : www.EasyEngineering.net

138 FINITE ELEMENT ANALYSIS

. .\ .

/\..1-.-----. .--'--.--
.~
/ .\
.
(b)
/
/
/\
\

(e)
\ .
(d)

ww ~. .------- .\.

w\~\
.
.I ~
.-------
.~
\

------
\
\. I
.Ea .---- .----.
.
. \
\.
.--.---
\ .-\

syE (e) (I)

FIGUR E 5.2 Some 2-D elements for in-plane loads

(g)

5.6
ngi
ASPECT RATIO

ne
In 2-D and 3-D elements, the displacement function is symmetric in x, y and z,
whether it is complete or not in terms of coefficients of a particular order as
given by Pascal triangle or Pascal tetrahedron. Hence, the shape of the finite
element in the idealised structure should also be oriented equally to all the
eri
relevant axes. For this purpose, certain conditions are generally specified in the
standard packages on the sizes and included angles for various elements.
Aspect ratio is defined for this purpose as the ratio of the longest side to the n g.n
A
Angle at A < 45(\
et
b , - - - I_

a
_- - - - - - '

alb> 5
bD a alb<5

A~ Angle at A < 60" C\

Shapes of elements, not preferred Preferred shapes of elements

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 1.39

shortest side. It is usually limited to 5, while the included angle is usually

limited to 45° to 135° for a triangular element and to 60° to 120° for a
quadrilateral or 3-D element. A few 2-D elements with valid and invalid shapes
are shown here.

5.7 INTER-ELEMENT COMPATIBILITY

The polynomial used to represent variation of displacement over the element

ww
should ensure compatibility of displacement along inter-element boundary.
If this condition is not satisfied, inter-element boundary of two adjacent

w.E
elements may overlap or show void on application of external loads, when
the displacement pattern of different elements with a common boundary are

asy
En
3
gi nee
rin
(al Unloaded elements (b) Loaded compatible clements

4'
g.n
e t
(e) Loaded meompatible overlappmg (d) Loaded incompatIble elements with void
elements

plotted separately. The inter-element compatibility condition is satisfied when

displacement at any point along a common edge, of all elements joining along
that edge, is a function of displacements of nodes on that edge only. This
concept is demonstrated here for displacement along x axis of a right angled

Downloaded From : www.EasyEngineering.net

140 FINITE ELEMENT ANALYSIS

triangular element, with the two sides of lengths a and b coinciding with the
coordinate axes.

u = [1 x
{
l
y] a 2 }
a3
:.
I
{Uu } =
2

u3
[I
l O b a3
l
I a0 00j{aa 2 } by substituting nodal coordinates

I~Jt:}
0

ww or {::H-Il/a
a3 -lib
l/a
0

w.E [_III a ~ Jfl[I_X/a-Y/b xla

Then,
0

asy
u =[1 x Y]
-lib
Y/b]tJ
lIa
0 lib u3

Enx,la t} a]{
At R(xt,O),
ginu = [1- XI / a 0] [1-x,/a xI / Uj }
U2

eer
or u at R(xt, 0) is a function oful and U2, displacements of the two end nodes of

ing
that edge only.

Similarly, at S(Xt,YI), ~=(a~xI)or 1-(:I)-(~)=O

Then, u~[O
.
x,la Y,/b]tJ ~ [x,/a y,/b]{::}
net

or u at S(xt, YI) is a function of U2 and U3, displacements of the two end

nodes of that edge only.

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 141

Similarly,

at T(O, Yl), u=[I-y\/b 0 y\/b1t:} = [1-y\/b Y\/bl{~:}

or u at T(O, Yl) is a function oful and U3, displacements of the two end
nodes of that edge only.

ww
The same logic holds good for v, displacement at any point of the element in
y direction.
To adequately represent stress concentration in some local regions, it is a

w.E
common practice to either increase number of elements or increase the order of
the polynomial of the displacement function. The first method ensures inter-

asy
element displacement compatibility but at a higher computational cost. The
second method may not always ensure inter-element displacement

E
compatibility. Transition elements are commonly used in such situations. These

ngi
are covered in more detail in section 7.10 of this book.

n
5.8 2-D ELEMENTS SUBJECTED TO BENDING LOADS
Plate bending element : It is a plate element in X- Y plane subjected to
bending load Pz and/or bending moments Mx , My. A thin plate (span> lOx eer
thickness) with small deflection « thickness/lO) follows Kirchhoffs theory and
is an extension of I-D beam element into two dimensions. It will have three ing
degrees of freedom at each node, displacement normal to the plate (w) and
rotations about the two major axes of the element represented by derivatives of .ne
w about x and y (ex and ey ).
For a triangular Plate bending element, normal deflection is assumed by the
polynomial.
t
w = al + a2.x + a3.y + at.x2+ a5.l + a6.x2y + a7.xy2 + as.x 3+ a9.l

Ox
Ow
= - = a 2 + 2a 4 ·x + 2a 6·xy + a 7·y
2
+ 3a s ·x
2
ax

Displacement function of triangular plate bending element is incomplete but

isotropic.

Downloaded From : www.EasyEngineering.net

1.42 FINITE ELEM I.N'( A N AI. YSIS

z
z

ww Plate bending element Thin shell element

w
Thin plate with large deflection is characterised by large tensile or
compressive stresses in the middle plane. The corresponding u and v

.Ea
displacements are given by

= -z ( ~ 1
u

s
= -z ( : ) ;

". ~(:l~-{y~~E
1
v

n
", ~(: l~-{~~1 gi n
~(: H~)~-2Z(:; 1 ee
1.,
rin
The stress-strain matrix is given by {cr} = [D) {E}
g.n
et
This is more commonly expressed in terms of moments per unit width (b= I),
also called stress resultants, using

MX } [1 v
3
Thus, My = - Eh v 1 2
{ 12(1- v ) 0
Mxy 0

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) EI.EMENTS 143

Thin shell element: It is a 2-D element subjected to in-plane loads as well as

bending loads. As in the case of a general beam, these two behaviours represent
uncoupled degrees of freedom. It can therefore be considered as a combination
of plane stress element and plate bending element. In the local coordinate
system, this element will have five degrees of freedom since moment about
normal to the plate is not included. However, if different elements are inclined
to each other, transformation of the combined stiffuess matrix of each element
with five degrees of freedom per node in local coordinate system results in six
degrees of freedom per node in the global coordinate system.

ww
5.9 3-D ELEMENTS

w.E
Similar to the elements described above, 3-D elements are of two types based
on the type of function used - solid element with three translational
displacement degrees of freedom per node without significant bending

asy
behaviour and thick shell element with six degrees of freedom. The
displacement functions normally used are given below.

E
For a 4-noded 3-D solid tetrahedron element
u = al + a2.X + aJ.y + a4·z
v = a5 + a6.x + a7.y + ag.z ngi
w = a9+ alO.x + all.y + al2.z
These functions are complete and isotropic.
nee
4
rin
8
.....,---_ _ _ _...,.,7
g .ne
3
51'-----t----r
t
2
For a 8-noded 3-D solid hexahedron element
u = al + a2.x + aJ.y + ~.z + a5.x2 + ~.l + a7.z2 + ag.xyz
v = ~ + alO.x + all.y + al2.z + au.x2 + al4.l + al5.i + al6.xyz
w = al7 + alg.X + a19.y + a20.z + a21.x2 + a22.l + a23. z2 + a24.xyz

Downloaded From : www.EasyEngineering.net

144 FINITE ELEMENT ANALYSIS

The function used is therefore incomplete but isotropic.

Isotropy condition in these elements involves symmetry w.r.t. X, Y and Z

axes. The displacement field can also be chosen by a different polynomial
u = a, +a2.x+a3.Y+84.z+as.x2 +<If;.l +a7.z2 +ag.xyz ; ...
Thus, there is no unique way of choosing an incomplete polynomial

ww 5.10 AxI-SYMMETRIC ELEMENTS

These are special cases of 3-D components where 2-D analysis can be carried

w.E out for evaluating displacements and stresses, saving lot of time and effort.
There are many components such as turbine casings, compressor casings,

asy
pressure vessels, cylindrical heat exchangers etc., which are 3-D components by
the relative dimensions of the component in the three coordinate directions.

E
However, each is symmetric about its axis of rotation and thus deflection and

ngi
stress along any 2-D radial plane (imaginary section planes A, B,.. in the
figure), will be identical. It is often more convenient to represent such

nee
components in cylindrical coordinate system, consisting of axial (usually
represented by z-axis), radial (or r-axis) and hoop (or circumferential or 8)

rin
direction. A section through r-z plane is considered for analysis. Unlike in the
case of plane stress or plane strain analysis of 2-D components, it need not be

g
constrained in the radial direction at any point in the component to suppress

.ne
rigid body motion. This is automatically taken care of by the closed geometry in
the hoop direction, thus providing a natural boundary condition.

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 145

ww
w .Ea
I
/ •••••••••• 1 •••••••••••\

syE
ngi
nee
rin
g.n
Also, displacement in the radial direction at any point in the component by
'dr' gives rise to a corresponding change in circumferential length by
et
21t(dr) dr
21t(r + dr) - 21£ r or 21£(dr) . It amounts to a hoop strain of - - - or - .
21£r r
Thus the stress-strain relation is a 4 x 4 matrix relating o' r , 0'0' o' z and 'trtJ
with Er , Eo, Ez and "f rtJ given by

O' r I-v v v 0 Er
<To E v I-v v 0 Ee
v v I-v 0 or {O'}=[O]{E}
O' z -(I+v)(I-2v) Ez
1-2v
0 0 0 --
'tro 2 "fro

Downloaded From : www.EasyEngineering.net

146 FINITE ELEMENT ANALYSIS

Sometimes, components like shell nozzle junction are analysed as

axisymmetric solids, to save time and effort, even though it is not symmetric
about nozzle axis or shell axis. For this purpose, cylindrical shell is replaced hy
a spherical shell of double the radius, because the larger stress component (hoop
stress) in a sphere is half the corresponding stress of a cylinder of the same
radius. The same is nottrue for thermal stresses and needs some moditication.

5.11 SUMMARY

ww • While 1-0 elements generally form a discrete structure, 2-0 and

3-D elements form part of a continuum. In the 1-0 element, where

w.E axial dimension is very large compared to the cross section, load is
assumed to act uniformly over the entire cross section. When one of

asy
the cross sectional dimensions, width, is significant compared to the
length of the member, it is considered as a 2-D element.

E
• Finite elements are classified into three categories - Simplex

ngi
elements obtained by joining n+ 1 nodes in n-dimensional space;
Complex elements whose displacement function consists of
quadratic or higher order terms; Multiplex elements whose

• nee
boundaries are parallel to the coordinate axes.
The function chosen to represent u and v displacements at any point

rin
in the 2-D element should satisfy the following conditions - The
function should be continuous and differentiable within the

g
element; should include constant term and linear terms; should

.ne
satisfy compatibility of displacement and its derivatives, across
inter-element boundaries and should satisfy geometric isotropy.
• A model with less number of higher order elements (with more than
t
two nodes along edges of the element) will give better results than
more number oflower order elements and is economical in terms of
computer memory and time.
• 2-D Plane stress element and 3-D solid element are similar to 1-0
truss element, whose degrees of freedom do not include slopes
while 2-D Plate bending element and 3-D thick shell element are
similar to 1-0 beam element, whose degrees of freedom include
slopes.
• Plane stress element applies to a thin plate with in-plane loads,
having zero stress in the normal direction, while plane strain
element is a thin slice of a large 3-D solid, where load acts in the

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 147

plane of slices and analysis of one slice represents solution of the

3-D solid, and has zero strain in the normal direction.
• Axisymmetric element is also a 2-D model of a 3-D solid with
axisymmetric geometry, loads and boundary conditions. It differs
from plane strain element, in having finite strain normal to the
plane of analysis as a function of radial displacement.
• Thin shell element is a superposition of plane stress element and
plate bending element, representing uncoupled degrees of freedom.

ww
OBJECTIVE QUESTIONS

w
I. Complete polynomial is _

.Ea
important, compared to symmetry of
displacement polynomial w.r.t. coordinate directions
(a) equally (b) more (c) less (d) unrelated

syE
2. A triangular element with cubic displacement function requires _ _ nodes
to represent the complete and symmetric polynomial
(a) 3 (b) 6

ngi
(c) 9 (d) 10
3. A triangular element with quadratic displacement function requires _ _
nodes to represent the complete polynomial
(a) 3 (b) 6 (c) 9
nee (d) 10

rin
4. A triangular 9-noded element will usually have _ _ cubic displacement
function
(a) symmetric & complete
(c) unsymmetric & complete
(b) symmetric & incomplete
(d) un symmetric & incomplete g.n
5. A constant term in the displacement function ensures
(a) rigid body mode
(c) zero stress
(b) constant strain mode
(d) zero deformation
et
6. Number of terms in the displacement function in relation to the number of
nodes in that element is
(a) more (b) equal (c) less (d) unrelated
7. A linear term in the displacement function ensures
(a) rigid body mode (b) constant strain mode
(c) strain varying in the element (d) stress varying in the element
8. All stiffness coefficients of a plate bending element have _ units
(a) same (b) different (c) any set of (d) depend on other data

Downloaded From : www.EasyEngineering.net

FINITE ELEMENT ANALYSIS

9. All stiffness coefficients of an axisymmetric element have _ units

(a) same (b) different (c) any set of (d) depend on other data
10. Displacement method can NOT be used with _ _ boundary conditions
(a) pressure (b) temperature(c) stress (d) displacement
11. A triangular plane stress element has _ D.O.F
(a) 6 (b) 9 (c) 12 (d) 15

ww 12. A thin shell element has _

element
no. of DOF, compared to a plate bending

w.E (a) same (b) more (c) less (d) unrelated

13. A plane stress element has _ no. of DOF, compared to a plate bending
element

a (a) same

syE(b) more (c) less (d) unrelated

14. An axisymmetric element has _ no. of DOF, compared to a plate bending
element
(a) same (b) more
ngi (c) less (d) unrelated

nee
15. A structural thin shell triangular element has _ DOF
(a) 3 (b) 6 (c) 9 (d) 18
16. A triangular plane strain element has _
(a) 3 (b) 6 (c) 9
DOF

rin
(d) 15

g
17. Number of displacement polynomials used for an element depends on
(a) No.ofnodes/element (b) No. of DOF/node
.ne
(c) No. of DqF/element (d) type of element
18. For a plate bending element, number of displacement polynomials and
number of D.O. Finode are
t
(a) 1,2 (b) 1,3 (c) 2,3 (d) 2,4
19. Accuracy of solution' in a 2-D component depends on
(a) included angle of elements (b) size of the component
(c) no. of DOFI node (d) type of load
20. Displacement of any point on a side is related to displacements of nodes on
that side only, ensures
(a) equilibrium (b) compatibility
(c) energy balance (d) continuity along inter-element boundary

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS jA9

21. Continuum analysis covers

(a) all 2-D trusses & frames (b) all 3-D trusses & frames
(c) all 2-D and 3-D plates, solids (d) only 3-D solids
22. Optimum number of elements in finite element model depends on
assessment of_ distribution in the component
(a) displacement (b) stress (c) strain (d) potential energy
23. Displacement function which matches function value at the specified nodes

ww is classified as
(a) Lagrange interpolation function (b) Serendipity function

w.E
(c) Hermite interpolation function (d) Pascal function
24. Displacement function which matches function value as well as its

asy
derivatives (slopes) at the specified nodes is classified as
(a) Lagrange interpolation function (b) Serendipity function

E
(c) Hermite interpolation function
25. Continuum analysis includes
(a) trusses (b) beams
(d) Pascal function

ngi
(c) plates (d) plates & solids

nee
26. Continuum elements and discrete members can be included in a single
model for analysis
(a) always true (b) sometimes true
rin
(c) never true (d) depends on matching DOF
27. Continuum elements in different analysis may vary in
(a) size (b) shape (c) size or shape
g
(d) size & shape .ne
28. Element formed with edges parallel to coordinate axes is called
(a) simplex element (b) complex element
t
(c) multiplex element (d) compound element
29. An element with no internal nodes is classified as
(a) serendipity element (b) Lagrange element
(c) Hermite element (d) Laplace element
30. An element with internal nodes is classified as
(a) serendipity element (b) Lagrange element
(c) Hermite element (d) Laplace element

Downloaded From : www.EasyEngineering.net

:1.50 FINITE ELEMENT ANAL YSIS

31. A concrete pedestal is represented by

(a) plane stress elements (b) plane strain elements
I

(c) 3-D solid elements (d) 3-D shell elements

32. Combination of plane stress element behaviour and plate bending behaviour
forms
(a) 3-D solid element (b) 3-D shell element

ww (c) Thin shell element

33. A 3-D dam is usually modeled with
(d) thick shell element

w.E (a) 2-D plane stress elements

asy
34. Element formed by joining n+ 1 nodes in n-dimensional space is cal!ed

E
(a) simplex element
(c) multiplex element
ngi
(b) complex element
(d) compound element
35. Element fonned with quadratic or higher order displacement polynomial
is a
(a) simplex element n eer
(b) complex element
(c) multiplex element (d) compound element
ing
36. Elements connecting lower order elements and higher order elements in a
mesh are called
(a) transition elements (b) sub-parametric elements .ne
(c) iso-parametric elements (d) super-parametric elements
37. Elements having mid-side nodes only on some sides are called
t
(a) transition elements (b) sub-parametric elements
(c) iso-parametric elements (d) super-parametric elements
38. Stress-strain matrix for plane stress element, if strain is represented by SiJ
and stress is represented by stu' is obtained from the condition
(a) Szz= 0 (b) Szx= 0 (c) stzx= 0 (d) slzz= 0
39. Stress-strain matrix for plane strain element, if strain is represented by SiJ' is
obtained from the condition
(a) Szz= 0 (d) slzz= 0

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS :l.5:l.

40. Stress-strain matrix for axisymmetric element is of order

(a) 3 x 3 (b) 4 x 4 (c) 6 x 6 (d) 9 x 9
41. Stress-strain matrix for plate bending element is of order
(a) 3 x 3 (b) 4 x 4 (c) 6 x 6 (d) 9 x 9
42. Elasticity matrix for _ behaviour is similar to 3-D elasticity matrix
(a) plane stress (b) plane strain

ww
(C) plate bending (d) axisymmetric
43. Plane stress element is an extension of
(a) truss element (b) beam element

w.E
(c) pipe element
44. Plate bending element is an extension of
(d) spring element

a
(a) truss element
(c) pipe element
syE (b) beam element
(d) spring element

ngi
45. Wrong sequencing of nodal connectivity in 2-D & 3-D problems leads to
(a) +ve Jacobian (b) zero Jacobian
(c) -ve Jacobian
nee
(d) No relation with Jacobian
46. Axisymmetric structures are usually modeled in
(a) element local coordinates (b) global cartesian coordinates
(c) global cylindrical coordinates (d) user specified system rin
47. A plate of 1em thickness with in-plane loads is modeled by
(a). plane stress element (b) plane strain element g .ne
(c) plate bending element (d) anyone of them
48. Actual thickness of plane strain element is
(a) very small (b) very large
t
(c) any specified value (d) assumed by software
49. Order of stifthess matrix for a plane stress model with 20 nodes is
(a) 10 (b) 20 • (c) 40 (d) 60
50. Order of stifthess matrix for an axisymmetric model with 20 nodes is
(a) 10 (b) 20 (c) 40 (d) 60
51. Number of stress components per node calculated for a plane stress
quadrilateral element is
(a) 2 (b) 3 (c) 4 (d) 5

Downloaded From : www.EasyEngineering.net

152 FINITE ELEMENT ANALYSIS

52. Number of stress components per node calculated for a triangular

axisymmetric element is
(a) 2 (b) 3 (c) 4 (d) 5
53. A general plate element is a superposition of elements
(a) plane stress & plane strain (b) plane strain & plate bending
(c) plane stress & plate bending (d) plate bending only

ww 54. An element with in-plane loads having 3 nodes along each side is a
(a) constant strain element (b) linear strain element

w (c) quadratic strain element

.Ea
SOLVED PROBLEMS
(d) constant displacement method

------------------------------------------------
Example 5.1
syE
Calculate displacements and stress in a triangular plate, fixed along one edge

ngi
and subjected to concentrated load at its free end. Assume E = 70,000 MPa,
t = 10 mm and v = 0.3.

Solution

Y,v nee lOON

rin
2

T
20mrn
SON

g.n
1 et
If the element OOFs are arranged in the sequence of[ul VI U2 V2 U3 v3f,
the stiffuess matrix from 5.3.1 is
13 a 2
0 a2
EtA 0 -vab b2 Symmetric
[K]= 2 2 2 2
a b (1_v ) -l3ab 0 0 13 b
-l3a
2
vab _b 2 l3ab
2
b + l3a
2

l3ab _a 2 vab -13 b 2 - ab(v + 13) a 2 +l3b 2

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 153

Substituting the given dimensions and material properties,

315

K _ 70000 x lDXe O
; 20 1 0
0
900
-180 400 Symmetric
[ ]- 30 2 X 20 2 x (1- 0.3 2 ) -210 0 0 140
-315 180 -400 210 715

ww After applying boundary conditions,

210 -900
UI
180 -140
= VI = U3 = V3 =
-390 1040
0, these equations

w
reduce to

{PXJ}
P Y2 100 .Ea
- = {50} = 641 026 [400
- . 0
0] {u 2 }
140 V 2

Therefore, U2 =-
syE
0.000195 mm and V2 =-0.001114 mm

ngi
Note : The given thick plate, trom university question paper, should not be
analysed as a 2-D problem. It can not be solved as a 3-D problem manually.

Example 5.2
nee
rin
Compute the plane strain stiffness matrix in terms of the ratio r = alb for the
rectangular element of sides a and b, using v = 0.2, r = I and displacement
model
g.n
Solution

By substituting coordinates of the four comer nodes in the displacement model,

et
With origin (0,0) at node I,
ul 1 0 0 0 0 0 0 0 al
u2 a 0 0 0 0 0 0 a2
u, a b ab 0 0 0 0 a3
u4 0 b 0 0 0 0 0 a4 or {uJ= [[A]. [0]]
VI 0 0 0 0 0 0 0 a5 [0] [A] {a}
v2 0 0 0 0 a 0 0 a6
V3 0 0 0 0 a b ab a7
v4 0 0 0 0 0 b 0 as

Downloaded From : www.EasyEngineering.net

154 FINITE ELEMENT ANALYSIS

y, v

4 3

2
'------=---+x, u

For the particular case of r = ~ = 1 , evaluating coefficients al to ag in tenns

ww of nodal displacements,
b

w.E al
a2
a3
1
-lla
-lla
0
lIa
0
0
0
0
0
0
II a
0
0
0
0
0
0
0
0
0
0
0
0
ul
u2
u3
a4
a5
lIa
0
2

asy 0
2
-l/a l/a
0
2
_\/a
0
2
0 0
0
0
0
0
0
u4

a6
a7
0
0
E
0
0
0
0
ngi
0
0
-\/a
-l/a
\/a
0
0
0
0
\/a
VI

v2
V3

nee
2 2 2 2
ag 0 0 0 0 l/a -lIa \ I a -l/a v4

or a _[rAr' rO] ]
{ }- [0] [Ar {u l e}

rin
Strain, {E} = r& }
av lOy
8u/Oy+0y/8x
[0
0 y 0 0 0
= 0 0 0 0 0 0 1
0 0 1 x 0 I 0
g.n
~Jra} =[BHu,}
e
where, t
[B]=ct~:
-(y-a) y -y 0 0 0 O
0
-x
0 0 x-a -x x
x -(x-a) y-a -(y-a) y
-<x-y-a)1
For a plane strain case, stress-strain relationship is given by

E_-ll~V
v
[D]= _ _ I-v
(1 + v)(1- 2v) 0
o

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 2.55

l
8
-~ 2 82 0 01 for v = 0.2
7.2 0
o 3
a b

[K] = f[B]T [0] [B] dv = t f f[B]T [0] [B] dx dy

v 0 0

a b

ww = f f[B]T [0] [B] dx dy

o 0
since t/= 1 for plane strain elements

w.E
44
-26 44 Symmetric

-22 4 44
E
86.4
4
15 asy-22
3 -15
-4 44
-3 44
-3 -15
En
3 15 4 44
-15
3
-3
15
-3
15
3
gin
-22 -26 44
-15 -4 -22 4 44

eer
when written corresponding to the displacement vector [UI VI U2 V2 U3 V3 U4 V4t

Example 5.3
Compute the plane strain stiffness matrix of a square, treating this as an ing
assembly of two triangular elements with the displacement field in these
elements expressed as u = al + a2 x + a3 Y and v = ~ + a5 x + ~ y .
.ne
t
Assume v = 0.2

Solution
Let the square of side 'a' be represented by two triangular elements identified
by nodes 1, 2, 3 and 1, 3, 4 respectively. Choosing node 1 as the origin of the
X-Y coordinate system,

y,v
4..-----."..3

w:;..._ _ _ _~--+ X, u

Downloaded From : www.EasyEngineering.net

156 FINITE ELEMENT ANALYSIS

For element-I,

Then
{~+r: : m::J
l I
a2 } rIO
=-I/a I/a 0
0 J{U } U2

ww {
a3

{E}={:;:
1 -I/a I/a

}=r~
u3
000

w.E
Strain,

100
ou/8y+8v/& Lo
0 000
0 o 1

a~mr
sy~IE
where, [R] 000
-I 1 -I
o
-I
1
OJ
1
0

nrl~VgiI~V
For a plane strain case, stress-strain relationship is given by

~. J
n
[O]= _ _E_ _
(I+V)(1-2V)L 0 0
eer
(I-2v)/2

=~r~ ~ ~J = ing
.ne
7.2 for V 0.2
003

[K] ==
v
f[8f [0] [8] dv = t

a b
a

o
b

f f[8f [0] [8] dx dy

0 t
= f f[8]T [0] [8] dx dy since t = 1 for plane strain elements
o 0

8
Symmetric
0 4
E -2 4 6
=
14.4 2 -4 -6 12
0 -4 -4 4 4
-2 0 2 -8 0 8

Downloaded From : www.EasyEngineering.net

CHAPTER 5 CONTINUUM (2-D & 3-D) ELEMENTS 157

when written corresponding to the displacement vector [UI VI U2 V2 U) V)]T

For element-2,

ww
w.E
asy 3
0 8
En
Symmetric

[K]- E
0
14.4 -3
-- 3
-2
0
2
8
0
-8
gi
3
3 12 nee
3 -8 2 -3 -5 11
rin
when written corresponding to the displacement vector

The stiffness matrix of the square plate is obtained by adding relevant

coefficients of the stiffness matrices of elements 1 and 2, as
g
[UI VI U3 V) U4 V4]T

.ne
3+8
0+0 8+4
t
-2 4 6 Symmetric

2 -4 -6 12
[K]- E
14.4 0+0 -2-4 -4 4 8+4
-3-2 0+0 2 -8 0+0 3+8
-3 2 0 0 -8 3 12
3 -8 0 0 2 -3 -5 11

corresponding to the displacement vector [UI VI U2 V2 U) v) U4 V4]T

Downloaded From : www.EasyEngineering.net

158 FINITE ELEMENT ANALYSIS

Note:

1. The stiffness matrix obtained for the combination of two triangular

elements in Ex. 5.3 is different from the one obtained for the square
element in Ex. 5.2.
2. In the first case, node 2 and node 4 are not linked and hence terms k37,
k38, ~7' ~8, k73, k74' k83 and k84 are equal to zero.
3. Even other terms are different, indicating that the results obtainell/or a

ww given problem will depend on the model used for analysis and differ
marginally.

w .Ea
syE
ngi
nee
rin
g.n
et

Downloaded From : www.EasyEngineering.net

CHAPTER 6
HIGHER ORDER AND
ISOPARAMETRIC
ww ELEMENTS
w.E
asy
8.1
E
HIGHER ORDER ELEMENTS ngi
nee
When geometry is modeled with CST elements, large number of small-size
elements need to be used, in order to accommodate variation of strains over the

rin
entire geometry. In view of the constraints on computer memory and time for
solving large size problems, an alternative method of using a small number of

g
higher order (refined) elements can also be considered. In these elements, a
higher order polynomial is used to include variation of strain over the element
by choosing additional nodes.
.ne
In most cases, axially loaded spars or truss elements have uniform cross
section and hence stress/strain in the element is constant along the length of the
member. A few special cases may involve stress/strain varying along the length
of a truss element, such as a vertical column with distributed self weight. For
t
such applications, a 3-noded truss element is used. However, in the case of
beam elements, elementary beam equation predicts deflection of a beam varying
parabolically along the length. Basic beam element in FEM caters to this
variation. Higher order beam element 'is not of any practical use and is,
therefore, not discussed further.
(a) 3-noded truss element: This element has 3 nodes including one extm
node in the middle of the element. 2nd order displacement function is
used. Stiffness hlatrix for this element is derived here for an
understanding ofthe subject.

Downloaded From : www.EasyEngineering.net

160 FINITE ELEMENT ANAl Y'il'i

_____ __-+_ x
k P, U

Substituting the values of x for the two end points and middle point of the
°
truss element, (x = at node i, x = L at node j and x = L/2 at node k), nodal

ww displacement vector is written as

w.E {~:}=[:
Uk I Ll2
~ ~ ]{:~}
L2/4 a3 .
or {Ue}=[G] {a}

a
Solving for the coefficients a and substituting in the earlier equation,

syE
u(x) = [f(x)]T [G]I {u e }

n-I~Lgi 4~L 1{~:}

1
2
=[1 X x] [_3 /L =[N]T {Ue}
2 I L2 2 I L2 - 4 I L!

nee
Ul

Strain, E = du = [f '(x)] T [Gr 1 {ue} = [ 8] {ue }

[-;/L 4~L j rin

where, [8] = [0. 1 2x]

(~2 )
2/L2
_lOlL
2/L2
g -4/L2
.ne
and
=

{5E} = [8] {5u e}

[(-3L + 4x) (-L + 4x) (4L -8x)]

; {5E}T = {5Ue}T[8]1'
t
Stress, {a} = [D] {E} = E {E} = E [8] fUel

:. {Pc} = [K6'] fUel

where, [Ke]:::: J[8]T [D][8] dv:::: JJJ8]T E [8] dx dy dz
v

= AE J[8]T [B] dx
L

since [B] is independent ofy and z and JJdY dz =A

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 181

=
1 -S]
AE [ 71 7-S
3L -S -S 16

A higher order truss element, which accounts for variable stress/strain, has
very few applications. Some of them are shown here.

Example 6.1

ww
Calculate the displacement at the free end of a 50cm long tapered bar of area of
cross section 1000 mm 2 at its fixed end and 600 mm 2 at the free end, subjected
to an axial tensile load of 1kN at the free end. Assume E = 200GPa.

w.E I---+P

asy
~---L---+

Solution E
---+~X,U

ngi
nee
In order to explain the advantage of a higher order element, the problem is
solved first by using basic truss element and then by using higher order element.

rin
(a) The bar is identified by two 2-noded elements and mean area of cross
section is considered for both elements.
At mid':point of the bar (node 2),

A = (1000 + 600) = SOOmm2 g .ne

t
2
For element 1, connecting nodes 1 and 2,
- (l 000 + SOO) -- 900 mm 2-- 9 cm2
A 1-
2
Stiffness matrix for element-l is,

K= AIE[
[]I
1 -l]=(E)[
L - 1 1
9 -9]
L - 9 9

Downloaded From : www.EasyEngineering.net

162 FINITE ELEMENT ANALYSIS

For element 2, connecting nodes 2 and 3,

AJ -- (800 + 600) -- 700 mm 2-- 7 cm 2

- 2

Stiffness matrix for element-I is,

A2E [ I -I] E [ 7 -77]

.[Kh =T -I I = L -7

ww Assembled stiffness matrix is then obtained as

w.Er [K]=~
9
-9
o
-9
9+7
-7
0
-7
7
1
a syE
Applying boundary condition UI = 0, reduced stiffness matrix is obtained as

ngi
Solving these two simultaneous equations, we get

{~:}= Dlet[~ I:]{IO~O}

nee
rin
where,
E
L
63E
Det = - (7 x 16 - 7 x 7) = -
L g .ne
", ~ (¥) ~II{~) t
and ", ~ 1(;i~O ~25{~)
Corresponding stresses are
E(u - u ) 2
0"1 = 2 1 = III - 0 = III N/cm constant along eJement-I
L

and 0"2 = E(u 3 - u2) = 254 - 111 = 143 N/cm2 constant along element-2
L

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 1.63

(b)The bar is modeled by one 3-noded element with uniform area of cross
section (mean of max and min areas). Thus, with A2 = 800 mm 2, element
stiffness matrix is derived as

[K]=( ~~E) [ 7
1 7
1 -8j{U
-8 u
1

3
}

-8 -8 16 u2

ww 1[:------, --::":-J----P
---------
w.E
Since the chosen model has only one element, assembled stiffness matrix

asy
is also the same. Therefore,

E ngi
n
Note: Care should be taken while numbering the load and displacement

eer
components, since the 3 rows of the stiffness matrix correspond to the
end nodes at x = 0, x = L and the mid point (at x = Ll2) respectively.
Applying the boundary condition, UI = 0, the above equations reduce to
ing
{P3}
P2
= {1000} = A2E [ 7
0 3L-8
.ne
Solving these equations, we get

{:}
6
~t ['8 ~]{'~O}
t
where, Det= A2E (16 x 7-8 x 8)= 128E
3L L

Therefore,
", ~ (~~ ~62.5(~l

Downloaded From : www.EasyEngineering.net

1.64 FINITE ELEMENT ANALYSIS

and
U, =l(~) =125(~1
Stress in the bar is given by

cr = E [B] {ue } =E (~2 ) [(-3L + 4x) (-L + 4x) (4L - 8x)]

uu: }
{u

ww = E (~2 ) [(-3L + 4x) UI + (-L + 4x) U3 + (4L - 8x) U2]

w.E =E (~2) [(-3L+4x).O+(-L+4x).{12S LIE) + (4L-8x).(62.S LIE)]

= (~)
asy
[(-12SL + SOOx) + (2S0L - SOOx)]

=
E
(~) (12SL) = 12SN/cm 2

ngi
Since a single element model with constant area of cross section is used,

nee
stress along the element remains constant and is not realistic.
(c) The bar is modeled by one 3-noded element with area of cross section

rin
varying along the length of the bar. The element stiffness matrix is obtained
from
[Ke] = J[B]T[D][B] dv = J[Bf E [B] A(x) dx = E J[B]T[B] A(x)dx
v L L g .ne
where [B]= (~2 ) [(-3L+4x) (-L+4x) (4L-8x)] as derived earlier.
t
On integration of the above with A(x) = Al - (AI - A3).xlL = a - b x,
where, a = Al and b = (AI - A3)/L, we get

[K]= JE(aUbX)
ge - 24Lx + 16x 2
3e-16Lx+16x2
3L2 -16Lx + 16x 2
L2 -8Lx + 16x 2
-12L2 +40LX-32X
2

- 4e + 24Lx - 32x 2 dx
j
[ -12e +40Lx-32x 2 -4L2 +24Lx-32x2 16e -64Lx+64x2

l4a-3bL 2a-bL -16a+4b Ll

=~ 2a - b L 14a - 11 b L -16a+ 12b L
6L
r -16a + 4b L - 16a + 12b L 32a-16b L

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 165

-12A, - 4A
E
128 r
16 -144
- 4A, -12A33J = - 16 96 -112
16A 1 + 16A 3 6L -144 -112 256 J
Since the chosen model has only one element, assembled stiffness matrix
is also the same. Therefore,

ww
w.E f---.P

asy
E
Note: Care should be taken while numbering the load and displacement

ngi
components, since the 3 rows of the stiffness matrii\. correspond to the
end nodes at x = 0, x = L and the mid point (at x = Ll2) respectively.

{P3} = {IOOO} = 8E [ 6
n
Applying the boundary condition, u, = 0, the above equations reduce to

eer
P 2 0 3L-7
Solving these equations, we get ing
u3 } = _1 [16 7] {IOOO} .ne
where,
{u

Det =
2 Det 7

(~~) (6 J(
6 0

16 - 7 x 7):: 125 (~)

t
7 x 1000 56L
Therefore,
= (125~) =T
2
u

16 x 1000 128L
and u3 = =--
(125~) E

Downloaded From : www.EasyEngineering.net

:166 FINITE ELEMENT ANALYSIS

tJ
Stress in the bar is given by

a ~ E[B]{u.} ~ E(~r [(-3L+ 4x) (-L + 4x) (4L- 8x)]

=E (~2) (-3L+4x)uI +(-L +4x) U3 +(4L-8x) U2J

ww =E(~2 )[(-3L+4X).0+(-L+4X).(12;L )+(4L_8X).(S:L)]

w.E = (~) [(-128L + S12x) + (224L- 448x)] = C~} (6L + 4x)

a
Stress at node 1, for x = 0, is

syE
Stress at node 2, for x = ~ ,is
0"1 =

0"2 =
(1~)

(1~)
(6 L + 4 x 0) = 96 N/cm 2

(6L + 4 x ~) = 128 N/cm

ngi (~) (6 L + 4 x L) = 160 N/cm 2

Stress at node 3, for x = L, is 0"3 =

nee
rin
Analysis of results
All the three models have 3-nodes, but gave different solutions. Exact solution
depends on how closely the assumed displacement field or stress distribution
matches with the actual displacement field. In this example, it is clear from
basic equilibrium condition that stress is linearly increasing from PIAl at node-lg .ne
to P/A 3 at node-3 and hence, displacement must increase in a parabolic form.
The results are tabulated below.

Model Displacements at Stresses (N/mm' at

t
Node-2 Node-3 Node-l Node-2 Node-3

Two2-noded III UE 254 LIE III lll1l43 143

Ave 127

One 3-noded, Constant area 62.5 LIE 125 LIE 125 125 125

One 3-noded, varying area 56 LIE 128UE 96 128 160

Exact P/A 1= 100 P/A2=125 P/A,=167

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 1.67

(i) The I sl model assumes linear displacement field and hence constant
stress over each of the two elements. At the common node between
elements, stress has a step function and the average value can be
taken to represent stress at that node. Results can be improved by
taking more such elements in the model.
(ii) The 2nd model uses a single 3-noded element, assuming constant
(mean) area of cross section along the element. Hence, even though
parabolic displacement is considered, linear displacement field and
constant stress are obtained.

ww (iii) The 3rd model consisting of a single 3-noded element, whose stiffness
matrix is derived for varying cross section area along the element,

w.E
gives parabolic displacement and linear stress representing the true
situation. Hence, this model gives best results.
(b) Higher order Continuum elements

asy
Higher order elements are more commonly used for analysing 2-D and 3-D
structures. Linear strain triangle (LST) will have six unknown coefficients

E
to include all terms upto second order, as shown in Pascal's triangle. For

ngi
evaluation of these six coefficients, six nodal values are required in each
element. Thus, the 6-noded element is formed by including midpoints of the

are
n
three sides as the additional nodes. The functions for u and v displacements

u(x) = a) + a2.x + a3.y + a4.x2 + a5. xy + ~.I ' eer

and v(x) = a7 + ag.x + a9.y + alO.x2 + all. xy + a\2. y2
The function for LST element, as in the case of CST element, ing
IS also
complete and isotropic.

.", .ne
/. .~
. I·
/ .
~.
t
/.----_.- ~_. . . ""
.-
.----- .
/
6-noded CST element lO-noded QST elemenl

Quadratic strain triangle (QST) will have ten unknown coefficients to include
all terms upto third order, as shown in Pascal's triangle. For evaluation of these
ten coefficients, ten nodal values are required in each element. Thus, the 10-
noded element is formed with 6 additional mid-side nodes and I internal node.
The functions for u and v displacements (complete and isotropic) are
u(x) = al + a2.x + a3.y + a4.x2 + a5. xy + ~.I + a7.x3 + ag.x2y + ~.xy2 + alO.y3

Downloaded From : www.EasyEngineering.net

168 FINITE ELFMENT ANALYSIS

and
VeX) = all + al2.x + al3.y + a14.x2 + a15. xy + al6.l + al7.x3 + als.x2y + al9.xl
+ a2o·i
Similarly, the displacement functions for a 10-noded 3-D solid tetrahedron
element with additional mid side nodes along its 6 sides (complete and isotropic
function) are

ww .j'\\.. \
w~-.-.-.\'-.-~
..Ea. ~

syE ~./
IO-noded tetrahedron clement

ngi 20-noded hexahedron element

U= al + a2.x + a3.y + ~.Z + a5.x2 + ~.y 2 + a7.Z2 + as.xy + a9.yz + alO.zx

nee
V = all + al2.x + al3'y + al4.z + a15.x2 + al6.l + a17.z2 + alS.xy + al9.yz + a20·zx
w = a21 + a22.x + a23.y + a24'z + a25.x2 + a26.l + a27. z2 + a2S· XY+ a29·yz + a30. zx

rin
and the displacement functions for a 20-noded 3-D solid hexahedron element
with additional mid side nodes along its 12 sides (incomplete but isotropic) are
U= a1 + a2.X + a3.y + ~.Z + a5.X2 + a6.y2 + a7.z2 + as.xy
g
+ ~.yz + alO.zx + all.x3 + al2.x2y + a13.xl + a14.y3 + al5.y2z + al6.yr
+ al7.z3 2 2 + a20.xyz .ne
V= a21
+ alS.x Z + a19'XZ
+ a22X + a23.y + a24.z + a2S· x2 + a26.y2 + a27.z2 + a2S·xy
+ a29.yz + a30·ZX + a31.x3 + a32.x2Y + a33. xy2 + a34·y3 + a3S'y2z + a36·yz2
t
+ a37.z 3 + a3S.x2z + a39.xz 2 + ~o.xyz
W = ~I + a42· x + a43.y + ~4'Z + a4S. x2 + a46.l + a47.r + ~s·xy
+ ~9·YZ + aSO·zx + aSI.x3 + aS2.x 2Y + aS3. xy2 + aS4.y3 + aSS.y2Z + aS6·yz2
_3 2 2
+ aS7.z + aSS'x z + a59.XZ + ~o.xyz
Note: Whenever possible, complete and isotropic displacement polynomial is
used. If, however, it is not possible, preference is given to isotropy rather than
completeness of terms of a particular order. In the above cases of higher order
elements with incomplete polynomials, there is no unique combination of the
terms and many other combinations of terms in the selected polynomial are
possible. Different software may thus use different polynomials.

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 1.69

6.2 ISOPARAMETRIC ELEMENTS

The derivation of stiffness matrix by the method described so far involves

integration of the strain energy over the surface or volume. For straight
boundaries, this integration can be carried out by numerical techniques. Higher
order elements are developed with better displacement functions so that
accurate results can 'be obtained with lesser number of elements. Inherent
disadvantage with these elements is that as the size of the element increases,

ww
accuracy of boundary representation reduces since edges of element boundary
are always assumed as straight lines. Also, as the size of the polynomial

w.E
increases, inversion of G matrix, linking nodal displacements to the coefficients
of the polynomial, takes more time.

asy
E ngi
nee
rin
(a) Unrefined elements
of simple shapes
(b) Refined elements
of simple shapes
g
(c) Refined elements
of complicated shapes
.ne
FIGURE 6.1 Representation of curved boundaries

A need was therefore felt to improve the method, in order to idealise the
given structure with curved boundaries more accurately.. In this method, element
t
geometry as well as displacements are interpolated over the element using
shape functions or interpolation functions Ni in terms of natural or
intrinsic or non-dimensional coordinates.
Two types of shape functions are commonly used.
• Lagrange interpolation function, which matches the function value
(displacement) at specified points or nodes
• Hermite interpolation function, which matches function value
(displacement) as well as its derivatives (slopes) at the specified nodes.

Downloaded From : www.EasyEngineering.net

i.70 FINITE ELEMENT ANAL YSIS

Curvilinear orthogonal coordinates ~, " and l;, whose magnitudes vary from
-1 to +1 in any element, are used in place of cartesian coordinates x, y and L or
cylindrical coordinates R, e and Z. The shape functions are then defined in
terms of the natural coordinates, which link displacement at any point in the
element to the nodal displacements thus avoiding the need for inversion of G
matrix. It is obvious that there will be as many shape functions as the number of
nodes in the element. Each shape function will have a value equal to unity at

ww one node and a value equal to zero at all other nodes

Three types of elements are possible.

w.E • If a higher order function is used to represent displacement and a lower

order function is used to represent geometry, it is called a sub-

• asy
parametric element.
If a lower order function is used to represent displacement and a higher

•
E
order function is used to represent geometry, it is called a super-
parametric element.
ngi
If functions of same order are used to represent displacement as well as

nee
geometry, then the element is called an iso-parametric element.
Of these, iso-parametric elements are most commonly used.

6.3 rin
g
STIFFNESS MATRICES OF SOME ISO-PARAMETRIC ELEMENTS

(a)I-D linear interpolation for a truss element

.ne
Let the non-dimensional coordinate for the 1-0 element be defined by
~= 2(x-x\)_1
(x 2 -x\)
,
t
so that at node 1, x = x\ and ~ = - 1 while at node 2, x = X2 and ~ = 1
Then, the shape functions N\ and N2 are defined in terms of~, as shown in
Fig.6.2, by
N\(~)= (l;~); N2(~)= (l~~)
It can be seen that the shape function N\ = 1 for x = x\ and ~ = - 1 at node 1
and N\ = 0 for x = X2 and ~ = + I at node 2, while N2 = 0 at node 1 and
N 2 = 1 at node 2.

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 171

Then, cartesian coordinate 'x' and axial displacement 'u' at any point on the
element are given by
x = N I XI + N2 X2 = [N] {x}
and u = NI UI + N2 U2 = [N] {q}

N2i~

ww ~=-I ~=o ~=+I

~
~=-I ~=O
JI.
~=+I ~

w FIGURE 6.2 Shape functions of a 2-noded truss element (Case-i)

.Ea
The strain E = du can now be expressed, using chain rule of differentiation, as

syE
dx

E= du d~ =[dN I U + dN 2 u ]d~
d~ dx d~ I d~ 2 dx

=[-UI+~] 2 ngi
=-U I +u 2 =[BHq}
2
1
2 (X2 - XI)

~] and
L
nee
rin
where, [8] = [-L L = X2 - Xl

Then,
g.n
This stiffness matrix is identical to the one obtained by polynomial method
described earlier.
Alternative coordinates - There is no restriction on the choice of the origin of
et
local coordinate system. For example, if the origin is taken at the left end of the
truss element, as shown in Fig. 6.3, then the non-dimensional coordinate ~ is
given by

~ = (x - XI) / (X2 - XI)

At node I, x = XI and ~ = 0 while at node 2, x = X2 and ~ = 1
The shape functions are now defined by NI(~) = I-~ ; N2(~) = ~
which give the values NI = 1 for ~ = 0 at node 1 and NI = 0 for ~ = 1 at node 2
while N2 = 0 at node 1 and N2 = 1 at node 2.

Downloaded From : www.EasyEngineering.net

:172 FINITE ELEMENT ANALYSIS

The Cartesian coordinate x and displacement u are again defined by

jN2 ~

ww 1;=0 1;=+1
~
1;=0
iI.
1;=+1 I;

w.E FIGURE 6.3 Shape functions of a 2-noded truss element (Case-2)

a E -

syE
_ du d~ -_(_ lI) + u ).
d~ dx
2
1
x2 - Xl
_
-
-lI)
L
+u 2 -[8]{
- q}

(~)
where, [8]= [-1

~ fB)'[D][B]dV ~
1];
n1gi L=Xl-XI

[BY E [B]A dx ~ ALE [ ~ 1 -11]

[K,]
nee
This stiffness matrix is also identical to the one obtained by polynomial
rin
method described earlier.

(b)l-D Quadratic interpolation for a truss element g .ne

The non-dimensional coordinate, with mid-point of the element (node 3) as
.. . . b): 2( x - X 3)
t h e onglO, IS gIven y ~ = ----'-----"-'--
(X2 -x)
t
so that with X2 - Xl = L, Xl - X3 = - Ll2 and X2 - X3 = Ll2 the non-
dimensional coordinates at the three nodes are ~) = -1; ~2 = 1; ~3 = 0
2
Then,
L

Corresponding shape functions N], N2 and N3 are given by

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 173

They are graphically represented in Fig. 6.4.

S=-l ~=O S=+l S

S=-l S=O

ww FIG U R E 6.4 Shape functions of a 3-noded truss element

The local Cartesian coordinate x and the displacement u are then given by

w x=N 1 x\+N2X2+N3X3=[N] {x}

.Ea
U = N\ UI + N2 U2 + N3 U3 = [N] {q}

E syE
Strain, expanded by the chain rule of differentiation, is given by

= du dS = [dN \ u + dN 2 U + dN 3 U ] dS
dS dx dS \ ds 2 ds 3 dx

~ [ _ (1- ~S)UI + (1 + ~S)U2 ngi _ 2SU 3] = [B] {q}

[B]=~[ (1- 2S)

nee
where,
L 2
(1 + 2S)
2
_ 2S]

r
l7 i-8]n
[Ke] = fiBY [O][B]dV = fA [BY E [B]dx = ~~
L
1
1
7 -8 g.n
v 0

This stiffness matrix is identical to the matrix obtained earlier for the 3-
-8 ":'8 16
et
noded truss element.

(c) I-D interpolation for a beam element

Shape functions for the beam elements differ from those of truss elements,
since derivatives of displacement (slopes) are also involved. Hermite
functions, H" which satisry deflection and slope continuity, are used so that
deflection at any point is given by

w = H\w\ + H 29\ + H3W2 + H492 or H\w\ + H 2(dw/dx)\ + H3W2 + ~(dw/dx)2

= H\w\ + (2/L) Hldw/dS)\ + H3W2 + (2/L) H 4(dw/dS)2

Downloaded From : www.EasyEngineering.net

174 FINITE ELEMENT ANALYSIS

since ~: =( }()=(~){ ~~)

Here, Cartesian coordinate x of any point in the element is related to non-
dimensional coordinate ~ by

where, x = XI (_I-_~) + X?
2 -
(_I+_~)
2
= (XI + x 2) + -,,~_(X---=2,--_X---,I,-,--)
2 2

ww for

and
-I :s; ~:s; + 1

dx = (X2 -XI) =(L/2) d~ 2

or - = -

w.E d~ 2
Hi = a, + b,~ + C,~2+ d,~3
dx L
V i=I,2,3,4

asy
By imposing the end conditions

~=-I

and
E WI =Oand (dW) =0 at
d~

W2 = 0 and (dW) = 0 at
I

ngi ~, = 1
d~ 2
n
these H, functions and their derivatives take the following values.
eer
HI
at ~ =-1 at ~ = +1
0 HI'
at ~ =-1
0 0ing
at ~ = +1

H2 0 0 H2' 0
.ne
t
H3 0 1 H3' 0 0
H4 0 0 H4' 0
Coefficients a, b, c, d can then be obtained as
1 3 1
a l =- b l =-- CI =0 d l =-
2 4 4
I -I -I 1
a 2 =- b2 =- c2 =- d 2 =-
4 4 4 4
1 3 -I
a 3 =- b 3 =- C3 = 0 d3 = -
2 4 4
-I -1 1 1
a --' b4 = - c 4 =- d 4 =-
4- 4 ' 4 4 4

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 175

_ (l-~)2(2+~) (2-3~+~3)
Th us, H I - -'---':":"--'--"'- or
4 4

H2 = (l_~)2(~ + I) (1 _ ~ _ ~ 2 + ~ 3 )
or
4 4

H3 = (l+~)2(2-~) (2+3~-~3)
or
4 4

ww H4 = (l+~)2(~-1)
4
or
(_1_~+~2 +~3)
4

w.E
Variation ofthese functions over the beam element is plotted in Fig. 6.5.

_________________ =
__ asy
En
~-Y:

Node I
Node}

gin
I eer
,'----=- ---------- ---------
Node I Node}
ing
FIGURE 6.5 Hermite Shape functions of a 2-noded beam element

(d) 2-D linear interpolation for a triangular element .ne

The non-dimensional coordinates ~ and 11 and the shape functions N I , N2
and N3 are also at any point P are given by above the figures.
t
Where A is the area of the triangle; Al is the area of the triangle formed by
points P, 2 and 3; A2 is the area of the triangle formed by points P, 3 and I;
and A3 is the area of the triangle formed by points P, 1 and 2, as shown in
Fig. 6.6. Hence, the shape functions Nh N2 and N3 are also called area
coordinates.
It can be seen that for any point P, A = Al + A2 + A3
and so N1 +N2 +N3 =1 or N3=1-NI':"N2='1-~-11

Downloaded From : www.EasyEngineering.net

1.76 FINITE ELEMENT ANALYSIS

The local Cartesian coordinates x and y and displacements u and v along

these Cartesian coordinates are given by
x=N I XI +N 2 X2 + N3 X3 = XI~+ x 211+ X3 (1-~-11)

Similarly, y = Y\3~ + Y2311 + Y3 where, Y\3 = YI - Y3 and Y23 = Y2 - Y3

..... (6.1)

ww The Cartesian coordinates as well as corresponding non-dimensional

coordinates are shown in Fig. 6.6 for a 3-noded triangular element.

w.E
y

2(0,1)

a syE x

ngi ~----------~~~
3(0,0)

nee
FIGURE 6.6 Mapping of a triangular element in ~-11 coordinate system

The displacements u and v can be represented in terms of the same non-

dipJensional coordinates as
rin
u = NI UI + N2 U2 + N3 U3
= (UI- U3)~ + (U2- u3)11 + U3
v =N I VI + N2 V2 + N3-V3
g .ne
= (VI- V3~ + (V2- v3)11 + V3
These equations can also be represented in matrix form by
t
UI
VI

{:}=[:I
0
NI
N2
0 N2
0 N3
0 :J u2
v2
u3
or {u} = [N] {q}

Shape functions are plotted in Fig. 6.7 with non-dimensional coordinates

indi~ed for each node. The three shape functions have a value of unity at

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 177

one node and a value of zero at all other nodes. In the figure, A), A2 and A3
indicate areas used for the calculation of non-dimensional coordinates I; and
II of any point P.
_ _---:---::::::::;.,3(0,0)

T I

1(1.0) ~

ww 2(0.1) 2(0.1 )

w.E
3(0,0)
1
J...-

1(1 ,0)

asy
E
2(0.1)

ngi 2(0.1)

FIGURE 6.7 Shape functions of a 3-noded triangular element

(e) 2-D quadratic interpolation for a triangular element

nee
the non-dimensional coordinates 1;, II and ~ remain same as above but the
number of shape functions are increased to six corresponding to the six
rin
g
nodes of the element and are given below
N I = 1;(21; - 1)
N2 = 1l(211 - 1) .ne
N3 =~(2~ - 1)

N4 = 41;1l
t
N s = 411~
N6= 4~1;
Lines with constant values of N I, N2 and N3 are shown in Fig. 6.8 while
variation of shape functions is plotted in Fig. 6.9.
The cartesian coordinates x and y and displacements u and v in the element
are given by
x = NI XI + N2 X2 + N3 X3 + N4 X4 + Ns Xs + N6 X6

Downloaded From : www.EasyEngineering.net

178 FINITE ELEMENT ANALYSIS

y = Nl Yl + N2 Y2 + N3 Y3 + N4 Y4 + Ns Ys + N6 Y6
U = Nl Ul+ N2 U2 + N3 U3 + N4 ll4 + Ns Us + N6ll6
V =N l Vl+ N2 V2 + N3 V3+ N4 V4+ Ns vs+ N6 V6

ww
w.E N3 = 0

asy
AGURE 6.8 Lines of constant shape function value over the triangular element

E ngi
n eer
ing
3

.ne
~----------~----------~2
4 t
FIG U R E 6.9 Shape functions of a 6-noded triangular element

(f) 2-D linear interpolation for a quadrilateral element

Unlike in the case of triangular element, which is identified by three non-
dimensional coordinates each having values in the range 0 to 1, a
quadrilateral element is identified by two non-dimensional coordinates each
having values in the range -1 to + 1, as shown in Fig. 6.10.
x = Nl Xl + N2 X2 + N3 X3 + N4 ~
Y = Nl Yl + N2 Y2 + N3 Y3 + N4 Y4

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 179

y
4(-1,1) , - - - - - t - - - - - , 3( 1.11

ww 1(-1,-1) 2( 1,-1)

w.E
FIGURE 6.10 4-noded quadrilateral element in ~ - 11 coordinate system

a= Nl
syE
The coordinates and displacements at every point in the element are
expressed in tenns of nodal values, using shape functions N), N 2, N3 and N4 as

ngi
u Ul + N2 U2 + N3 U3 + N4 U4

Shape functions at each node of a 2-D element can be derived as the product
nee
rin
of shape functions along ~ direction and II direction passing through the
particular node.

Thus,
g.n
where Nl!; and Nl'l are the shape functions of I-D elements with

~ = -1 to +1 and II = -1 to + 1
et
N = (1-~) (1-11) = (1- ~)(1 -ll)
I 2' 2 4

Similarly we can get N = (1 + ~)(1-11) . N = (1 + ~)(1 + ll)

, 2 4 ' 3 4

and N4 = (1- ~Xl + ll)

Their values are graphically represented in Fig. 6.11.

Downloaded From : www.EasyEngineering.net

180 FINITE ELEMENT ANALYSIS

ww
w.E FIGURE 6.11 Shape functions of a 4-noded quadrilateral element

a syE
(g) 2-D quadratic interpolation for a quadrilateral element

ngi
This quadrilateral element, with 4 corner nodes and 4 mid-side nodes, is also
identified by two non-dimensional coordinates each having values in the

nee
range -I to + I . But, the coordinates and displacement of any point in the
element are expressed by using 8 shape functions (Ref. Fig. 6.12), as

rin
g.n
where,
2
et
N 1 = - (1- ~)(I -1l)(1 + ~ + ll) N 5 = (1- ~ )(1-11)
4 2

N 2 = - (I + ~)(1-11)(1- ~ + ll) N 6 = (1 + ~)(1 -1l2)

4 2

= (1- ~
2
N 3 = - (1 + ~)(1 + 1l)(1- ~ -ll) N7 )(1 + ll)
4 2

N4 = - (1- ~)(1 + 1l)(1 + ~ -ll) N 8 = (1- ~)(1-112 )

4 2

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISQPARAMETRIC ELEMENTS 181

4( -1,1) ...-----+-----, 3(1, I)

7(0, I)

8(-1,0) 6(1,0)

5(0,-1 )

ww x
1(-1,-1 ) 2(1,-1)

w.EFIGURE 6.12 Mapping of 8-noded quadrilateral in ~-T} coordinate system

(h)

asy
3-D linear interpolation
• A 8-noded 3-D element is identified by three non-dimensional coordinates
each having values in the range - I tQ + t, as shown Fig. 6. t 3.
y
E ngi 11 3(1,1,-1)

nee
rin 6(1,-1,1)

~---------------------+X
5(-1,-1,1) g .ne
FIGURE 6.13 Mapping of B-noded 3-D solid element in ~-1l coordinate system

The coord.inates and displacement functions are given by

t
x = NI XI + N2 X2 + N3 X3 + N4 ~ + N5 X5 + N6 X6 + N7 X7 + Ns Xs
Y = N 1 YI + N2 Y2 + N3 Y3 + N4 Y4 + Ns Ys + N6 Y6 + N7 Y7 + Ns Ys
Z = N I ZI + N2 Z2 + N3 Z3 + N4 Z4 + Ns Zs + N6 Z6 + N7 Z7 + Ns Zs
and
u = NI UI + N2 U2 + N3 U3 + N4 U4 + Ns Us + N6 ll6 + N7 U7 + Ns Us
v= NI VI + N2 V2 + N3 V3 + N4 V4 + Ns Vs + N6 V6 + N7 V7 + Ns Vs
W = NI WI + N2 W2 + N3 W3 + N4 W4 + Ns Ws + N6 W6 + N7 W7 + Ns Wg

Downloaded From : www.EasyEngineering.net

1.82 FINITE ELEMENT ANALYSIS

Shape functions at each node of a 3-D element can be derived as the product
of shape functions of I-D' elements along ~ direction, 11 direction and
~ direction passing through the particular node.

Thus,
NI = NI!; . NIl] . NI~
where NI!;, NIl] and NI~ are the shape functions of I-D elements with
~ = -I to + I, 11 = -I to + 1 and ~ = -I to + 1

ww NI = (l-~)(I-11) (l-s)
222

w.E =-,--(1_-..=.;..~)-,-(1_-11--=..:. ).:. . (1_-s.=.:. . )

a
Similarly, we can get N2

syE N3
= (I +~)(1-11)(1-s)

= (1 + ~)(I + 11)(l-s)
8

N4 ngi
= (l-~)(l+11)(1-s)
8

N s = (1- ~)(I - 11)(1 + s)

8
nee
rin
8
N6 = (l + ~)(l-11)(1 + s)

N7 = (l + ~)(I + 11)(1 + s)
8

8
g .ne
Ns = (l-~)(1 +1)(1 + ~)
8 t
SA JACOBIAN

In order to have unique mapping of elements, there should be only one set of
cartesian coordinates for each set of corresponding non-dimensional coordinates.
Fo[..a 2-D plate element in x and y coordinates, using chain rule. for partial
derivatives,
au au ax au ay
-=--+-- Similarly for au
a~ ax'~ ay'a~ ~

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 183

Expressing them in matrix fonn,

{aulOrt
au/~} [Ox/~ Oyla~]{auIOx} {au'Oxl
= OxIOrt OyIOrt aulOy =[J] aulOyf
where [J] is called the Jacobian or a matrix of partial derivatives of Cartesian
coordinates of the element w.r.t. non-dimensional local coordinates of the
element

From eq. 6.1, [J] = [XI3 YI3] or [Jrl = _1_[ Y23

ww X 23 Y23 Det J - x23

If nodes are numbered counter-clockwise, in a right-handed coordinate

w
system, det J is +ve.

.
For a 3-noded triangular plate element, from eq. 6.1

Ea
[J]=l~ ~J==[XI3
Ox Oy X23
syE
YI3]
Y23
Ort Ort
or [J]-I =_1_[ Y 2l ngi
where,
DetJ -X23

Det J = Xu y 23 - X23 Y13 nee

It can also be expressed as
Xl Yl 0 Xl -X3 YI -Y3 0 x13 rin
DetJ= I x2
1 x3
Y2 = 0 x2 -x3
Y3 1 x3
Y2-Y3 == 0 x23
g.n
et
Y3 1 x3
== XI3 Y23 - YI3 X23
1 1
Area of the element, A =="2 Det J =="2 (XI3 Y23 - X23 .Y 13)
1 I 1
and [K] == f[B]T [D][B] dV == f f f[B]T [D][B] Det [J] d~ dT) dt;
v -I -I -I

For a 3-D element in x, Y and z coordinates, these relations can be expressed

in matrix fonn in a similar way as

au/~}
{au 1at;,
[Ox/~ Oyla~ f)z/~J{auIOx} {au lOx}
au IOrt == OxIOrt Oy 1Ort f)zIOrt aulOy == [J] au lOy
Ox 1at;, Oy 1at;, Oy 1at;, au 1f)z au 1f}z

Downloaded From : www.EasyEngineering.net

:L84 FINITE ELEMENT ANALYSIS

where,
ax las
[1] = ax I Or)
fax I at"
The same Jacobian relates derivatives of displacement components v and w

ww
also.

6.5 STRAIN-DISPLACEMENT RELATIONS

w.E
(a) For a l-noded triangular element

asy
E ngi
n eer
Y23 aulOl; + (-YI3) au/iJIl
=(~tJ){ (-x2])c3v/OS+ xl3 aulOr)
ing }

(-X 23 )aulOS + xl3 au/iJIl + Y23 avla~ + (-YI3) av/ilrJ .ne

Since

u= N, ~, + N2 U2

aulas = u, -
+ N3 U3 = (u,- U3) ~ + (U2- U3)" + U3
U3 = UI3; aula" = U2 - U3 = U23
t
Similarly, av;a~ = v, - V3 = VI3; avla" = V2 - V3 = V23

Therefore,

Y23
{t}=-
1 DetJ r0
X32

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS US

VI
[Y~3
0 Y31 0 Yl2
=_1 u2
DetJ
x32
x32 0
xl3
xl3 0
X:l1 v2
=[8]{q}
Y23 Y31 x21 YI2
u3
v3

ww [yn
where strain-displacement matrix,
0 Y31 0 YI2

w.E [8]=_1_
DetJ
0
x32
x32
Y23
0
xl3
xl3
Y31
0
x21
X:l1
YI2

asy
The elements of [8] are constants and not functions of coordinates (since
shape functions are linear in x and y).and hence, strain in a 3-node triangular

E
element is constant over the entire element.

ngi
Element stiffness matrix [Ke] is obtained from [Ke] = J [8]T [D] [8] dV.
Here, the integration is carried out on non-dimensional coordinates with the

nee
limits 0 to +1 or -1 to + I, depending on the particular case. Element with
curved edges also is mapped into an element with straight edges in the non-
dimensional coordinate system. Hence, for an element with curved boundaries,

rin
integration becomes much simpler in non-dimensional coordinate system
compared to the integration in Cartesian coordinate system.

(b) For a 4-noded quadrilateral (2-D) element

Following the same procedure, we can get [8] platrix. The local non-
g .ne
dimensional coordinates ~ and 11 take values from -1 to +1.
Here, x = NI XI + N2 X2 + N3 X3 + N4 ~
Y == NI YI + N2 Y2 + N3 Y3 + N4 Y4
t

Downloaded From : www.EasyEngineering.net

186 FINITE ELEMENt ANAI.YSIS

Then, J =[Ox/~ fJy/~]=[JII J 12 ]

[] Ox/en, fJy/en, J 21 J 22

where, J II = (~) 1-(I-T)xl +(I-T)x2+(1 +T)x3-(1 + T)x.11

J 12 = (~) [- (I - T) YI + (I - T) ) Y2 + (I + T) Y3 - (I + T) Y4]

ww J21 = (~) [- (I -~) XI - (I +~) X2 + (I +~) X3 + (I -~) '4]

w .Ea
J22 = (~) [- (I-~)YI -(I +~)Y2+(1 +~)Y3+(I-~)Y4]

Now, { au/fJy syE

au/Ox} = (I/0etJ) [ J22
-J 21
-J I2 ] {au/fff,}
J I1 au/Ori

and {8v/Ox} = (I/DetJ) [ J 22

8v/fJy -J 21 ngi -J 12 ]{8v/~}
J I1 8v/Ori

r& }
{e}= 8v/fJy
au/fJy+8v+Ox
nee
rin
r -J 12 0

J~I 1
au/~
g.n
= O;tJ
J"
0
-J 21
0
J II
-J 21
J 22 -J 12
au/Ori
8v/~
8v/Ori
= [B]{q}
et
where,

[B]=_I_ J"
OetJ
.r
0
-J 12
0
0
-J 21 J~I 1x
-J 21 J I1 J22 -J 12

[-1(1-~)
-(1~~)1
'0 (1- T) 0 (1 +T) 0 -(1 +T)
1 -(l-~) 0 -(1 +~) 0 (I +/;) 0 (I-/;)
4 0 -(I-T) 0 (1- T) 0 (l +T) 0
0 -(l-/;) 0 -(I +S) 0 (l +S) .0 (I-S)

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 187

Elements of this matrix are functions of coordinates, since shape functions

include higher order terms and hence strain in this element is not constant
over the element.

(c) For a 8-noded quadrilateral (2-D) element

Following the same procedure, we can get [B] matrix. The local non-
dimensional coordinates; and 1') take values from -1 to + 1.
x = NI XI + N2 X2 + N3 X3 + N4 Xi + Ns Xs + N6 X6 + N7 X7 + Ns Xs

ww y = NI YI + N2 Y2 + N3 Y3 + N4 Y4 + Ns Ys + Nt'y6 + N7 Y7 + Ns Ys
u = N I UI+ N2 U2 + N3 U3 + N4 U4 + Ns Us + N6 U6 + N7 U7 + Ns Us

w.E
where,
V =N I VI+ N2 V2 + N3 V3+ N4 V4 + N5 V5+N6 V6 + N7 V7+ Ns Vs

asy
NI = - (1- ;)(1-1')(1 +; + 1')
4 .
2
N s = (1_; )(1-1')
2

E
N 2 = - (1 + ;Xl-1')(1- ; + 1')
4

N 3 = - (1 + ;)(1 + 1')(1-; -1') ngi

N 6 = (1 + ;Xl _1')2 )
2
= (1-;2)(1+1')
nee
N7
4 2
N = (1_;)(1_1')2)
rin
N 4 = - (1- ;)(1 + 1')(1 + ; -1')
g
4 2
(d) For a 4-noded tetrahedron (3-D) element
Following the same procedure, we get g .ne
{~=
au/fJx
auliJy
Ow 1&
= [Jr
l
aula;
aul8rJ
Ow I fX:,
= [B]{q}
t
liJy+iNlfJx au 18rJ + iN I iJt,..
iNl&+OwliJy iN I fX:, + Ow 18rJ
Ow/fJx+au/& Ow 10; + au I fX:,

Here, [J] =[::: ~:: :::1

X34 Y34 Z34
l
If the elements of [Jr are designated by the constants A jj , then

Downloaded From : www.EasyEngineering.net

188 FINITE ELEMENT ANALYSIS

[All
l
[Jr = A21
AI2
A22
An]
An
A31 A32 A33

and AI' = All + AJ2 + AI3 ; A 2' = A21 + A22 + An; A3' = AJI + Au + A33
then,
All 0 0 AI2 0 0 AI3 0 0 -A; 0 0

ww [B]=
0
0
A21
0
0
A31
0
0
A22
0 A32
0 0
0
A23
0
0
A33
0
0
-A;
0
0
-A;

w.E A31
0 A31
0
A21
AJI
0
A32
A32
0
A22
AI2
0
A33
A33
0
A23
AI3
0
-A;
-A;
0
-A;
-A;

asy
A21 All 0 A22 AI2 0 A23 AI3 0 -A; -A; 0

Here again, elements of [B] matrix are constants since shape functions are

element. E
linear functions and, hence, strain in this element is also constant throughout the

ngi
6.6 SUMMARY

• nee
Higher order elements are broadly classified as - Serendipity

rin
elements, having no internal nodes and Lagrange elements, having
internal nodes which can be condensed out at the element level before

•
assembling.
g
Higher order elements use higher degree displacement polynomial
.ne
•
and can represent true situation with lesser number of elements than a
model with lower order elements.
Isoparametric elements model displacement as well as boundary with
t
the same polynomial. A small number of higher order isoparametric
elements can model curved boundaries of component accurately.
Even a large number of lower order and non-isoparametric elements
can only approximate curved boundary by a set of straight lines.
• Isoparametric elements use non-dimensional coordinates / shape
functions and facilitate programming for numerical integration on a
computer.

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 189

OBJECTIVE QUESTIONS

1. Curved boundary is better modeled by usi"g

(a) non-dimensional shape functions (b) higher order elements
(c) more number of simple elements (d) isoparametric elements
2. Sum of shape functions at a point is
(a) 1 (b) 0 (c) any +ve integer (d) any -ve integer

ww
3. When fewer nodes are used to define the geometry than are used to define
the displacement, the element is called _ _ element

w.E
(a) subparametric
(c) superparametric
(b) isoparametric
(d) complex

asy
4. When same number of nodes are used to define the geometry and
displacement, the element is called _ _ element

E
(a) subparametric (b) isoparametric
(c) superparametric (d) simple

ngi
5. When more nodes are used to define the geometry than are used to define the
displacement, the element is called _ _ element
(a) subparametric
(c) superparametric
n
(b) isoparametric
(d) complex eer
6. Derivatives of displacement function with respect to element coordinate
system and non-dimensional coordinate system is given by ing
(a) Lagrangian (b) Poisson
.ne
t
(c) Gaussian (d) Jacobian
7. Number of shape functions for a triangular plane stress element are
(a) 2 (b) 3 (c) 4 (d) 6
8. Number of shape functions for a quadrilateral plane stress element are
(a) 2 (b) 3 (c) 4 (d) 8
9. Number of shape functions for a 8-noded quadrilateral' plane stress element
is
(a) 2 (b) 3 (c) 4 (d) 8
10. Shape functions for a triangular plane stress element are also called
(a) r-s coordinates (b) area coordinates
(c) volume coordinates (d) x-y coordinates

Downloaded From : www.EasyEngineering.net

190 FINITE ELEMENT ANALYSIS

SOLVED PROBLEMS

Example 6.1
For a point P located inside the triangle shown in figure, the shape functions NI
and N2 are 0.15 and 0.25 respectively. Determine the x and y coordinates of
point P.

Solution

ww A triangular element will have three natural or non-dimensional coordinates N),

N2 and N3 such that N I + N2 + N3 = I or N3 = I - N 1- N 2.

w.E y 3

asy 2

En ~-----------------------+x

gi
Hence, coordinates of point P i.e., (xp, yp) are given by
Xp = Nlxl + N2X2 + N3X3
nee
= Nlxl + N2X2 + (1- N I - N 2)X3
= 0.15 x I + 0.25 x 4 + (I - 0.15 - 0.25) x 3 = 0.15 + 1.0 + 1.8 = 2.95
yp = NIYI + N 2Y2 + N 3Y3
rin
= NIYI + N 2Y2 + (I - N I- N 2)Y3
= 0.15
Example 6.2
x 1+ 0.25 x 2 + (1 -0.15 -0.25) x

g.n
5 = 0.15 + 0.5 + 3.0 = 3.65

The coordinates and function values at the three nodes of a triangular linear
element are given below. Calculate the function value at (20,6).
Node I Coordinates (13,1) Function value 190
e t
Node 2 Coordinates (25,6) Function value 160
Node 3 Coordinates (13,13) Function value 185
y 3

~-----------------------+x

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 191

Solution
A triangular element will have three natural or non-dimensional coordinates Nt,
N2andN3suchthatNI+N2+N3=1 or N3= I-NI-N2
Hence, coordinates of point P i.e., (xp, yp) are given by
Xp = Nlxl + N2X2 + N3X3 = NI (XI - X3) + N2 (X2 - X3) + X3
20 =N I (13 -13) + N 2(25 -13) + 13

ww yp = NIYI + N 2Y2 + N 3Y3 = NI (YI - Y3) + N2 (Y2 - Y3) + Y3

6 =N I (1-13)+N 2(6-13)+13

w.E =-12NI-7 x L7J+13 or N 1-~

-
144

Hence, asy
E ngi
Function value at (xp, yp) = NIV I + NN2 + NN3

nee
= (35 )XI90+(l-.-)XI60+( 25 )XI85=171.632
144 12 144

rin
Example 6.3
The nodal coordinates of the triangular element are (1,2), (5,3) and (4,6). At the

g
interior point P, the x-coordinate is 3.3 and the shape function at node 1 is 0.3.
Detennine the shape functions at nodes 2 and 3 and also the y-coordinate at the
point p~
.ne
Solution
A triangular element will have three natural or non-dimensional coordinates NJ,
N 2 andN3 suchthatN t +N 2 +N3 = 1 orN3= I-N I -N 2.
t
Hence, coordinates of point P i.e., (x p, yp) are given by
Xp= Ntxt + N2X2 + N3X3 = Nt (XI - X3) + N2 (X2 - X3) + X3
3.3 =N t (1-4) + N 2(5 -4)+ 4 = 0.3 x (-3) + N2 + 4
=N 2 + 3.1 or N2 = 0.2
Hence, N3 = 1 - N t - N2 = 1 - (0.3) - (0.2) = 0.5
yp = NtYt + N 2Y2 + N3Y3 = NI (Yt - Y3) + N2 (Y2 - Y3) + Y3
;, 0.3 x (2 - 6) + 0.2 x (3 - 6) + 6 = -1.2 - 0.6 + 6 = 4.2

Downloaded From : www.EasyEngineering.net

1.92 FINITE ELEMENT ANALYSIS

Example 6.4
Triangular elements are used for stress analysis of a plate subjected to in plane
load. The components of displacement along x and y axes at the nodes i, j and k
of an element are found to he (-0.001, 0.01), (-0.002, 0.01) and (-0.002, 0.02)
cm respectively. If the (x, y) coordinates of the nodes i, j and k are (20, 20),
(40, 20) and (40, 40) in cm respectively, find (a) the distribution of the two
displacement components inside the element and (b) components of

ww
displacement of the point (xp, yp) = (30, 25) cm.

Solution

w (a) Distribution of displacement components u and v inside the element are

.Ea
given by
Up = NI UI + N2 U2 +N3 U3 = -0.001 NI - 0.002 N2 - 0.002 N3
and
syE
Vp = NI VI + N2 V2 +N3 V3 = 0.01 NI + 0.01 N2 + 0.02 N3
(b)A triangular element will have three natural or non-dimensional coordinates
Nt. N 2 andN 3 suchthat
ngi
N I +N 2 +N3 =1 or N3=I-NI-N2
Hence, coordinates of point P i.e., (xp, yp) are given by

30 = 20 NI + 40 N2
nee
Xp = Nlxl + N2X2 + N3X3 = Nlxl + N2X2 + (1 - N I - N 2)X3
+40 (I - N I - N 2)
=40-20N I
rin
Therefore,
Similarly,
20N I = 10 or NI =0.5
yp = NIYI + N 2Y2 + N 3Y3 = NIYI + N 2Y2 + (1- N I - N 2)Y3 g.n
25 = 20 NI + 20 N2 + 40 (1 - N I - N 2)
= 40 - 20 N I - 20 N2
et
Therefore, 20N2 = 15-20NI = 15-10 or N2 =0.25
and N3 = 1 - N I - N2 = 1 - 0.5 - 0.25 = 0.25
The displacements u and v at (xp, yp) are given by
Up = NI UI + N2 U2 + N3 U3
= 0.5 x (-0.001) + 0.25 x (-0.002) + 0.25 x (-0.002)
=-0.0015 cm
and Vp = NI VI + N2 V2 +N3 V3
= 0.5 x 0.01 + 0.25 x 0.01 + 0.25 x 0.02 = 0.0125 cm

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGIIER ORDER AND ISOPARAMETRIC ELEMENTS 193

Example 6.5

The nodal coordinates and the nodal displacements of a triangular element,

under a specific load condition are given below.
XI = 0, YI = 0, XJ = I mm, Yj = 3 mm, Xk = 4 mm, Yk = I mm
UI = I mm, VI = 0.5 mm, UJ =-0.05 mm, VJ = 1.5 mm, Uk = 2 mm, Vk =-1 mm
If E = 2 x 10 5 N/mm 2 and !.1 = 0.3, find the stresses in the element

ww
Solution

w.E
On plotting coordinates of the three nodes of the triangular element, the nodes I,
J and K are identified as I, 3 and 2 respectively to represent the nodes in the
counter-clockwise direction around the element.
Y23 = YKJ = I - 3 =-2
asy Y31=YJ(=3-0=3 Y12=YIK=O-I=-l
X32 = XJK = I - 4 =-3
E X13=XlJ=O-1 =-1

Jacobian of the triangular element, J = [Xl3

ngYi
X21 =XKI=4-0=4

I3 ]=[-1
Y23 3-2
-3]
nee
X23

rin
g .ne
~~--------------------------+x
K(4,1)

t
Determinant of Jacobian, [J] = (-I )(-2) - (-3)(3) = II
In a 3-noded triangular element, stresses are constant throughout the
element. The three stress components in the element are given by
{cr} = [0] {E} = [0] [B] {u}
0 Y31 0 Y12
=[0]_1 0
IJ I. X lV" X32 0 Xl3 0 xO" Jrll )
32 Y23 Xl3 Y31 X 21 Y12

Downloaded From : www.EasyEngineering.net

194 FINITE ELEMENT ANALYSIS

1.0
0.5

1.~ [-2
~Ij
0 3 0 -I
-0.05
~ o
[I v1 0 -3 0 -I 0
=(I_EV ' ) 1.5
0 (l-v)/2 11 -3 -2 -I 3 4
2.0
-1.0

ww = 2xl0
5
03
[I 0.3
I oo ]{-4.15}
-70
11 x (1- 0.3 ) ~.
2
0 0.35 9.55

w.E 5 -625 }
= 2xl0 -8:245 N/mm 2
10.01 {
asy
3.3425

Example 6.6
E ngi
x, y, z coordinates of nodes of a tetrahedron element are (30,0,0), (0,10,0),

nee
(0,0,20) and (20, 20, 10). Formulate strain-displacement matrix [8].

Solution

r 1 rin
Yl4
Y24
Y34
ZI4j 10 -20 -
10
Z24 = -20 -10 -10
Z34 -20 -20 10
g .ne
l
All
[Jr =[A]= A21
[
AI2
A22
A 13'j
A23 =(13~~0)
[-300 400 100
400 -100 300
1 t
A31 A32 A33 200 600 -500

A ,_ -(All + AI2 +AI3) _ -(-300+ 400+ 100) _ -2

and
I- 13000 - 13000 - 130

A ,_ -(A21 +A22 +A 23 ) _ -(400-100+300) _ -6

2- 13000 - 13000 - 130

A ,_ -(A3\ +A32 +A33) _ -(200+600-500) _ -3

3- 13000 - 13000 - 130

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 1.95

Then,
All 0 0 A12 0 0 Al3 0 0 -A; 0 0
0 A21 0 0 A22 0 0 An 0 0 -A'2 0
0 0 A31 0 0 A32 0 0 A33 0 0 -A;
[B]=
0 A31 A21 0 A32 A22 0 A33 An 0 -A; -A~
A31 0 All A32 0 A12 An 0 Al3 -A; 0 -AI
A21 All 0 A22 A12 0 An Al3 0 -A~ -A; 0

ww
Thus,
-3 0 0 4 0 0 1 0 0 -2 0 0

w
[B]=C~~ .E
0
0
4
0
0
2
0
0
-1
0
0
6
0
0
3
0
0
-5
0
0
-6
0
0
-3
0
2
2
0 asy
4
-3
0
6
6
0
-1
4 -5
0 -5
0
3 0
-3
-3
0
-6
-2

Example 6.7
4 -3 0
E -1 4 0

ngi
3 0 -6 -2 0

Detennine the deflection at the point of load application using a one-element

model for the configuration shown in figure.
nee
Solution
Numbering the nodes of the element in counter-clockwise direction as shown, rin
Y23=-20; Y31 =0
X 32 = 30 ; Xl3 = -30 ; X21 = 0
; Y 12 =20
g .ne
Jacobian of the triangular element, [J]=[X
X 23
Determinant of Jacobian IJI = (-30) (-20) = 600
Y\3]_[-30
Y 23 - -30
13
0]
-20 t
lOON
3

T
20mm
SON

t= 10 mm
E = 70 Gpa

1 v =0.3

Downloaded From : www.EasyEngineering.net

196 FINlTE ELEMENT ANALYSIS

The nodal displacements {u} can be calculated from {P} = [K] {u}. So, the
stiffness matrix [K] has to be calculated from [K] = [8]T [0] [8]
where,

-20
ww =_1_ 0

r 30
o
30
o
o -30
o 20
0

w.E 600 -20 -30 o o

o 0

asy 30 0
-20 -30

E ngi -9 20
-30 6
0~1

-20 nee
o 30
o 0

IKJ=tArB]T [DUB1= IOx300x70x10

3
o
0
3; =~~ [-20 rin 9
o
o

g
-6 30
600x600xO.91 0 -30 0
20
o
o
o
0
20
10.5 - 7 - 10.5

.ne
715

-315
-390 -315
-3901040
210 315
210
180
-900
o
-400
180
o
210
-140
-210
t
=641
180 -900 0 900 -180 o
-400 180 0 -180 400 o
210 -140 -210 o o 140

After applying boundary conditions, UI = VI = U2 = V2 = 0, we get

or 50} [400 0
{ -100 = 641 0 140
]{u 3
V3
}

Downloaded From : www.EasyEngineering.net

CHAPTER 6 HIGHER ORDER AND ISOPARAMETRIC ELEMENTS 1.97

which give the displacements at node 3 as

50
u 3 =( )=0.000195mm
641 x 400
-100
and V3=( )=-0.0011143mm
641x140

By the conventional displacement polynomial method

ww
(Note: This solution was already given in chapter-5. However, for a quick
comparison between these two methods, it has been repeated here.)

w.E {u} = [I x y] {u} and {v} = [I x y} {~}

The coefficients UJ, U2, U3, ~J, ~2 and ~3 are evaluated from the above in
terms of nodal displacements as

a
r} l' 0s2°lr}
u2 = 1
yE 0 0 u2
u3 I 30 20

n
r 0 600 0It}gilO
u3

neem~:}
60
or r}u2 = 6~0 - 20 0
1
20 - u 2. = 60 - 2 0
u3 L 30 - 30 0 u3 3 -3
Similarly,
rin
{:}[
0
0 ~Ol{:: }
30 20 ~3
g .ne
or
{::}~ 6~ol-~0
600
0 °lr}
20 lO 1
v 2 = 60 - 2 _0
@ 2 Ole}v2
t
~3 30 -30 o V3 3 3 0 V3

{e} = r& }lO

av/8y
8u/8y + av/fJx
= 0 0 0 0 0
0 0
0 0 0

0 rl
U2
U3
~J
~2
~3

Downloaded From : www.EasyEngineering.net

198 FINITE ELEMENT ANALYSIS

0 60 0 0 0 0 UI

-2 0 2 0 0 0 u2

r~
I 0 0 0
3 -3 0 0 0 0 u3
=_1 0 0 0 0 = [8]{q}
60 0
0 0 !J 0
0
0
0
0
0 -2
0 60
0
0
2
VI

v2
0 0 0 3 -3 0 V3

ww where, [8] = _1
60 r-0
2 0
0
2
0 3
0 0
-3
~J
w.E 3 -3 0 -2 0
This matrix is same as the one obtained through iso-parametric approach,

asy
except that the elements of the stiffness matrix in this method correspond to the
displacement vector [UI U2 U3 VI V2 V3]T whereas in the previous method
they correspond to the displacement vector [UI VI U2 V2 U3 v3f. The

En
sequence of load components in the load vector has to be correspondingly
modified. Rewriting the [8] matrix to correspond with the displacement vector
[UI VI U2 V2 U3

gi
V3],

715
-390
-390 -315
1040 nee
210
180
-900
-400
180
210
-140

[K] = t A [8t [0] [8]= 640

-315
180
210
-900
315
0 900
0
rin 0
-180
-210
0
-400 180 0 -180 400
g.n
0

e
210 -140 -210 0 0 140
After applying boundary conditions,
{P}R = [K]R {uh
UI = VI = U2 = V2 = 0, we get
t
or {-~~O} -(0.0~156)[4~0 I~O1{~~}
or U3 = 0.000195 mm
V3 = -0.001114 mm
Check Reactions are obtained from the assembled stiffness matrix,
corresponding to the fixed degrees of freedom, and checked with
R lx +R2X +50=0; R ly +R2y -100=0

Downloaded From : www.EasyEngineering.net

CHAPTER 6 III(illr'R ORDER AND ISOPARAME'r RIC ELEMENTS 1.99

0
R 1X 715 -390 -315 180 -400 210 0
R 1y -390 1040 210 -900 180 -140 0
= 641
R 2X -315 210 315 0 0 -210 0
R2Y 180 -900 0 900 -180 0 0.000195
-0.001114

ww -199.96
122.47
N

w.E
149.96
-22.50

asy
Static equilibrium relations using calculated reactions are
R 1x + R 2X + 50 = -199.86 + 149.96 + 50::::: 0

E
R 1y + R2y -100 = 122.47 - 22.50 - 100::::: 0

ngi
Note: The given thick plate, from university question paper, should not be

n
analysed as a 2-D problem. It can not be solved as a 3-D problem manually.
Finite element analysis gives approximate resull.\' for the engineering
eer
problems. In this simple example, this was evident from the fact that the first
condition was completely satisfied while a small discrepancy was seen in the
second condition with the results obtained. The discrepancy increases with the
order of the reduced stiffness matrix due to numerical rounding off errors ing
associated with matrix inversion techniques using digital computers.
.ne
t

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
En Left Blank"
"This page is Intentionally
gin
eer
ing
.ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 7
FACTORS INFLUENCING
SOLUTION
ww
w.E
7.1. asy
DISTRIBUTED LOADS

E
In the discussion so far, analysis has been confined to structures subjected to

ngi
specified nodal loads. But, many engineering problems include distributed loads
like
•

•
self weight of beams n
Loads along the length of 1-0 elements such as wind load on columns,

eer
Loads along the edges of 2-D elements such as in-plane pressure on

ing
edges of plates or axi-symmetric solids; pressure (bending) load normal
to the surface of the plate
• Loads on surfaces of 3-D elements such as pressure on one or more
surfaces of a thick solid component .ne
• Loads on volumes of 3-D elements such as self weight, centrifugal
force on a rotating component.
Such loads are usually represented by equivalent loads, based on force
t
equilibrium, in strength of materials.
For example, a uniformly distributed load 'p' on a beam AB of length 'L', as
shown in case-l of Fig. 7.1, is approximated by two equal parts of the beam as
shown in case-2. The distributed load on both the parts is transferred to the ends
of the beam as point load of pL/2 and moment due to the distributed load
represented by the resultant load of pL/2 acting at a distance of L/4 from beam
end, as shown in case-3. Thus,

pL
PI =P2 = - and
2

Downloaded From : www.EasyEngineering.net

202 FIN IT EEL E MEN TAN A L Y SIS

These statically equivalent loads are shown in case-4.

p over L

A Case-I B

p over Ll2 p over Ll2 pLl2 at Ll4 pLl2 at Ll4

i ~ ~ ~ ~ ~ ~ ~ ~ ~
from end A
1 from endB

ww A
Case-2
B
A
Case-3
B

w.E
asy Case-4

FIGURE 7.1 Statically equivalent loads

E ngi
If beam AB is simply supported at its two ends, then the reactions based on
the static force equilibrium conditions L F = 0 and LM = 0 will be equal and
opposite to these equivalent loads.

n eer
Finite element method is based on minimum potential energy theory for the
calculation of stiffness matrix or load-displacement relations. It will, therefore,

consistent loads used in FEM, give displacements identical with those from
closed form solutions. ing
be consistent if the equivalent loads are also based on energy. It is noticed that

7.2 .ne
t
STATICALLY EQUIVALENT LOADS VS. CONSISTENT LOADS

Statically equivalent loads, even though satisty force and moment equilibrium,
do not give the same nodal displacements as the actual loads. Consistent loads,
based on energy equivalence, give the same displacements as obtained with the
actual loads, in addition to satistying force and moment equilibrium. For this
reason, consistent loads are used in FEM. This can be verified by the following
example.
Consider a simple cantilever AB of length L, fixed at end A and subjected to
uniformly distributed'load p, as shown in case-l of Fig. 7.2. The statically
equivalent load system is shown in case-2 of Fig. 7.2 while the consistent load
system is shown in case-3 of Fig. 7.2. In these two cases, a point load and a
moment at ends A and B can replace the distributed load. Since end A is fixed,
the load and moment at A add to the reactions at A without contributing to

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFI.ULNCIN(i SOLUTION 203

displacement at any point on the cantilever. Hence, load and moment at end B
only are shown in these figures.

1.\ i i i
Casc·l
~ ~ --.t
x~
II
I, ~*
pl.~g(~l
Casc-2
1;\ ~*
PI'ilJ x~
l ,,,c-3

FIGURE 7.2 Distributed and equivalent loads

ww
Case-1: Actual load
At any section, distance x from 8, from simple theory of bending

w.E ~ pox{; )~ ~ 1
,Bending moment M EI( ::;

asy
where v is the displacement normal to the axis of the beam
E is the modulus of elasticity of beam material
and
En
I is the moment of inertia of the beam cross section

.
Integratll1g,
gin
- EI -dv == -
dx 6
3
P x- + C 1

dv
eer
The constant C I is evaluated from the houndary condition - == 0 at x = L

i
dx

Thus,
1
P L3
C = - - and -EI - = - - -
6
dv
dx
p X 3 P L3
6 6 ng.
px 4 pL3 X
Integrating again, - EI v = - - - - - + C
24 6 2

The constant C 2 is evaluated from the boundary condition v = 0 at x = L

net
pL4 pL4 3pL4 pL4
Thus, C ==--+-==--==-
2 24 6 24 8
The maximum displacement is obtained at the free end, i.e., at x = 0
P L4
-Elv=C 2 or v=---
8 EI

Case-2 : Statically equivalent loads

As explained in section 7.1, loads PB == P Land M B == P L2 ,which satisfy static

2 8
equilibrium conditions, are used as equivalent loads.

Downloaded From : www.EasyEngineering.net

204 FINITE ELEMENT ANALYSIS

At any section, distance x from B, from simple theory of bending

. moment M = (PL)
BendIng -
2
pe
.x - -
8
- = - EI - -
dx 2
(d 2V)
. dv p L x 2 P L2 x
IntegratIng, - EI- = - - - - - + C]
dx 4 8

ww The constant C] is evaluated from the boundary condition dv = 0 at x = L

w.EThus, C] = - P L3 and _ EI dv = P Lx 2 _ P L2 X _ P e
8 dx 4 8
pLx 3 pL2 x 2 pex
8

asy
Integrating again, - EI v = - - -
12 16
- - - + C2
8

E
The constant C 2 is evaluated from the boundary condition v = 0 at x = L

Thus, C 2 = -pL4) S L4
- .(-4+3+6)=-P-
( 48 48 ngi
n eer
The maximum displacement is obtained at the free end, i.e., at x = 0
-SpL4

ing
- EI v = C2 or v=--'----
48 EI
It can be seen that the displacement v in this case is different from that of
case-1 0

.ne
Case-3 : Consistent loads
pL pL2
Loads PB =-andM B = - - ,based on energy equivalence, as explained
2 12
later are used to represent distributed load.
t
At any section, distance x from B, from simple theory of bending
2
. moment M = (PL)
Bendmg - pe
.x - - d v
- = -EI-
2 12 dx 2
. dv P Lx 2 p ex
IntegratIng, - EI - = - - - - - + C]
dx 4 12

The constant C] is evaluated from the boundary condition dv =0 at x = L

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 205

Thus C = - P L3 and _ EI dv = P L x 2 _ P L2 X _ P e
'1 6 dx 4 12 6

pLx 3 pL2 x 2 pL3 x

Integrating again, - EI v = - - - - - - + C2
12 24 6
The constant C 2 is evaluated from the boundary condition v = 0 at x = L

p L4] 3 P L4 pC'
Thus, C 2 = - .(-2+1+4)=--=-

ww [ 24 24 8

The maximum displacement is obtained at the free end, i.e., at x = 0

- EI vw.E = C2 or v=--
_p L4
8E1

asy
It can be seen that the displacement 'v' in this case is identical to the
correct value represented by case-I and hence consistent loads. based on

E
work or energy equivalence. are preferred in FEM

ngi
The shape functions used to define displacement, in natural coordinate
system, over a finite element can also be used to calculate nodal loads

n
vector consistent with the loads distributed over an edge or a surface of

eer
an element. These consistent loads are calculated for different types of
distributed loads (on edge, area or volume), as explained below.
(a) Consistent nodal loads corresponding to distributed body forces are
given by
ing
{PB }= nNY{X}dv
v
.ne
where {X} indicates distributed load over the volume of the body
For a beam element with uniformly distributed self weight,
[N] = [L12(3 - 2L1) L12L2 L22(3 - 2L2) -L1Ll]
t
{p.l= bh I[NY {pldL=(b ~i L lIn
(b) Consistent nodal loads corresponding to distributed surface or traction
forces are given by

{ps}= nNf {T}ds

where {T} indicates distributed load over the surface of the body

Downloaded From : www.EasyEngineering.net

206 FINITE ELEMENT ANALYSIS

For a beam element with uniformly distributed pressure load {p}, in

units of force per unit .length (w), assuming that the load is uniformly
spread across the width 'b' of the cross-section,

Jpds=b JpdL= JWdL

S L L

Thus,

ww {PS }= b ~Nr
L
{p}dL

w .Ea bLp
12
6
L
6
wL
--
12
6
L
6

7.3
syE
CONSISTENT LOADS FOR A FEW COMMON CASES
-L -L

(a) Load on a beam ngi

1.
nee
Beam of length 'L' with a concentrated load 'P' at distance 'a'

rin
from node 1 (r = aiL)

2
g.n
p) = P (l - 3r + 2r\
M) =-P L (r- 2r2 + r 3);
P2 = P (3r2 - 2r3)

M2 = -P L (- r2 + r)
et
For the particular case when the load is at the center of the beam,
r= ~ and

P P
p)=- P2 =-
2 2
M) =-PLI 8 M2 =+P LI 8

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 207

2. Beam of length 'L' with a uniformly distributed load, w. Let P =

w L be the total load
wi
~1--------------~2
I
p_P_wL P wL
1- - 2 P2 = - = - -
2 2 2
-PL M _PL
M I =--

ww 3.
12 2 - 12

Beam with a linearly varying load (0 to w). Let P = w L / 2 be the

w.E total load

P-~
1- asy
E
10
-PL
M I =--
15 ngi
(b) Load along an edge of a plate
nee
Moments at nodes are not relevant here, since rotations of nodes are not
treated as DOFs.
rin
4.
'w' along an edge. Let P = w L be the total load
'--______---'1 w
g
Plate (modeled with CST elements) subjected to uniform pressure

.ne
P
PI =-
2 "
I

P2 =-
P
2
2
t
5. Plate (modeled with CST element) subjected to linearly varying
wL
pressure (0 to w) along an edge. Let P = - - be the total load
2
~w
2

_2P
P2-
3

Downloaded From : www.EasyEngineering.net

208 FINITE ELEMENT ANALYSIS

6. Plate (modeled with LST elements) subjected to uniform pressure

'w' along an edge. Let P = w L be the total load
w I I
1 3 2
PI=~' p-p. P =2P
6' 2- 6 '
3 3

Note: Corresponding statically equivalent nodal loads would be PI = ~;

ww P 2 = P and P3 = P, distributing equally P over each half of

4 2 2

w .Ea
7.
beam (I -3 & 3-2).
Plate (modeled with LST element) subjected to linearly varying
wL
pressure (0 to w) along an edge. Let P = - - be the total load.
2

syE-========:::=J
1 2 2
w

ngi
PI=O;
P
P2 =3;
2P
P3 = 3

r
surfuce, w. Let P = w A be the total load.
3

ing
~ .ne
9.
1

At comer nodes, PI = P2 = P3 =-
P
3
2

6-noded Triangular Plate with uniform pressure normal to the

t
surface, w. Let P = w A be the total load.
3

~
1 4 2
At comer nodes, PI = P2 = P3 = 0
P
At mid-side nodes, P4 = P5 = P6 =-
3

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFI lIENCING SOLUTION 209

10. 4-noded Quadrilateral Plate with uniform pressure normal to the

surface, w. Let P = w A be the total load.
3

ww At comer nodes, PI
P
= P2 = P3 = P4 ="4

w.E
II. 8-nodcd Quadrilateral Plate with uniform pressure normal to the
surface, w. Let P = w A be the total load.

asy
3

E 8

ngi
At comer nodes, PI n
= P2 = p) = P4 = - -
P
12 eer
At mid-side nodes, Ps = P6 = P7 = P8 = ~
3 ing
7.4 ASSEMBLING ELEMENT STIFFNESS MATRICES
.ne
Solution of any practical problem by finite element method involves a very
large number of simultaneous equations and hence, computer memory needs to
be effectively utilised. The following techniques are therefore utilised to
effectively use the available computer memory.
t
Element stiffness matrices are symmetric. Therefore, assembled stiffness
matrix of the entire structure is also symmetric. Storing half the matrix is hence
adequate. But, for storing even half of n x n stiffness matrix allocation of as
many memory locations for the variable K(n, n), no saving of computer memory
required.
(a) Storing a banded matrix - Stiffness matrix is usually a banded matrix,
in addition to being symmetric, depending on the nodal connectivity of
the elements, as explained in Fig. 7.3 for the simple case of a
rectangular plate modeled with 4-noded quadrilateral elements.

Downloaded From : www.EasyEngineering.net

210 FINITE ELEMENT ANALYSIS

Element-I is linked to nodes I, 2, 7 and 6. Hence, contribution of

element-I in the assembled stiffness coefficients will be found in the
rows and columns associated with these four nodes only. Likewise, each
of the other elements will have contributions in the rows and columns of
the assembled stiffness matrix corresponding to their respective four
comer nodes only. None of the elements 2-8 are linked to node-I.
Hence, assembled stiffness mat~ will not have non-zero terms in (i, j)
positions, where row 'i' and column 'j', correspond to the degrees of
freedom associated with nodes 3,4,5, 8, 9, 10, 11, 12, 13, 14 and IS.

ww While assembling such a stiffness matrix in the computer, significant

number of memory locations can be saved by avoiding entries

w.E corresponding to these zero values.

2 3 4 5

a0 sy 0 G 0
7 8 9

EG n [2] 0
6 IO

0
1l 12 gin 13 14 15

eer
FIGURE 7.3 Numbering of nodes in a model

in
Half Bandwidth 'b' = (Max. node number difference for anyone
element + 1) x Number ofDOF per node
.Storing a symmetric banded matrix requires n x b memory locations asg.n
against n x n memory locations required otherwise. It is a rectangular
matrix of n rows and b columns, first column representing diagonal
elements of the square matrix. The algorithm in the program properly
e
identifies elements of the matrix. For example, ~,4 is stored as ~,I ; ~,5
t
as ~,2; ~,6 as ~,3 and so on till all non-zero elements are covered. In
general, kn,m of the assembled matrix is identified by kn,m-n+l in the
banded matrix. The banded matrix of order 12 x 7, as stored in this way,
is represented below with diagonal element indicated by 'd'. Since the
symmetric half of the band matrix has width 'b' in the first few rows
and reduces to I in the last row, the banded matrix wiII have some zero
values in the last few rows as shown in Fig. 7.4.

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 211

d x x x x x d x x x x x
d x x x x x Os beyond d x x x x x
bandwidth
d x x x x x d x x x x x
d x x x x x d x x x x x
d x x x x x d x x x x x
d x x x x x d x x x x x

ww d x
d
x
x
x
x
x
x' x
x d
d
x
x
x
x
x
x
x
x
x
0

w.E
symmetric d x x x d x x x 0 0
d x x d x x 0 0 0
x x

asy
d d 0 0 0 0
d d 0 0 0 0 0

E
Half of a symmetric matrix

ngi
Storing in band matrix form

FIGURE 7.4 Storing half of a banded matrix in different forms

nee
(b) Minimising bandwidth of stiffness matrix - Bandwidth of the
assembled stiffness matrix can be minimised by renumbering node
numbering sequence of a finite element model.
As an example, let us consider a ring modeled by a number of curved
beam el~ments. Case-I and case-2, shown in Fig. 7.5, follow two rin
different node numbering schemes.
g .ne
t

Case 1 Case 2
FIGURE 7.5 Minimising bandwidth by renumbering nodes of a simple ring

Downloaded From : www.EasyEngineering.net

212 FINITE ELEMENT ANALYSIS

In case-I, Maximum node number difference = 9 - 1 = 8

and half bandwidth b = (8 + 1) x No. OOF/node
while, in case-2, Maximum node number difference = 3 - 1 or 5 - 3,. =2
and b = (2 + 1) x No. ofDOF/node
For all other elements, node number difference is less than or equal to
the above values.
Let us consider one more example of a plate with 2 OOF per node. It is

ww shown with two node numbering sequences as case-l and case-2 in

Fig.7.6.

w.E 23 24 25 26 27 28 29 30 31 32 33

12 13 14 15 16 17 18 19 20 21 22
3

'I
6 9 12 15 18 21 24 27 30 33

Is Is In 114Jnr0 1"1'61'9132
a syE
1234567891011
Case-l
1 4 7 10 13 16 19 22 25 28 31
Case-2

ngi
Figure 7.6 Minimising bandwidth by renumbering nodes of a plate

In case-I, Half bandwidth, b = (24- 12 + 1) x 2 = 26 for top left element

nee
In case-2, Half bandwidth, b = (6 - 2 + 1) x 2 = 10 for the same element
The bandwidth of this element happens to be equal to the bandwidth of

rin
any other element in the model and, hence, represents the maximum
bandwidth for these two cases.
In this simple model, total number of OOF = 33 x 2 = 66.
g .ne
If entire stiffness matrix is stored, computer memory required = 66 x 66
If banded stiffness matrix is stored,
computer memory required = 66 x 26 in case-l and 66 x lOin case-2.
Several algorithms are developed for selecting appropriate node
numbering sequence in an actual problem so as to minimise bandwidth.
t
Some of thtUn are used, even without the knowledge of the end user, in
many general purpose commercial software.
(c) Skyline method of assembly - Bandwidth may not be the same for all
the elements in a practical problem with irregular geometry. Then
within the maximum bandwidth, b, used for computer memory
requirement starting zeroes of all columns can be avoided by following
a different method of assembly. Thus, requirement of computer memory
can be minimised. This method also avoids the need to store the zero
values of the last few rows of the banded matrix. Here, the banded
matrix K is stored as a column vector A, with columns of banded matrix

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 213

stored one after the other. Each column is of variable length and starts
with the first non-zero value in that column. Intermediate zeroes need to
be stored. Another array variable (10) stores the sequential address of
the last term of each column (diagonal elements) so that the program
can identify location of each stiffness coefficient in the column vector.
The elements are identified as shown in the following example, for a
simple 7 x 7 stiffness matrix. In actual practice, the saving of computer
memory will be very large, since the stiffness matrix wi1\ be of a very

ww large size. The square and banded stiffness matrix

kll k12 0 k14 0 0 0

w .Ea
[K]=
k22 kn
k33
k24
k34
k44
k25
0
0
0
k36
k46
0
0
0

syE
symmetric k55 k56
k66
k57
k67

is stored in a single column matrix as ngi k77

{K} = [k ll kJ2 k22 kn k33 k14.k24 k34

o nee
0 k55 k36 ~ k56 ~ k57 ~7 k77]T
~4 k25

rin
Note that starting elements k\3' k 15 , k 16 , k26 , k17, k27, k37 and ~7 in 3 , 5th,
61h and 71h columns having value zero, are not stored while elements k35
Td

and ~5, also having value zero, are stored since they are included in
between elements of a particular column with non-zero value. g.n
Another vector {IO}, ith element of which corresponds to the position of
the ilh diagonal element in [K], helps in identifying each element of {K}
with its corresponding position in [K]
et
{IO} = [1 3 5 9 13 17 20]T
Position of any element in the jlh column of the original stiffness matrix
is identified by its position w.r.t. the diagonal element in the j'h column
represented by IOU). Position of element in ilh row of this column (i ~j)
is given by [ ID(j) - G-i) ].
Ex: Position ofk33 in the array {K} = 10(3) = 5
Position ofkn in the array {K} = 10(3) - (3 - 2) = 5 - 1 = 4
Position of~6 in the array {K} = 10(6) - (6 - 4) = 17 - 2 = 15

Downloaded From : www.EasyEngineering.net

2:1.4 FINITE ELEMENT ANAL YSIS

7 .5 AUTOMATIC MESH GENERATION

One of the difficult tasks in the analysis of a component by finite element

method is the need to discretise the component into a large number of elements
connected by nodes and specitying the nodal coordinates as well as element
connectivity. In most commercial software, this is done by the program from
the description of the component geometry as a solid model either as a

ww combination of some primitive shapes or through key points, lines connected by

key points, areas connected by lines and volumes specified by the enclosing

w.E
areas. Type of element (1-D truss, ] -D beam, Plane stress element, plate
bending element, thick shell element, ... ) that the component closely matches in
its behaviour, also needs to be specified by the user.

asy
An automatic mesh generation program generates the locations of the node

En
points and elements, labels the nodes and elements and provides the element
node connectivity relationships. A set of nodes is identified to represent the

gin
component, based on the relative dimensions of the component, choosing a
minimum number of nodes across the smallest dimension, which varies with the

eer
software. In many programs, the user can also specity the minimum size of the
elements to be generated and the ratio of sizes of adjacent elements, for

ing
generating mesh which varies from coarse to fine, in the areas of stress
concentration.

.ne
Two methods are explained here for forming elements, from the given set of
nodes.
In the Tessellation method, program starts connecting user defined nodes t
starting with an arbitrary point on the boundary. It creates a simplex element
using the neighbouring nodes which give the least distorted element shape.
Then, it proceeds to form the next element. An example of Tesselation method
is given here. Fig. 7.7 (a) gives nodal set of a component while Fig. 7.7 (b),
Fig. 7.7 (c) and Fig. 7.7 (d) show one simplex element formed by joining three
nodes. The element in Fig. 7.7 (d) is the least distorted element and is
finally selected before generating other elements (shown by dotted lines) in a
similar way.

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 215

x x
x x

x x x x

~,
x x

x x
x
(a)
(b)

ww x x
x x
x

w .Ea
(c) syE (d)

ngi
FIGURE 7.7 Mesh generation by Tessellation method

nee
In the Dctree method, a three dimensional cube is assumed around the
object. If the object is partially occupied by the cube, it is subdivided into small
cubes and each cube is checked. If any cube is full (completely occupied by the

rin
object) or empty, then the cube is not subdivided further. It gives elements of
different sizes, irrespective of the stress distribution in the object and hence, is
not very much popular.
g.n
7.6 OPTIMUM MESH MODEL

Best possible mesh has to be used to obtain solutions as accurately as possible,

while minimising the requirement of computer resources. In many cases, it can
et
not be decided before the analysis is completed. In time-dependent iterative
problems, mesh refinement between different steps becomes very important to
ensure convergence of the solution. There are many mesh refinement methods
available.
(a) Mesh refinement method or h-method - Refines the element size
based on solution gradients.
(b) Mesh movement method or r-method - Grid points are moved around
(mesh redistribution) to provide clustering in certain regions, based on
error indicators.
(c) Mesh enrichment method or p-method - Refines degree of
polynomial (interpolation function) based on user-specified error

Downloaded From : www.EasyEngineering.net

22.6 FINITE ELEMENT ANALYSIS

tolerance. It is particularly important when singularities are encountered,

such as near crack tips in fracture analysis. Side or interior nodes are not
installed physically, but higher order modes of the polynomial
corresponding to these nodes are combined with corner nodes by means
of static condensation (similar to elimination of rows associated with
DOFs corresponding to the specified nodal displacements), so that
compatibility with adjacent elements is not affected.

ww 7.7 GAUSSIAN POINTS & NUMERICAL INTEGRATION

Closed form solutions for integration, associated with many FEM problems, are

w not possible using digital computers. Hence, numerical integration is generally

.Ea
used in FEM. Integration of a simple function f(x) over the range (rI. r2)
amounts to calculating area under the curve. In numerical' integration, this is
approximated by the sum of areas of a few rectangles (products of functions

syE
values and the local range or weight) at a few sampling points as shown. This
sum naturally approaches true value of the function area as the number of
sampling points increase. Gauss quadrature method of numerical integration is

I ngi
proved to be the most useful in finite element applications.
n

1= ff(r)dr=Lw i f(rJ
-I i=1
nee
with l:w, =1-(-1)=2

rin
where WI is the "weight" or range associated with the ith sampling point and n is
the number of sampling points within the element.

f(r)
g.n
et

For I-point integration, rl =0.0 WI =2.0

1
2-point integration, rJ, r2 = ± 0.5774 (or.± J3)
3-point integration, r), r3 =:!; 0.7746 wI. W3 = 0.5556
r2 = 0.0 W2 = 0.8889

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 217

4-point integration, rj, r4 ± 0.8613

= WI. W4 = 0.3479
r2, r3 = ± 0.34 W2, W3 = 0.6521

etc ..

1----
X

ww o

w.E
a
In a similar way, for double integration involving two shape functions rand s,

1= J
-1-1
I I

syE m n

Jf(r,s )dr ds == I I w, WJ f(r" sJ

1=1 J=I

ngi
and for triple integration involving three shape functions r, sand t,

nee
1 1 1 I m n
1= J J Jf(r,s,t)drdsdt== I I I w i Wj wk f(ri,Sj,t k )
-\-1-1 i=1 j=1 k=1

Since an arbitrary 2-D quadrilateral element is mapped into a square element

and an arbitrary 3-D solid element is mapped into a cube in natural (non- rin
dimensional) coordinate system, m = nand / = m = n are commonly used.
(a) Integration in Natural coordinate system g .ne
In terms of non-dimensional shape functions (area coordinates),

1= J Jf(L p L 2, L3)dA ==
A
tI~
i
wi f(L\, L 2, L'3) t
(i) For n = 1 (3-noded triangle), Gaussian point of integration is at
the center '0' of the triangle given by the coordinates LII = L21 =
L3 1 = 1/3 with the associated weight WI = 1

Downloaded From : www.EasyEngineering.net

218 FINITE ELEMENT ANALYSIS

(ii) For n = 2 (6-noded triangle), Gaussian points of integration are at

the mid-points of the three sides of the triangle given by the
coordinates

ww A

w.E
asy
En
and
gi
The associated weights are nee
(c)
rin
For a quadrilateral, with two Gaussian points along each coordinate, 2 x

g.n
2 points of integration are obtained by extrapolation of 1-0 two-point
values to 2-D in natural coordinates 1; (± 0.5774) and TJ (± 0.5774) as :
1 (-0.5774, -0.5774); 2 (+ 0.5774, -0.5774)
3 (+0.5774,+0.5774) and 4 (-0.5774, + 0.5774)
The associated weights with each point are Wi = Wj = 1.0
e t
i.e., the function value calculated at the above Gaussian points IS
associated with a quarter of the square area.
Similarly, with three Gaussian points along each coordinate, 3 x 3
coordinates and weights are obtained as extensions of 1-0 three-point
values
1; = -0.7746 at points 1, 4, 7 with weights Wi = 0.5556
= 0.0 at points 2, 5, 8 with weights Wi = 0.8889
and = + 0.7746 at points 3,6,9 with weights Wi = 0.5556

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 219

Similarly,
T) = -0.7746 at points 1,2,3 with weights w, = 0.5556
= 0.0 at points 4, 5, 6 with weights Wi = 0.8889
and = + 0.7746 at points 7, 8,9 with weights w, = 0.5556

a!rr
•
4 3
. . ~

ww I 2

w.E 2 x 2 Gaussian integration points

. llt
~
.
asy I 7

4
9

E
5 6
.I .
2

ngi
3

3 x 3 Gaussian integration points

In finite element method, displacement over an element is assumed by nee

an algebraic polynomial and integration of terms of [8] matrix for
evaluating element stiffness matrix is carried out term by term, each
rin
term being a product of different coordinates in various powers (in
general, of the form IJ, .L~.l;k .L~. Some mathematicians have derived
simple formulae for such integrations.
g .ne
(i)
L
JL~.qdL=L p!q!
(P+q+l). t
Ex:
L
JL( L2 dL=-;6 JL( dL=-3
L
L
2 L

(ii)
AJJL~.Lq.L'k dA= (p+q+r+2).
J
2A.p!q!r! (FeJippa)

Ex: L L L dA= 2A.l!.l!.l!= 2A =~

AJJ ( 2 3 5! 120 60

Downloaded From : www.EasyEngineering.net

220 FINITE ELEMENT ANAL YSIS

2A.3!.0! .O! 2Ax6 A

JJL~ dA= 5!
=--=-
120 10
A

2A.2!.0!.0! 2Ax2 A
J JL~ dA= =--=-
A
4! 24 6

(iii)
JIJ'rL~.L~.Uk.VJ dV = (p+q+r+s+3)!
v
J
6V p!q!r!s! (Clough)

ww Ex:
v
1 2 3 4
6V = ~
JIJL L L L dV = 6V I !7!I !11 11 = 5040 840

w.E IJJrL dV=6V2!0!0!0!

v
2
I 5! 120
=6Vx2=~
10

7.8 asy
MODELLING TECHNIQUES

En
Finite element method does not give a unique solution for any problem. The
accuracy of solution depends on many aspects like modelling of the actual

gi
component, number and type of elements used, approximation of loads and

nee
boundary conditions, solution techniques, etc .. It is for this reason that design
validation of products by the statutory safety codes of many countries is not
based on FEM results. Therefore, the engineer who uses this method (or uses

rin
any general purpose software based on this method) should check for the
correctness of the results with the expected trend Oat select locations. In this

g.n
chapter, important aspects of modelling and boundary conditions are discussed.
An element library is created, in each FEM based software, with each
element assumed to have a particular type of deformation. Appropriate types of
elements are selected to represent the component, based on the dimensions of
e
the component in different directions as well as on the nature of deformation of
the component.
t
For example, if a component has very small dimensions in the transverse
directions compared to its length, it can be modeled by I-D elements. In
addition, these I-D elements may be treated as axial ioaded elements (truss
elements or torsion elements), laterally loaded elements (beam elements or pipe
elements), depending on how close behaviour of the component is to the
behaviour assumed for these elements. It may be recalled that most of the
trusses analysed do not have pin joints, but their resistance to bending. is
negligible. A water tank, shown in Fig. 7.8, is an example of one dimensional
idealisation when it is analysed for wind loads or seismic loads. In this case,
nodal displacements normal to the axis alone are significant. Stresses aroun<l thp,

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION

circumference at any particular section have to be obtained from beam theory

(0 = M y / I), based on the bending moment obtained from this analysis and
geometric properties at that section.

ww
w.E
asy
E I
I
I
I ngi
nee
I I
I
I
I
I
I
I
I
I
I
I

rin
I
I
I
I
I
I
I
I
I
I

g
I
I
I
I

.ne
I I
I
I
I
I

FIGURE 7.8 Modelling of a water tank for dynamic analysis

t
Similarfy, 2-D elements with very "mall thickness compared to the other two
dimensions may be modeled with plane stress, plane strain, plate bending or
thin shell elements depending on the shape of the component as well as the
loads and boundary conditions applicable to the component.
Component with equally significant dimensions in all the three directions
can be modeled with 3-D solid elements or thick shell elements depending on
whether the component is subjected to significant bending deformation or not. It
may be noted that 3-D solid element wilt also have bending deformation
included but of a lesser degree (usually first order) whereas shell element has a

Downloaded From : www.EasyEngineering.net

222 FINITE ELEMENT ANALYSIS

more significant bending deformation (displacement modeled by a cubic

polynomial) as well as restraining slopes independently (simple supports or
fixed supports).
The number of nodes and elements required for modelling a discrete
structqre leave very little flexibility or ambiguity to the engineer. But, the
number of nodes and elements used in the analysis of a continuum model leaves
lot of flexibility to the analyst. The number of elements should be
commensurate with the criticality of the component or the desired accuracy of

ww results and is also dependent on the computer time and memory available for
this analysis. A lot of work is done to assess whether using a large number of
lower order elements is preferable or using a small number of higher order

w.E
elements is preferable. In addition, aspect ratios and included angles of 2-D and
3-D elements affect the results significantly. Their effect is demonstrated
through the following two examples. Rigid guidelines applicable to all kinds of

asy
problems can NOT be specified. These examples are only meant to highlight
dependence of results on various parameters.
Example 7.1
E ngi
A simple beam subjected to bending moment is simulated through a varying
load on the ends of a small rectangular plate. It is analysed, with the boundary

nee
conditions u = 0 at all nodes along x (neutral axis) and along y (symmetry) and
v = 0 at the origin (for suppressing rigid body modes), using six different mesh

rin
models. Displacement u at A and v at B obtained using FEM are tabulated
below.
y

g .ne
t
y y

~--~L-----~--~~X X
Case I Case 2

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 223

~--------~~----~X X
Case 3 Case 4

ww y y

w .Ea
L-~~"-~~~~~~-'X

C~5
syE C~6
X

ngi
Element Number of FEM Value I Exact Value
Case
type Nodes Elements u at A vat B
1 CST 12 12 0.84 0.812
2
3
4
CST
CST
CST
12
15
18
12
16
24
nee
0.812
0.778
0.940
0.916
0.825
0.951
5
6
CST
LST
35
9
49
2 rin
0.946
1.0
0.960
1.0
Example 7.2
g.n
A square plate in x-y plane simply supported on ail four sides and subjected to a
uniformly distributed normal load p is analysed using plate bending elements. A
quarter plate is modeled taking advantage of symmetry about x and y axes. The
results obtained at the center of the plate (C) are tabulated below with
et
quadrilateral elements (cases 1-3) and triangular elements (cases 4-7).

Downloaded From : www.EasyEngineering.net

224 FINITE ELEMENT ANALYSIS

y y

~--~L---~~--~~X X
Case 1 Case 2

ww y

w.E
a syE
""""-_ _ _ _ _....;;:0,......::;.._ _.......

Case 3
X
Case 4
X

ngi y

nee
L-~~'-~~~~~-+X
rin X

Case->
Element type
Case 5
1
Quad Quad
2
Quad
3 4
Triangle
Case 6
5
Triangle
g
6
Triangle
.ne
7
Triangle
No. of nodes
No. of elements
Displacement, w
at C (FEM Value
9
4
1.07
15
8
1.045 1.02
25
16
9
8
0.966
15
16
0.989
25
32
0.995
25
32
0.996 t
IExact Value)

7.9 BOUNDARY CONDITIONS FOR CONTINUUM ANALYSIS

Another important choice is to select the region of analysis. If the entire

component is not modeled, care should be taken to apply suitable end
conditions to simulate the true situation. For example, while analysing a nozzle
between a cylindrical shell and a fluid pipe, axial load across the cross section
due to fluid pressure need to be applied. Also, reasonable length of the
component has to be included in the model around the region of geometric
discontinuity to ensure that the results of analysis in the region of discontinuity

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 225

are not adversely influenced by the imposed end conditions. Some design codes
indicate guidelines for the same.
This aspect is explained through the example of a long cylindrical shell, such
as a boiler drum, shown in Fig. 7.9. The cylindrical shell is made up of two
parts with a circumferential joint and has a mismatch of radius of the two parts
by 0 at the joint. This joint has been analysed to calculate the stresses generated
due to this change of radius. Length of the cylinder considered on either side of
the joint influences the stresses at the joint. Analyses carried out with three

ww
different models (L) have given th~e different results. The model in which
uniform stresses are observed at the two ends is considered as the most

w.E
appropriate, since stresses without this discontinuity are uniform along the
length and the effect of any local discontinuity should vanish beyond some
distance.

14 asy
L ~I Offset ~ L ---1
TI ~ E
ngi FiT
R
nee 1
__~__________________________________________________________
R+o

rin
,,
,
g .ne
~\ ~
~
------
1
--_ ......

,i
f
\\
I'
I \\ "

\ "--
\

\
\;
~
t
\
<JJ \
CIl
QJ \

b
CIl ~ ~
L \ L,
,
L, \ L,
,
\
i
0
\
\

FIGURE 7.9 Effect of size of model around a discontinuity

Downloaded From : www.EasyEngineering.net

226 FINITE EI:EMENT ANALYSIS

(a) Symmetric boundary conditions

In some components with one or more lines or planes of symmetry,
modelling the entire component for analysis will result in waste of time
without any additional information regarding displacements or stresses
as compared to the analysis of a symmetric part of the component.
Boundaries of this model along these lines/planes of symmetry are
represented by a suitable boundary condition to indicate that the model
is only a part of the component. Displacements of corresponding nodes

ww on either side of the line of symmetry in the direction normal to the line
of symmetry are equal and opposite. Hence, on nodes along the line of
symmetry, displacement normal to the line of symmetry is zero. A

w.E rectangular plate with a circular hole at the center, shown in Fig. 7.10,
is a typical example of symmetry along two lines and hence, only a
quarter of the plate can be modeled for analysis, saving time and effort.

asy i
y

E ngi
--- - ------0---------
i

_--+x

nee i
!

rin i
i
Actual plate

g.n
v = 0 at all points
along this line
e
u = 0 at all points
along this line t
ISymmetry about X-axis I Symmetry about Y-axis

• ~. n"~L---''---
" .11 p..
along this line "
_ _ _ _....J1

v = 0 at all points
along this line

Quarter plate model - Using symmetry about X and Y axes

FIGURE 7.10 Symmetric boundary conditions

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INflUENCING SOLUTION 227

It should be noted that .\ymmetry should exist not only in the

geometry of the component bllt also in the loads ami boundary
C(}Iulitions of the component, so that results of one part of the
component are applicable to the remaining parts of the component. A
few such models, where symmetry can be used in modelling the
component for analysis, are shown below.
This aspect is explained through the example of arch of a factory gate,
Fig. 7.11. One half of the arch about the vertical line of symmetry can
be modeled, if the arch is analysed for a symmetric load such as self

ww weight (Case-a). But the model will not be adequate if it is to be

analysed for wind load, say from right to left (Case-b).

w .Ea
syE
Case (a) SymmetrIc arch with
symmetric (gravity) load
Case (b) Symmetric arch \\ 1Ih
unsymmctnc (wind) load

ngi
FIGURE 7.11 Dependence of model on geometry and loads

nee
Symmetry of a component to be analysed need not be limited to the
coordinate axes. If an octagonal structure such as a chimney is to be
analysed for 2-D heat conduction through its wall, symmetry can be

rin
lIsed to model just 11161h of the cross section as shown in Fig. 7.12.

g.n
et
Actual geometry Model for analysis

Figure 7.12 Model of a c~lmney for 2-D heat conduction

(b) Cyclic or Sector symmetry

In rotary components like turbines, fans, compressors, '" one sector
covering hub, one blade and rim can be modeled and the boundary

Downloaded From : www.EasyEngineering.net

228 FINITE ELEMENT ANALYSIS

condition will be different from that of the symmetry boundary

condition. Displacements of nodes along one boundary (A-A) of the
sector in cylindrical coordinate system are equal to the displacements of
corresponding nodes on the other end (B-B) of the sector. The example
of a fan, shown in Fig. 7.13, explains the cyclic symmetry condition.

ww
w.E L_---;---1i
"
Hub

asy
En
gin
eer
ing
FIGURE 7.13 Cyclic symmetry of a fan

7.10 TRANSITION ELEMENT

.ne
The purpose of any analysis is to evaluate stresses at all points in a component
when subjected to external or internal loads. Uniform mesh model is adequate
when the stress variation in the component is small. However in areas of
structural discontinuity or localised loads of high intensity such as thermal
t
stresses during welding, uniform mesh model may not be appropriate. Such
situations can be effectively modeled either by choosing fine mesh in the areas
of high stress or by using higher degree displacement model for elements in the
high stress areas. The latter option will require less computer memory and time.
Compatibility conditions are not satisfied on the common edges if elements
with linear displacement formulation (with 2 nodes along the common edge) are
used on one side and elements with quadratic displacement formulation (with
more than 2 nodes along the common edge) are used on the other side.

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 229

0
11
4 3

0 [TI ~ 0
5 2

ww I
x
2

w .Ea
FIGURE

syE
7.14 An example of a transition element

Transition elements, satisfying inter-element compatibility along the

ngi
common edges, are used to connect different types of elements. These elements
will not have same number of nodes on all their edges. For example, a

nee
quadrilateral element I, shown in Fig. 7.14, connecting linear displacement
elements 2, 3 and 4 with a quadratic displacement element 5 is one such
transition element. Displacement polynomials
2
u = al + a2 x + a3 Y + a4 x + a5l
v -- ~ + a7 x + as y + fu) x 2 + alO Y2 rin
g.n
an d
are used for that element, ensuring symmetry w.r.t. coordinate axes x and y. In

et
iso-parametric formulation, shape functions of this particular 5-node
quadrilateral transition element are given by
NI = ~(~-I)(I-11)
4

N2=~(~+I)(I-11)
4

N3 = (~ + 1)(1 + 11)
4

N = (1- ~)(1 + 11) .

4
4 '
We can have, in a similar way, transition quadrilateral elements with 6 or
7 nodes depending on the displacement function used in elements 2, 3 and

Downloaded From : www.EasyEngineering.net

230 FINITE ELEMENT ANALYSIS

4 surrounding the transition element I. The terms in the displacement

polynomial for each type of such elements will be different to ensure symmetry
w.r.t. coordinate axes. The same logic can also be extended to other 2-D and
3-D elements. Transition elements are not relevant for discrete structures.

7.11 SUBSTRUCTURING OR SUPER ELEMENT ApPROACH

In the finite element analysis of large systems, the number of equations to be

ww solved for an accurate solution will be very large. Computational cost of matrix
inversion is proportional to cube of the order of stiffness matrix. Thus, it is

w.E
cheaper to invert three square matrices of order 500 than inverting one square
matrix of order 1500. It also reduces computer memory requirement, since all
substructures are not si~ultaneously processed. Method of substructures can' be

asy
used to reduce the number of equations. In this method, the structure is divided
into a number of parts, called substructures or super elements, each of which

E
can be subdivided into a large number of elements. Each substructure is treated

ngi
as one large element with many interior and boundary nodes. Assembled
stiffness matrix of a substructure is rearranged to group displacements of all
boundary nodes (elements with suffix 'b') and displacements of internal nodes

n
(elements with suffix 'i') separately. Using static condensation procedure. this

eer
stiffness matrix is reduced to include modified contributions of boundary nodes
only, depending on the type of elements being used, as explained below.

ing
From the first set of equations [K II ] {u,} + [K ,b] {Ub} .ne
= {P,}, we can express
{U,} as
{u,} = [K,ir l ({ P,} - [K ,b ] {Ub})
t
Substituting these values of {u,} in the 2 nd set of equations
[Kb,]{u,} + [Kbb]{Ub} = {Pb},
l
we get [Kbl][Kllr ( {P,} - [K,b] rUb} ) + [Kbb]{Ub}= {Pb}
or
This can be written in a different notation as [K*] rUb} = {po}
where [K*] is the condensed stiffness matrix of the substructure, including
OOFs associated with boundary nodes only
and {po} is the corresponding modified load vector.

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 231

This process of condensation is carried out, in many software, by Gauss

elimination procedure. These reduced matrices of different substructures can be
assembled together to get the stiffness matrix of the complete structure. The
remaining procedure of applying boundary conditions and solving for the
unknown displacements and stresses is same. There can be many levels of
substructures, the highest level substructure consisting of ordinary finite
elements only.
A typical application of this procedure can be seen in the case of an aircraft,

ww
where fuselage (central body), nfain wings, tail wings, etc .. are meshed as
independent substructures. They are assembled together, after static
condensation of the stiffness matrix of each substructure, to analyse the entire

w.E
aircraft structure for the specific loads. This method saves considerable time and
memory of the computer.

7.1.2
a syE
DEFORMED AND UNDEFORMED PLOTS

One problem with FEM is the generation of a large output conslstmg of

ngi
displacement at all nodes and in alI active degrees of freedom. It is difficult to
scan for useful values from this large output to arrive at any meaningful
conclusion. To overcome this difficulty, many general purpose software include

nee
options for plotting deformed shape as well as iso-stress and iso-temperature
contours. Deformed shape of the compoftent is better appreciated when the

rin
same is superimposed on undeformed geometry, with or without node/element
numbers. Data errors related to load direction and area of application as well as

g
boundary conditions are often checked with these deformed plots.
Within the elastic limit, the displacements in a component are so small that
the deformed and undeformed plots coincide. Hence, nodal displacements are
.ne
usually multiplied by a factor so that maximum displacement at any point in the
element is about 20% of the component size or about 1 cm on A-4 size plot.
Thus, deformed plot forms a vital check on the analysis of the component, to
assess whether the results are sensible and meaningful or not. Deformed plot is
t
qualitative and not quantitative, since physical dimensions of the component
and nodal displacements are not plotted to the same scale.
Another visual anomaly is in the deformation of frames. Plots are generated
by the post processor of the software from the nodal displacements obtained in
the solution phase. Thus, the deformed plot of a beam member is a straight line,
generated trom its two nodal displacements, irrespective of the end conditions
(simply supported or fixed ends). Cubic displacement polynomial as welI as end
conditions on slope are not reflected in these plots. (Ref. Fig. 7.15).,

Downloaded From : www.EasyEngineering.net

232 FINITE ELEMENT ANALYSIS

------------------------ ~

(a) With ~all1pk' '1II'POJts (b) With fixed supports

FIGURE 7.15 Plot of deformation of beam

Note : Plots of defonnation of beam between nodes 1 & 3 as well as between

ww 2 and 3 in both the cases of end conditions are linear even through both of them
are cubic functions. Even the displacement value at node 3 appears to be of the

w.E
same magnitude in both cases, though the calculated values are different.

7.13
asy
SUMMARY

• Consistent loads, based on energy equivalence, are used in FEM to

E
represent distributed loads and give better results than equivalent loads

ngi
• Nodes, forming elements, are numbered to minimise maximum node
number difference in the elements. This affects bandwidth of assembled

•
stiffness matrix
nee
Many Software, based on FEM, store one half of banded, symmetric

• rin
stiffness and mass matrices to minimise computer memory requirement
Modelling a physical problem involves selection of proper types of

g
elements, from the element library of any commercial software, which
assume same displacement behaviour as the actual problem. Details, with
zero or less influence on the results, can be omitted to simplifY the .ne
•
physical model while ensuring a meaningful analysis with minImum
computer cost
Shape and size of elements influence the results significantly
t
• Use of symmetry in geometry, loads and boundary conditions help in
modelling a smaller part of the component, saving computer time and
memory
• Substructuring helps in the analysis of very large components by
reducing size of matrices
• Defonnation plots in many FEM software are qualitative (with magnified
displacements), plotting only displacements without indicating correct
slopes.

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INFLUENCING SOLUTION 233

OBJECTIVE QUESTIONS
I. A symmetric structure can be analysed by modelling one symmetric part
(a) depending on applied loads
(b) depending on boundary conditions
(c) always yes
(d) depending on applied loads & boundary conditions

ww
2. Anti-symmetric boundary condition along an edge of a 2-D structure
implies, applied loads are __ on either side of the edge

3.
w (a)
(c)
opposite

.Ea
equal and opposite
(b)
(d)
equal
unrelated
Sector symmetry boundary condition implies _ _ along two radial
edges of the sector
(a) syE
same radial displacements in cartesian coordinate system
(b)
ngi
same circumferential displacements in cylindrical coordinate
system
(c)
nee
equal and opposite radial displacements in cartesian coordinate
system
(d)
rin
equal and opposite circumferential displacements in cylindrical
coordinate system
4. Cyclic symmetry boundary condition implies _ _ along two edges of
g.n
et
the sector
(a) same radial displacements in cartesian coordinate system
(b) same circumferential displacements in cylindrical coordinate
system
(c) equal & opposite radial displacements in cartesian coordinate
system
(d) equal and opposite circumferential displacements in cylindrical
coordinate system
5. An octagonal section chimney with hot gases inside can be analysed
using _ model
(a) full section (b) one half of section
(c) one quarter of section (d) lISth of section

Downloaded From : www.EasyEngineering.net

234 FINITE ELEMENT ANALYSIS

6. Use of non-dimensional coordinates helps in

(a) numerical integration (b) displacement calculation
(c) stress calculation (d) strain calculation
7. Gaussian points are used for
(a) numerical integration (b) displacement calculation
(c) stress calculation (d) strain calculation
8. Quadrature means

ww (a)
(b)
calculation of area of element
calculation of element stress

w 9.
(c)

.Ea
(d)
numerical integration for getting stiffness coefficients
calculation of nodal displacements
Accuracy of stiffness matrix improves with
(a)
(b)
syE
more number of Gaussian points
more number of nodes
(c)
(d)
size of elements
shape of elements ngi
10. Sector symmetry and cyclic symmetry differ
(a) in the shape of sector edges
nee
(b) in the size of sector edges
rin
(c)
(d)
in radial displacements along two sector edges
in circumferential displacements along two sector edge!> g.n
II. Using symmetry condition __ ; but gives same solution
(a)
(b)
saves computer time
saves computer memory
et
(c) saves effort of data preparation
(d) all of them
12. Symmetry boundary condition about an edge is applicable when
(a) normal loads & normal displacements at nodes along the edge are
zero
(b) loads & displacements along the edge are zero
(c) normal loads & normal displacements at nodes on either side of the
edge are equal & opposite
(d) loads & displacements along the edge are same

Downloaded From : www.EasyEngineering.net

CHAPTER 7 FACTORS INlllll:NCING SOLUTION 235

13. A cantilever beam can be analysed as a plate with __ boundary

conditions
(a) Cartesian symmetric (b) sector symmetry
(c) cye! ic symmetry (d) cartesian anti-symmetric
14. Number ofDOF for 2-node cantilever and propped cantilever are
(a) 1,2 (b) 2,1 (c) 3,4 (d) 2,4

ww
15. Number of DOF for 3-noded simply supported beam and fixed beam are
(a) 1,2 (b) ~,3 (c) 3,4 (d) 4,2
16.
w.E
Small region of interest in a big component can be analysed using free
body end conditions
(a)
(c)
always true
never true asy (b)
(d)
sometimes true
depends on other data
17.
En
_ model of a rectangular plate with a circular hole at the center, and

18.
(a) full (b) gi
loaded uniformly along the four edges, is adequate for analysis
1/2 (c) 1/4

nee
(d) 1/8
_ model of a square plate with a circular hole at the center, and loaded
uniformly along the four edges, is adequate for analysis
(a) full (b) 112 (c) 1/4 (d) 1/8 rin
19. _ model of a square plate with a rectangular hole at the center (edges
parallel to the edges of the plate), and loaded uniformly along the fourg.n
edges, is adequate for analysis
(a) full (b) 1/2 (c) If4 (d) 1/8
e t
20. In statically equivalent loads, free end moment of a cantilever of length
'L' with uniformly distributed load of value 'p' is
(a) pL2/4 (b) pL2/8 (c) pL2/12 (d) pL2116
21. In consistent loads, free end moment of a cantilever of length 'L' with
uniformly distributed load of value 'p' is
(a) pL2/4 (b) pL2/8 (c) pL2112 (d) pL2116
22. In statically equivalent loads, end moment of a simply supported beam of
length 'L' with a concentrated load 'P' at the mid point is
(a) PLl4 (b) PLl8 (c) PLlI2 (d) PLl16

Downloaded From : www.EasyEngineering.net

236 FINITE ELEMENT ANALYSIS

23. In consistent loads, end moment of a simply supported beam of length 'L'
with a concentrated load 'P' at the mid point is
(a) PLl4 (b) PLl8 (c) PLlI2 (d) PLlI6
24. In statically equivalent loads, end moment of a simply supported beam of
length 'L' with a uniformly distributed load of value 'p' is
(a) pL2/4 (b) pL2/8 (c) pL2/12 Cd) pL2/16
25. In consistent loads, end moment ofa simply supported beam of length 'L'

ww with a uniformly distributed load of value 'p' is

(a) pe/4 (b) pL2/8 (c) pL2/12 (d) pL2//16

w.E
26. Consistent loads for a LST element with uniform pressure 'p' along an
edge of length' L' , at the two end nodes and mid-node are

asy
(a)
(c)
pLl2, pLl2, 0
pLl4, pLl4, pLl2
(b)
(d)
pL/3, pLl3, pLl3
pLl6, pLl6, 2pLl3
27.
E
The process of reducing number of mid-side or internal nodes before

ngi
assembling element stiffness matrices is called
(a) Gauss reduction (b) Jacobi reduction

28.
(c) Choleski reduction
n (d) static condensation

eer
Lengths of longest side and shortest side of a 2-D or 3-D element decide
the
(a) aspect ratio
ing
(b)
(c)
shape function
order of displacement polynomial .ne
29.
(d) included angle
Number of nodes along the side ofa 2-D or 3-D element decide the
(a) aspect ratio
t
(b) shape function
(c) order of displacement polynomial
(d) nature of deformation

Downloaded From : www.EasyEngineering.net

CHAPTER 8
DYNAMIC ANAL YSIS
(UNDAMPED FREE VIBRATION)

ww
w .Ea
Dynamics is a special branch of mechanics where inertia of accelerating masses

syE
must be considered in the force-deflection relationships. In order to describe
motion of the mass system, a component with distributed mass is approximated
by a finite number of mass points. Knowledge of certain principles of dynamics

ngi
is essential to the formulation of these equations.
Every structure is associated with certain frequencies and mode shapes of

nee
free vibration (without continuous application of load), based on the distribution
of mass and stiffuess in the structure. Any time-dependent external load acting
on the structure, whose frequency matches with the natural frequencies of the

rin
structure, causes resonance and produces large displacements leading to failure
of the structure. Calculation of natural frequencies and mode shapes is therefore
very important.
Consider ilh mass m, of a system of connected rigid bodies and the force g.n
components Fj G= 1,2, .. 6) acting upon it in three-dimensional space. If the mass
mj is in equilibrium at rest, then l:fj =0 .
If mass mJ is not in equilibrium, it will accelerate in accordance with
et
Newton's second law i.e., F·J = m·1 ii.J

The force (-mj.liJ) is called the reversed effective force or inertia force.
According to D' Alembert's principle, the net external force and the inertia
force together keep the body in a state of 'fictitious equilibrium'
i.e., J=
L(FJ - m ii 0 .
If the displacement of the mass mj is represented by oUJ G= 1,2, .. 6), then the
virtual work done by these force components on the mass mj in equilibrium is
given by

Downloaded From : www.EasyEngineering.net

238 FINITE ELEMENT ANALYSIS

D' Alembert's principle rewritten in the form,

8Wi = L Fj . 8uj - L (m iij) . 8uJ = 0 is a statement of virtual work
for a .~ystem in motion.
For a simple spring of stiffness 'k' and a lumped mass 'm' under steady state
undamped condition of oscillation without external force, the force equilibrium
condition of the system is given by
k u(t) + m li(t) = 0,

ww where, F, = - k u(t) is the reactive elastic force applied to the mass.

Displacement in vibration is a simple harmonic motion and can be

w.E
represented by a sinusoidal function of time as
u(t) = u sin rot

asy
where, ro is the frequency of vibration in radians/sec
It is more often expressed in 'f cycles/sec or Hertz (Hz) where ro = 21t f

Then, velocity
En
u(t)= ---(ou cos rot

and acceleration tl(t) = ---(0

gin
u sin rot = ---(0
2
u(t)

K.u(t) + m OCt) = (k - ro 2m) u(t) = 0

eer
In general, for a system with on' degrees of freedom, stiffness 'k' and mass
'm' are represented by stiffness matrix [K] and mass matrix [M] respectively.
Then, ([K] - ro 2 [M]) {u} = {O} ing
or ([Mrl[K] - ro [I]) {u} 2
= {O}
.ne
Here, [M] is the mass matrix of the entire structure and is of the same order,
say n x n, as the stiffness matrix [K]. This is also obtained by assembling
element mass matrices in a manner exactly identical to assembling element
stiffness matrices. The mass matrix is obtained by two different approaches, as
t
explained subsequently.
This is a typical eigenvalue problem, with ro 2 as eigenvalues and {u} as
eigenvectors. A structure with 'n' DOF will therefore have 'n' eigenvalues and
'n' eigenvectors. Some eigenvalues may be repeated and some eigenvalues may
be complex, in pairs. The equation can be represented in the standard form,
[A]{xL = A, {xL. In dynamic analysis, ro, indicates ith natural frequency and {Xli
indicates ith natural mode of vibration. A natural mode is a qualitative plot of
nodal displacements. In every natural mode of vibration, all the points on the
component will reach their maximum values at the same time and will pass

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 239

through zero displacements at the same time. Thus, in a particular mode, all the
points of a component will vibrate with the same frequency and their relative
displacements are indicated by the components of the corresponding
eigenvector. These relative (or proportional) displacements at different points
on structure remain same at every time instant for undamped free vibration
(Ref. Fig. 8.1). Hence, without loss of generality, {u(t)} can be written as {u}.

w~I"L-~_--I_+-
ww t=t l t=t 2
_ _ _ _":::""'=-_ _ _ _ _ _ _
t=t l t=t 2 ~
~

.Ea FIGURE 8.1 Mode shape

syE
Since {u} = {O} forms a trivial solution, the homogeneous sy~tem of
equations ( [A] - A, [I] ) {u} = {O}gives a non-trivial solution only when

ngi
which implies
([A]-A,[I])= {O},
Det ( [A] -),[1] ) = o.

nee
This expression, called characteristic equation, results in nth order
polynomial in A, and will therefore have n roots. For each ~, the corresponding
eigenvector {u}. can be obtained ITom the n homogeneous equations
rin
represented by ([K] - A, [M]) {u} = {O}. The mode shape represented by {u(t)}
gives relatives values of displacements in various degrees of freedom.
It can also be represented as g.n
where,
[A][X] = [X][A]
[A] = [Mr l [K]
[X] is called the modal matrix, whose ith column represents ith eigenvector
e t
{Xli
and [A] is called the spectral matrix with each diagonal element
representing one eigenvalue, corresponding to the eigenvector of that column,
and off-diagonal elements equal to zero.

8.1 NORMALISATION OF EIGENVECTORS

The equation of motion of free vibrations ([K] - 00 2 [M]) {u} = {O} is a system
of homogeneous equations (right side vector zero) and hence does not give
unique numerical solution. Mode shape is a set of relative displacements in

Downloaded From : www.EasyEngineering.net

240 FINITE ELEMENT ANALYSIS

various degrees of freedom, while the structure is vibrating in a particular

frequency and is usually expressed in normalised form, by following one of the
three normalisation methods explained here.
(a) The maximum value of anyone component of the eigenvector is
equated to 'I' and, so, all other components will have a value less than
or equal to ' I ' .
(b) The length of the vector is equated to 'I ' and values of all components

ww are divided by the length of this vector so that each component will have
a value less than or equal to 'I'.

w (c) The eigenvectors are usually normalised so that

.Ea
T
{uh [M] {uh = I and
T
{uh [K] {uh = AI
For a positive definite symmetric stiffness matrix of size n x n, the

i.e., syE
eigenvalues are all real and eigenvectors are orthogonal
{u}IT[M] {uh=O and {u}jT[K] {u}j=O \;f i:;t:j

8.2
ngi
MODELLING FOR DYNAMIC ANALYSIS

nee
Solution for any dynamic analysis is an iterative process and, hence, is time -
consuming. Geometric model of the structure for dynamic analysis can be
significantly simplified, giving higher priority for proper representation of

rin
distributed mass. An example of a simplified model of a water storage tank is
shown in Fig. 8.2, representing the central hollow shaft by long beam elements

g.n
and water tanks at two levels by a few lumped masses and short beam elements
of larger moment of inertia.

8.3 MASS MATRIX

Mass matrix [M] differs from the stiffness matrix in many ways:
et
(i) The mass of each element is equally distributed at all the nodes of that
element
(ii) Mass, being a scalar quantity, has !!!!!!.£ effect along the three
translational degrees of freedom (u, v and w) and is not shared
(iii) Mass, being a scalar quantity, is not influenced by the local or global
coordinate system. Hence, no transformation matrix is used for
converting mass matrix from element (or local) coordinate system to
structural (or global) coordinate system.

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 24:1

® 22
21
® 20
@
@ 19
@ 18

ww @
@)
17
16
15

w.E CD>
@)
14

asy ~II
13
12
11

E I ngi
@
®
CD
G)
10
9
8
7

I
I
I
I
I
0
0) nee 6

rin
I
I
I
I 0 4

g
I
I
6)

.ne
I
I 3
I
I
CD
t
I
I
2
I
I CD

Actual structure Idealised structure

FIGURE 8.2 Finite Element Model of a water tank for dynamic analysis
Two different approaches of evaluating mass matrix [M] are commonly
considered.
(a) Lumped mass matrix
Total mass of the element is assumed equally distributed at all the nodes
of the element in each of the translational degrees of freedom. Lumped
mass is not used for rotational degrees of freedom. Off-diagonal
elements of this matrix are all zero. This assumption excludes dynamic

Downloaded From : www.EasyEngineering.net

242 FINITE ELEMENT ANALYSIS

coupling that exists between different nodal displacements. Lumped

mass matrices [M] of some elements are given here.
Lumped mass matrix of truss element with 1 translational DOF per
node along its local X-axis

[M]= PAL[I
201
0]
Lumped mass matrix of plane truss element in a 2-D plane with

ww 2 translational DOF per node (Displacements along X and Y coordinate

axes)
I 000

w.E [M]= pAL 0

200
o
I
0
o

a syE
000
Please note that the same lumped mass is considered in each
translational degree of freedom (without proportional sharing of mass

ngi
between them) at each node.
Lumped mass matrix of a beam element in X-V plane, with its axis

nee
along x-axis and with two DOF per node (deflection along Y axis and
slope about Z axis) is given below. Lumped mass is not considered in
the rotational degrees of freedom.

[M]= pAL o
I 0 0 0
rin
g
0 0 0
2 0 0
o 0
I
o
0
0 .ne
nd
Note that lumped mass terms are not included in 2 and 4th rows, as
well as columns corresponding to_rotational degrees of freedom.
Lumped mass matrix of a CST element with 2 DOF per node. In this
t
case, irrespective of the shape of the element, mass is assumed equally
distributed at the three nodes. It is distributed equally in all DOF at each
node, without any sharing of mass between different DOF
0 0 0 0 0
0 1 0 0 0 0
[M]= pAL 0 0 1 0 0 0
3 0 0 0 1 0 0
0 0 0 0 0
0 0 0 0 0

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANAL YSIS 243

(b) Consistenttnass matrix

Element mass matrix is calculated here, consistent with the assumed
displacement field or element stiffness matrix. [M] is a banded matrix of
the same order as the stiffness matrix. This is evaluated using the same
interpolating functions which are used for approximating displacement
field over the element. It yields more accurate results but with more
computational cost. Consistent mass matrices of some elements are
given here.

ww
Consistent mass matrix of a Truss element along its axis (in local coordinate
system)

w.E{u} T = [u
[Nf = [NI N 2]
v]

where, NI = (I-I;)
2 asy
and N2 = (1+1;)
2 ET L ngi T
[M]= fiN]p[N]
v
dV= fA[N]p[N]
0
nee
dx=
+1

fAp[N][NY (detJ)(dx/dl;}d1;
rin
Here,
-I

g .ne
and dx = dx .dl; = det J
dl;
d~ = (~) d~
2
t
Using the values of integration in natural coordinate system,

Downloaded From : www.EasyEngineering.net

244 FINITE ELEMENT ANALYSIS

= PAL[8/3 4/3]= PAL[2 21]

8 4/3 8/3 6 1
Con.dstent mas.~ matrix (if a Plane Tr"ss element, inclined to global X-axis -
Same clements of I-D mass matrix are repeated in two dimen~ions (along X and
Y directions) without sharing mass between them. Mass terms in X and Y
directions are uncoupled.
201 0

ww [M]= pAL 0 2 0
6 1 0 2 0

w.E 002

Consistent mass mqt,.ix of a Space Tr"ss element, inclined to X-V plane) -

a
Same elements of 1-0 mass matrix are repeated in three dimensions (along X, Y

syE
and Z directions) without sharing mass between them.
2 0 0 0 0

[M]= pAL 0
0 2 0 0
0 2 0 0 1 ngi 1 0

6 0 0 2 0 0
0 1 0 0 2 0 nee
0 0 0 0 2
rin
Cmui.~tent mass matrix of a Beam element

[M]=PA(~) f{H}T {H}d~ with Hermite shape functions

g {H} .ne
as used in a

beam element.
2(2 -3~ +~3)
t
L(1 - ~ + ~ + ~
2 3
pAL ) x
= ]28
2(2 + 3~ _ ~3)
L(-1-~+~2 +~3)
[2(2-3S+s3) L(]-S-S2 +S3) 2(2+3S-S 3 )

= pAL
r 156
22L
22L
4L2 13L
54 -13L
-3L2
420 54 13L ]56-22L
L - 13L -3L2 -22L 4L2

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 245

Consistent mass matrix of a CST element in a 2-D plane

[NY =[N) 0 N2 0 N3 0]
o N) o N2 0 N3

[M]= fiN]p[N]T dV =t fiN]P[N]T dA

2 0 I 0 1 0

ww =
ptA
2 0
2 0
1 0
1 0
1

w 12

.Ea Sym
2 0
2 0
1

syE
Note: Natural frequencies obtained using lumped mass matrix are LOWER
than exact values.

ngi
Example 8.1 : Find the natural frequencies of longitudinal vibrations of the

nee
unconstrained stepped shaft of areas A and 2A and of equal lengths (L), as
shown below.

rin
2A A

g.n
I" L L

Solution : Let the finite element model of the shaft be represented by 3 nodes
~I
et
and 2 truss elements (as only longitudinal vibrations are being considered) as
shown below.

2 3
CD

[KL =(2A)(~)[_: -:]=( A~)[_~ -2].

2 '

[K]2 =( ALE)[_~ -:]

Downloaded From : www.EasyEngineering.net

246 FINITE ELEMENT ANAL YSIS

Using consistent mass matrix approach

[M] = p(2A )L [2 I] = pAL [4 2]

I 6 12 624;

[M]2 = P~L [ ~ ~]
Assembling the element stiffness and mass matrices,

ww [K]= :E[_~ -~ ~l -d
w.E o -I

2 0]
a syE
6
1 2
1

Eigenvalues of the equation ([K] - 00 2 [M] ) {u} = {O} are the roots of the
characteristic equation represented by
2 ngi 2

nee
2AEIL-00 4ppALI -2AEIL-00 2ppALI 0 ,
2 2 2
2AEIL-00 2ppALI 3AEIL-00 6ppALI -IAEIL-00 pALI =0
o - AEIL-00 2 pALl6 AEIL-00 2 2ppAL

rin 2
Multiplying all the terms by (LlAE) and substituting J3= pI3oo

2(1-2J3) -2(1+J3) O!
6E
g .ne
or
-2(1+J3) 3(1-2J3) -(I+J3 =0

18
o
~ (~ - 2) (1 -
-(1 + J3}
2~) =0
(1-2J3) t
The roots of this equation are ~ = 0, 2 or ~ or 00 2 = 0 12E2 or 3E2
2 ' pL pL
Corresponding eigenvectors are obtained from ( [K] - 00 2 [M] ) {u} = {O} for
different values of 002 as [1 1f for J3 = 0, [I 0 _2]T for J3 = ~ and
2
[I -\ 1]TforJ3=2.
The first eigenvector implies rigid body motion of the shaft. One component
(u I in this example) is equated to '1' and other displacement components

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 247

(U2 and U3 in this example) are obtained as ratios w.r.t. that component,
following one method of normalisation. Alternatively, they may also be
expressed in other normalised forms.
Note: Static solution for such an unconstrained bar, with rigid body motion,
involves a singular [K] matrix and can not be solved for {u}, while dynamic
analysis is mathematically possible.

Example 8.2

ww
Find the natural frequencies of longitudinal vibrations of the same stepped shaft
of areas A and 2A and of equal lengths (L), when it is constrained at one end, as
shown below.

w.E
a 2A

syE
A
J

ngi
L L

Solution

nee
Let the finite element model of the shaft be represented by 3 nodes and 2 truss
elements (as only longitudinal vibrations are heing considered) as shown below.

1 [0 2 0 rin
3

[K]I=(2~E)[_~ -~]=(~~)[_~ -~] g .ne

[K]2=(ALE)[_~ -~] t
Using consistent mass matrix approach

[M]
I
=P(2A)L[2
6 12
1]= PAL[4
624;
2]
Assembling the element stiffness and mass matrices,

[Kl<EH ~~ -1 [Ml~p~L[~ ! ~j
Downloaded From : www.EasyEngineering.net
Downloaded From : www.EasyEngineering.net

248 FINITE ELEMENT ANALYSIS

After applying boundary condition at node I, I SI row and I SI column of the

assembled matrix can be deleted. Eigenvalues of the equation ( [K] - 0/ [M] )
{u} = {O} are the roots of the characteristic equation represented by
2
3AEIL-00 6pALI6 -AEIL-00 2 pALI6 =0
- AEI L - 00 2 pALl 6 AEI L - 00 2 2pALl6
2 2
13 = pL oo
ww Multiplying all the terms by (LIAE) and substituting

3(1- 213) -(1+13~_0

w.Eor
1
-(1+13) (1- 213)1-
11 ~2 - 14~ + 2 = 0

asy
The roots of this equation are ~ = 0.164, 1.109 or 00 2 = 0.98~ E or 6.6542 E

En pL
Corresponding eigenvectors are obtained from ([K] - 00 2 [M]) {u} = {O} for
pL

Example 8.3
gi
different values of 00 2 as [0 1 1.732]T.

nee
.

rin
Find the natural frequencies of longitudinal vibrations of the constrained
stepped shaft of areas A and 2A and of equal lengths (L), as shown below.
Compare the results obtained using lumped mass matrix approach and
consistent mass matrix approach.
g.n
Solution
Let the finite element model of the shaft be represented by 3 nodes and 2 truss
elements (as only longitudinal vibrations are being considered) as shown below.
e t
::;
::;::;
::;;:;
::;::
2A A

~
:-:::
/" L -\.. L -\
2 3

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 249

[K]I = ( 2~E )[ _ ~ - ~] = ( ALE)[ _ ~ - ~]

[K]2 =(ALE)[ _~ -~]
(a) Using lumped mass matrix approach

ww [M] = P(2A)L[1 0]= PAL[2 0]

1201202;

Assembling the element stiffness and mass matrices,

w.E 2 - 2
[K]=ALE -2
OJ
3 -1 ; [M]=P~L [20 0 OJ
3 0

a [
o -I 1
syE
0 0

Application of boundary condition (node 1 constrained) eliminates row

1 and column I, thus reducing the size of stiffness and mass matrices to

ngi
2 x 2. Eigenvalues of the equation ([K] - (02 [MD {u} = {O} are the
roots of the characteristic equation represented by
3AEI L - (02 3pALI 2
-AE/L
-AE/L
AEI L - (02 pALl 2
=0
nee
rin
Multiplying all the terms by (LlAE) and substituting f3= pe(02

3(1-f3) -I 1=0
2E
g .ne
or
1
-I (1-f3)
3 f32 - 6 f3 + 2 = 0 t
. are
Th e roots 0 f t h ·IS equation (.l-_
I-' (3 ±v'3) or 0.423, 1.577
3
Corresponding eigenvectors are [0 -0.57734 I]T for f3 = 1.577
and [0 0.57734 I]T for f3 = 0.423
(b) Using consistent mass matrix approach

[M]I=P(2~)L[~ ~]=P~L[~ ~]

Downloaded From : www.EasyEngineering.net

250 FINITE ELEMENT ANAL YSIS

Assembling the element stiffness and mass matrices,

ww
l
[K]= ALE -2
2 -23
o -1
01
-I ; [M]=
1
P~L [42
0

Application of boundary condition (node 1 constrained) eliminates row

1 and column 1, thus reducing the size of stiffness and mass matrices to
2
6
1 2
011

w.E 2 x 2. Eigenvalues of the equation ( [K] - o} [M] ) {u} = {O} are the
roots of the characteristic equation represented by

3AE/L-ro 2 6pALl6 -AE/L-ro 2 pALl6 =0

asy -AE/L-ro 2 pALl6 AE/L-ro 2 2pALl6

E ngi
Multiplying all the terms by (L/AE) and substituting /3= pL ro

3{! - 2/3) - (1 + 13~ - 0

6E
2 2

1
-(1+/3) (1-2/3)1-
nee
or 11 ~2 - 14 ~ + 2 = 0

13= (7 ± 3.J3) rin

The roots of this equation are

Corresponding eigenvectors are [0

11
g
or l.10874, 0.16399

-0.57734 I]T for 13 = 1.577

.ne
and [0 0.57734 1f for /3 = 0.423
Note : Natural frequencies obtained with lumped mass matrices are
LOWER than those obtained with consistent mass matrices, while the mode
t
shapes are practically same.

Example 8.4
Find the natural frequencies of vibrations of a simple cantilever beam.

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 251

Solution
Let the finite element model of the beam be represented by 2 nodes and I beam
element to facilitate manual calculation. After applying boundary conditions,

[K]=(~!)[ _I:L ~} [M]=( )[ _I~:L -42~L]

Eigenvalues of the equation ([K] - (j} [MD {u} = {O} are the roots of
12a - 78~ - 6La + II L~ = 0

ww I-6La + IIL~ 4L2a - 2L2~

where
w.E a=-3
EI
L
and

35p2·-204ap+12a2 =0
or

asy
The roots of this equation are ~=-
2a
or
202a

or (0
2 12EI
= - - or
pAL4
E 35
1212EI
pAL4
35

ngi
Corresponding eigenvectors are [ 0.983 n
1.36/L]T for
eer
~= 2a
35

and
35ing
[1.006 7.716/L]T for p = 202 a

8.4 SUMMARY .ne

• A distributed mass system will have as many natural frequencies and
mode shapes as the number of OOF, 'n'.
• Free undamped vibrations involve a set of n homogeneous equations.
t
Such equations will not give a unique solution. A mode shape consists of
relative displacement values at (n-l) OOF, obtained w.r.t. the chosen
displacement value at one OaF. The mode shapes (eigen vectors) are
usually normalised.
• The n natural frequencies may be real or complex (in pairs). Some of
them may be zero (indicating rigid body mode) or repeated.
• Only first few frequencies (lower values) are significant and are usually
calculated by iterative methods. Hence, a coarse mesh is adequate for
dynamic analysis.

Downloaded From : www.EasyEngineering.net

252 FINITE ELEMENT ANALYSIS

• They can be calculated using lumped mass matrix or consistent mass

matrix, based on shape functions used for assumed displacement field.
Each element of mass matrix of an element represents scalar mass,
associated with a particular DOF and no transformation is involved
between element (local) coordinate system and structure (global)
coordinate system unlike stiffness matrix of vector elements.
• Lumped mass matrix is diagonal; has no components in the rotational
DOF; mass of element is equally distributed at all the nodes of the
element, irrespective of shape of the element; same mass at a node is

ww •
taken along all translational OOF at the node and not shared in fractions.
Consistent mass matrix is square, symmetric and banded, just like

w stiffness matrix.

.Ea
OBJECTIVE QUESTIONS

1.
syE
An unconstrained 3-D frame with 4 nodes has
frequencies
number of zero

2.
(a) (b) 2

ngi(c) 3 (d) 6
A frequency of value _ indicates rigid body motion along one dof

3.
(a) zero (b) (c)
nee
infinity (d)
Principal modes of vibration of a multi-dof system are
less than zero

(a) parallel (b)

rin
orthogonal

4.
(c) integer multiples (d) fractional multiples
g.n
With lumped mass matrix, the differential equation of vibration refers to
(a)
(c)
elastic coupling
mode superposition
(b)
(d)
inertia coupling
both inertia and elastic coupling
et
5. With consistent mass matrix, the differential equation of vibration refers
to
(a) elastic coupling (b) inertia coupling
(c) mode superposition (d) both inertia and elastic coupling
6. Normalising eigenvector w.r.t. mass matrix is useful in
(a) mode superposition (b) evaluating natural frequencies
(c) frequency response (d) damped vibrations

Downloaded From : www.EasyEngineering.net

CHAPTER 8 DYNAMIC ANALYSIS 253

7. An unconstrained 2-D frame with 4 nodes has number of zero

frequencies
(a) (b) 2 (c) 3 (d) 6
8. A 4-noded cantilever gives _ number of frequencies
(a) 3 (b) 4 (c) 6 (d) 9
9. A 3-noded simply supported beam gives _ number of frequencies
(a) 3 (b) 4 (c) 5 (d) 7

ww
10. A natural mode of vibration represents _
(a) absolute displacements
at each node

w (b)
(c)
(d)
relative displacements

.Ea
proportional displacements
absolute strain.

syE
ngi
nee
rin
g.n
et

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
En Left Blank"
"This page is Intentionally
gin
eer
ing
.ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 9
STEADY STATE HEAT
CONDUCTION
ww
w.E
asy
Application of FEM is not limited to structural analysis. Availability of faster
computers with large memory have facilitated in generalising mathematical

E
concepts involved in finite element analysis and applying them to many

ngi
different engineering fields. It is now possible to use the same finite element
model of the component for steady state as well as transient thermal analysis,
structural analysis due to static loads as well as dynamic loads etc.

nee
The major difference between structural analysis and thermal analysis by
FEM is in the number of unknowns. While in a structural analysis, the primary

rin
unknowns are vector displacement components ranging from 1 to 6 at any node
in the model depending on the type of component and loads, thermal analysis

g
deals with a single unknown, scalar temperature, at every node in the model.
In many practical situations, thermal load as well as mechanical loads will be
simultaneously acting on a component. It is to be understood that thermal .ne
expansion of a component induces stresses only when the expansion is
partially or completely constrained.

9.1 GOVERNING EQUATION

t
In Cartesian coordinates
•
Consider a small element (a cube of dimensions dx, dy and dz) in a solid body.
The energy balance during time 'dt' can be stated by,
Heat inflow + Heat generated = Heat outflow + Change in internal energy
(qx + qy + qz) dt + q (dx dy dz) dt
= (qX+dX + qY+dY + qZ+dZ) dt + pCp dT (dx dy dz)
or [(qx - qX+dX) + (qy - qy t dY) + (qz - qZ+dZ)] dt + q (dx dy dz) dt
= pCp dT (dx dy dz)

Downloaded From : www.EasyEngineering.net

256 FINITE ELEMENT ANALYSIS

where, k = Thennal conductivity

Cp = Specific heat at constant pressure
and p = Mass density
Considering heat flow through the body along X-direction and using

Fourier's law of heat conduction q = -k.A( ::) along each direction,

qx+dx = qx + (a:: ) dx

ww = qx -(!)[ (:)]dX
kx Ax

w.E = qx - ( !) [ k x ( : )] dx dy dz

or
asy
(qx - qX+dX) dt = (a/ax) [kx (aT/ax)] dx dy dz dt

by dx dy dz dt
E
Considering similar expressions for qY+dY and qZ+dZ and dividing all tenns

ngi
nee
For a homogeneous material, kx = ky = kz = k and hence

~; + : ; + ~; +(i)=(p~p )(~)=(~)(~) rin ••••. (9.1)

where, a = ~ is called thennal diffusivity

pCp
g .ne
For the steady state condition, the time differential on the right hand side
becomes zero.
Then,
t
2 2
a T +a T +a T +
ax 2 ay2 az2 k
=0
2
(.9.) ••••• (9.2)
2
i.e., k V T + q = V . (-k VT) + q = 0 •••• (9.3)
2 2 2
2 a T a T a T
where, VT= -+-+- for 3-D heat conduction
2 ax ay2 az2

a2T a2T
=-+- for 2-D heat conduction
ax2 ay2

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 257

= for 1-0 heat conduction

dx 2
Laplace equation V2T = 0 is a particular case of steady state problem when
q=O
Helmholtz equation

In general, a static field variable problem in terms of the unknown scalar

function 8 can be represented by Helmholtz equation, given by

ww -a.( -OO} +a- (k -00) +-

ax k Xax
a (k -OO) +A.8+q=O
ay Yay az zaz

w.E
This equation represents steady state heat conduction problem when 8
represents nodal temperature; kx , ky and kz are thermal conductivities of the

asy
material along x, y and z directions; A. = 0 and q is the heat source or sink. The
equation can then be written as

~(k
ax x aT}+~(k
ax ay aT}+~(k E
az az z aT}+q
az =0
y
ngi
• Specified temperature T = To nee
Boundary conditions associated with a thermal analysis are:
(at x = XI)
• Specified heat flux (insulated boundary) q = 0
rin
(atx=xJ)
• Convection heat transfer
(on fluid solid interface) g
(at x = Xk)

.ne
In Cylindrical coordinate system

In the case of solids of revolution, eq. (9.1) can be used more conveniently in
cylindrical coordinates (r, 8, z coordinates) as given below
t
.... (9.4)

9.2 I-D HEAT CONDUCTION

In this case, temperature is considered along the length of a rod or thickness of
wall representing the direction of heat flow through conduction. Therefore,
temperature is a function of only one linear dimension, x. Heat conduction in

Downloaded From : www.EasyEngineering.net

258 FIN IT EEL E MEN TAN A L Y SIS

the other two directions is neglected. Steady state equation with no heat source
then reduces to
2
dq=d T =0
dx dx 2

where, heat flux q = -k( ::) is the Fourier's law with -ve sign indicating

reduction of temperature with increasing x.

The geometric model consists of 2-noded elements with heat conduction

ww along the element. Such problems are broadly categorised into three types,
depending on the possibility of convection heat transfer along the length of the

w element.

9.2.1.
.Ea
HEAT CONDUCTION THROUGH A WALL

In this case, conduction across the wall thickness through unit area of cross

syE
section is considered and the waIl is assumed to have very large dimensions in
the other two directions. On the two surfaces of the wall, specified temperature
or specified convective heat transfer from the ambient fluid medium form the
boundary conditions.
ngi
Using the iso-parametric method of derivation of element stiffness matrix,

nee
we can now obtain thermal conductivity matrix, designated as [KT ], for an
element of length L between nodes I and 2. A comparison of the method of

rin
deriving element stiffness matrix and element conductivity matrix for a 1-0
element is given below for better appreciation.

Thermal conductivity matrix, Stiffness matrix, g.n

[KTI
Q/A = q = -k (dT/dx)
T(S) = N) T) + N2 T2 = [N]T {T}
PIA = a = E (duldx)
[KI

u(S) = N) u) + N2 U2 = [Nf {q}

et
dT/dx = [BT]T {T} du/dx = [B]T {q}
dx = (L/2) dS dx = (L/2) dS

N)= (I - S)/2; N 2= (I + S)/2 N)= (1 - S)/2; N2 = (1 + S)/2

[KT] = f [BTf k [B T] (L/2)dS [K] = f [Bf E [B] A (L/2) dS

=k [
L -1
1 -;] = AE[
L -1
1 -;]

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 259

For a 2-element model (Ref. Figure 9.1),

Element - I Element - 2

ww T, ---+~ Heat flow

w .Ea
syE
FIGURE 9.1 1-0 heat conduction through a wall

where ngi
Temperature T at any point in the element is defined by T = [N] {Te}
[N] are the shape functions, as used for displacement function
and {Te} is the nodal temperature vector
nee
:: =(:~)(::) rin
=(d[NY).{TJ 2 g.n
d~

=( ~ )[- 1
(X2 - XI)

1]. {Te}
et
= [BTl {Te}

where [BT]=(~).[-1 1]
Element conductivity matrix,

Since

Downloaded From : www.EasyEngineering.net

260 FINITE ELEMENT ANALYSIS

dx 2 XI X L
and -=--+-= --
dl; 2 2 2

[KT 1-k [ 1 -1]1

= L -1

Element heat rate vector, {R T t =2QL }Nr dl; =2QL {I}I

ww Assembling conductivity matrices and heat vectors of all the elements of a

w.E
structure,
[KTJ {T} = {R}

asy
Similar to the assembled stiffness matrix, assembled conductivity matrix is
also symmetric, banded and singular. Number of these equations is reduced by

E
applying boundary conditions, as detailed below, and the equations are solved

ngi
for the unknown nodal temperatures.
(i) Specified temperature r* (at node m)
•
nee
Penalty approach: A large value C = max (KI J) x 104 is added to the
mth diagonal element (in the mth row and mth column) of the

rin
conductivity matrix [KTJ. In addition, C x T* is added to the element
in mth row of {R}
•
g
Elimination method: K j m x T* is added to the ith element of {R},

.ne
where 'i' ranges from I to total number of rows. In addition, mth row
and mth column of the conductivity matrix [KTJ and heat vector {R}
are deleted.
(ii) Specified convection heat transfer from ambient fluid
q = h (Tm - Too) from node 'm' to the ambient medium at Too
t
The film coefficient value 'h' is added to the element in the mth row and
mth column of the conductivity matrix and the value h Too is added to the
mth element of {R}.

Example 9.1
Consider a brick wall of thickness 0.3 m, k = 0.7 W/m OK. The inner surface is
at 28°C and the outer surface is exposed to cold air at -15°C. The heat transfer
coefficient associated with the outside surface is 40 W/m 2 OK. Determine the
steady state temperature distribution within the wall and also the heat flux
through the wall. Use two elements and obtain the solution.

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 26:1

Solution
Considering a two-element model across half the thickness as shown, heat
conduction matrix of the two elements is obtained as

ill 2 rn
----=::....---- x ----=::....---- convection
3~

ww
w.E
where

,\ssembled conductivity matrix is obtained by adding corresponding terms

as,
a [KT ]= 1;
s[ 1y-1E0]
-1 1+ 1 - 1
o -1 1
ngi
nee
We have 2 boundary conditions of constant temperature T\ = 28 DC and
convection at free end (node 3) with heat flow q given by

rin
q = h (T3- Ta,). = 40 [T3 - (-15)] = 40 T3 + 600
To include these effects, 40 T3 is added on the left side while - 600 and
contribution of T \ are added on the right side.
g.n
Using these conditions, we get modified relations for the 2 unknown
temperatures as
e t
Solving them, we get T2 = 7.68 DC and T3 =-12.63 DC
Heat flow Q can be calculated from the first equation as

Q=(~)(T\ -T )=C34 )(28-7.68 )=94.83 W/m 2

Check: Q = (~ )(T2 - T3 )=C34 ) [7.68-(-12.63)]=94.78 W/m2

Q = h(T3 - To)= 40[(-12.63)- (-15)]=94.8 W/m2

Downloaded From : www.EasyEngineering.net

262 FINITE ELEMENT ANAL YSIS

Example 9.2
A composite slab consists of three materials with thermal conductivities of 20
W/m oK, 30 W/m oK, 50 W/m oK and thicknesses 0.3 m, 0.15 m and 0.15 m
respectively. The outer surface is at 20 DC and the inner surface is exposed to
the convective heat transfer coefficient of25 W/m 2 oK and a medium at 800 DC.
Determine the temperature distribution within the wall.

ww
Solution

Since the plate can be considered infinite, heat transfer can be assumed to be
one-dimensional across the thickness. Heat conduction matrices of the three

w.E
elements covering the three materials are obtained as

asy
2 23m 4 Free
x - - - - - x - - - - - convection

E ngi
=~=~=
where kl

k
L
30
0.3

k2 =-=-=200=66.7x3 W IOC
n
66.7 W IOC

eer
L 0.]5

=~=~=333=66.7x5 ing
.ne
k3 W IOC
L 0.15
The equations after assembling conductivity matrix and heat vector are
obtained by adding corresponding terms as,
[KT] {T} = {R}
t
-1 0 --0 TI 0
-1 1+3 -3 0 T2 0
Or 66.7 =
0 -;3 3+5 -5 T3 0
0 0 -5 5 T4 0

We have 2 boundary conditions of constant temperature T4 = 20 DC and

convection at free end (node I) with heat flow q given by
q= h (T 1 - Tx,). = 25 (TI - 800) = 25 TI - 20000
= 66.7 x 0.375 TI - 20000

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 263

Penalty method of applying boundary conditions

C = max (K 1j ) x 104 = 66.7 x (3 + 5) x 104 = 66.7 x 80000
Then
1+0.375 -1 0 0 T) 20000
-1 1+3 -3 0 T2 0
66.7
0 -3 3+5 -5 T3 0

ww 0 0 -5 5+80000 T4
Solving them, we get T) = 304.6 °C; T2 = 119°C; T3 = 57.1 °C; T4 = 20°C.
66.7 x 80000 x 20

w.E
Elimination method of applying boundary conditions
To include the effect of boundary conditions, 25 TJ is added on the left side

asy
while 20000 and contribution ofT4 are added on the right side.
Using these conditions, we get modified relations for the 3 unknown
temperatures as
1+ 0.375 E
-~ _~l{~:} ={n20~00
66.7
[
-1
o - 3 J 8 gT3
ine
5 T4 x 66.7
}

Simplifying, we get
eri
[
1.375
-1
-1 0j{TJ }_{300}
4 -3 T2 - 0
ng.
or T) =
o -3

304.6 °C
T2 = 119°C
8 T3 100
net
T3 = 57.14 °c

Example 9.3
Heat is generated in a large plate (K = 0.4 W/m 0c) at the rate of 5000 W/m 3
The plate is 20 cm thick. Outside surface of the plate is exposed to ambient air
at 30°C with a convective heat transfer coefficient of 20 W1m 2 0c. Determine
the temperature distribution in the wall.
Solution
Since heat transfer through convection takes place on both the sides of the plate,
half the plate can be considered for analysis with the mid-plane as the plane of
symmetry. Heat transfer can be assumed to be one-dimensional across the

Downloaded From : www.EasyEngineering.net

264 FIN IT EEL E MEN TAN A L Y SIS

thickness while the other dimensions of the plate can be considered infinite.
Considering a two-element model across half the thickness as shown, heat
conduction matrix and heat generation vector of the two elements are obtail1ed
for unit area of cross section.

Line of ~f-I_-=[TI=I=--_ ~ ____rn_2____3 Free

Symmetty convection

ww
w.E
where k
k \=-=( 0.4
L 5 x 10- 2
)=8W
/oC

as, asy
Assembled conductivity matrix is obtained by adding corresponding terms

E8 n-8 OJ
r gi
[K]= -8
o
8+8
-8
-8
8

nee
The nodal load vector consists of heat generation and is given by

{R\}={R 2}=(;L ){:}={:~~} rin

Since
2
QL = 5000 x (5 x 10- ) = 125 W
2 2
g .ne
We have the boundary condition of convection at free end (node 3) with heat
flow q given by
q = h (T3-Too) = 20 (T3 - 30) = 20 T3 -600
t
To include this effect, 20 T3 is added on the left side while 600 is added on
the right side.
Using these conditions, we get modified relations for the 3 unknown
temperatures as

- 88 8 +
-88 - 0j{T\}
8 T2 = { 125 + 125
125}

ro -8 8 + 20 T3 125 + 600

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 265

Caution: Previous type of check on nodal temperatures, based on heat

conducted through different elements, is not applicable here, since the quantity
of heat flowing through node-l is increased by generation of heat between
nodes I and 2 and so heat entering 2nd element at node-2 will be more. Similar
inequality exists between heat flowing through node-2 and node-3.

Example 9.4
A composite bar of 3 different materials, rigidly fixed at both the ends, is

ww
subjected to a uniform temperature rise of 80°C. In addition, axial loads are
applied at two points on the bar as shown. Determine the displacements,

w.E
stresses and support reactions.

Solution

asy
The finite element model of this problem consists of 3 axial loaded elements as
shown.

co En
60 kN
gi
~_ _[I]_2_-.!..::!...~---QJ_3-------B
nee
Material
Section-l
Bronze
Section-2
Aluminium
rin
Section-3
Steel
Area of cross section (mm 2)
Length (mm)
Modulus of elasticity(GPa)
2400
800
83
1200
600
70
600
400
200 g.n
Coefficient of thermal
expansion ( / 0c)
18.9 x 10-6 23 x 10-6 11.7 x 10-6
e t
OJ
•
2
--.t--~~
I'
x
60kN 75kN ~

Stiffness matrices of elements 1,2 and 3 (connected by nodes 1 & 2; 2 & 3

and 3 & 4 respectively) are given by,

3
where k\ =AjE j =2400x83xl0 =249xl0 3
Lj 800

Downloaded From : www.EasyEngineering.net

266 FIN IT EEL E MEN TAN A L Y SIS

3
k2=A2E2 =1200x70x10 =140x10 3
L2 600
3
and k - A3 E 3 600x200xlO =300x10 3
3 - L3 400

Assembled stiffness matrix is obtained by adding corresponding terms as,

ww k,
-k,
-kl
k, + k2 - k2
0 0
0

w.E [K]=
0
0
- k2
0
k2 + k~
- k3
- k3
k3

asy
= 103
-249
249 -249
249 + 140
0
-140
0
0

En 0
0
-140 140 + 300 -300
0 -300 300

gi nee
The nodal load vector consists of loads applied at nodes 2 and 3 as well as
loads due to constrained expansion. These loads are calculated for each element
based on thermal expansion of that element.
Load in element I
rin
= AI EI UI ~T = 2400 x 83 x 103 x 18.9 x 10-t> x 80 = 301.2 kN
Load in element 2 g.n
= A2E2u2~T= 1200 x 70 x 103 x 23 x 10-t> x 80
Load in element 3
= A3 E3 U3 ~T = 600 x 200 x 103 x 11.7 x IO-t> x 80 = 112.3 kN
= 154.6 kN
e t
Direction of load at the ends of each element should result in expansion of
the element. At nodes 2 and 3, mechanical load also should be added to the
thermal load from elements on either side.
Load-displacement relations can thus be written now as
249 -249 0 0 ul -301.2+ RI
-249 389 -140 0 U2 301.2-154.6-60
103 = 103
0 -140 440 -300 u3 154.6 -112.3 -75
0 0 -300 300 u4 1123.+R 4

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE IIEAT CONDUCTION 267

Applying the condition that nodes 1 and 4 are fixed and hence displacements
at these nodes are zero, load-displacement relations corresponding to the
unknown displacements U2 and U3 can be written as

389 -140] {u 2 } ={301.2 -154.6 -60}

[ -140 440 1I3 154.6-112.3-75

Solving these two simultaneous equations, we get

tl2 = 0.222 mm and

ww U3 = -0.0012 mm

w.E
Reactions can be calculated from the two deleted equations

249 ,- 249 0 0] u 2
u,

asy
= {-301.2+R,}
[ o 0 -300 300 u 3 112.3+R4

Therefore. R, = 245 kN
Stresses in the elements,
E and
ngi
R4 = -111.9 kN

0, = E,t, = E, [B,] {q} '-2

J{
nee
=83xl0 3[-800-I - 1
800
0 }
0.222
rin
= 22.62 N/mm2

02 = E2t2 = E2 [B 2] {qh3
g .ne
=70xl0 3 [-
1-
600
- 1
600
J{- 0.222
0.0012
} t
= -26.04 N/mm 2
03 = E3 t 3 = E3 [B3] {qh-4

= 200 x 10 3 [ _ _1_
400
_I_J
400
{- 0.000 12}
= 0.6 N/mm 2

Check: L F = 0 or R, + ~ + P2 + P = 245 - 3 111.9 - 60 - 75 ~ 0

Downloaded From : www.EasyEngineering.net

268 FINITE ELEMENT ANALYSIS

9.2.2 HEAT TRANSFER THROUGH A FIN

A fin is of finite lateral dimensions (Fig. 9.2), unlike infinite lateral dimensions
of a wall. A rod of small cross sectional area, an axi-symmetric plate around the
periphery of an I.C. engine cylinder with heat flow in the radial direction and a
flat plate with its ends across the width insulated can be analysed as 1-0 fins.
The heat transfer is essentially 1-0 heat conduction along the length with heat
loss through its periphery. Equations relating heat flow to nodal temperatures
will therefore include additional matrix of convection heat transfer to represent
heat flow Q through periphery of each element. This convection matrix [H] is a

ww function of variable temperature over the element, which can be expressed in

terms of nodal temperatures using shape functions, and is a square symmetric

w
matrix of the same order as [Kr]. It forms the first (unknown) part 'h. T' of
convective heat transfer h (T-Ta) where Ta is the ambient temperature, while the

.Ea
known second part 'h Ta' is added to the heat flux vector on the right side of the
equation ([Kr]+[HD {T} = {R}.

syE
ngi
nee
Insulated edge

_I- Finite plate with _~

rin
-t--+---.Heat flow insulated edges
direction
Insulated edge

g.n
et
FIGURE 9.2 Different models of heat transfer by conduction and convection

where, [Kr] = (k t )[_~ -~]

{R}=(Ph~ooL ){a

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 289

L l-~ Pdx=hP R
Jh ()2 2] dx
l-~+;
2

ww since
o L 0 L

x-x 2 X3]L
L

hPL

w.E = h P [ -L-+ 3L2 0 = 3

asy
E ngi
where,
and nee
P is the perimeter of the cross section of the element or fin
A = Area of cross section of fin
The coefficients of [KT], [H] and {R} can all be divided by A. These
~, rin
coefficients are thus modified respectively to

books.
L 6A 2A
g
PhL and PhTooL in some

.ne
For a fin of rectangular cross section of width 'w' and thickness 't',
(Ref. Fig. 9.3)
P = 2 (w + t);:::: 2 w; A = w.t
t
Then PIA;:::: 2/t
For afin of circular section of radius 'r'
P = 21t r; A = 1t ~ and PIA = 2 I r
For a tapered fin of length 'L' and rectangular cross section,
{AJ -AJx
A (x ) =A + --'---''----'--
I L

Downloaded From : www.EasyEngineering.net

270 FINITE ELEMENT ANALYSIS

where P, and PJ are the perimeters at nodes I and J

/ ...:......,.:..-Fin

@P A \ /

ww
w.E FIGURE 9.3 Rectangular and circular section fins

Following the same procedure as earlier explained and integrating the terms,
we get
asy -I]; [H] P
E ngi 1
= (~)[3PI +
12
J
P, + PJ

where,
nee
and
rin
Example 9.5
A metallic fin, with thermal conductivity 360 W/m oK, 0.1 cm thick and 10 cm
g .ne
long extends from a plane wall whose temperature is 235°C. Determine the
temperature distribution along the fin if heat is transferred to ambient air at
20°C with heat transfer coefficient of 9 W1m 2 OK. Take width of the fin as 1 m.
t
Solution
Neglecting temperature variation across thickness as well as along tI.<J.1 width,
conduction can be considered along the length of the fin only while heat loss
through convection takes place around the periphery. of the flat fin. Thus, using
I-D model, heat conduction and heat convection matrices of the three elements
are obtained as

[] 2 I] 3 [] 4
- - - - = - - - x -----"=----- x - - - = - - - - Free
convection

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 271

ww
For each element, L =.2.:.!. m
3

w.E ~=.~=10800
L (0.1/3)

asy hL= 9(0.113)=100

3t 3xO.001

hTooL =9x20x (0.1/3) =6000

t 0.001 En
Assembling the element matrices, we get

or
gi nee
([K] + [H] ) {T} = {R}

1
-1 1+ 1
-1 0
-1
0
0
2
1 2+2
0
1 0
0 TI
T2 rin
10800
0
0
-1
0
1+1
-1
-1
+100
0
0
1 2+2
0 1 2
T3
T4 g .ne
=6000
1+ 1
1+ 1
t
1

Substituting the boundary condition of constant temperature TI = 235°C

and dividing throughout by 100, we get

l 220
- 107
o
-107
220
-107
OJ {T2} {120+107T1}
- 107 T3 =
110 T4
120
60

Downloaded From : www.EasyEngineering.net

272 FIN IT EEL E MEN TAN A L Y SIS

Solving these equations, we get

T2 = 207.35 DC; T3 = 190.2 °C and T4 = 185.6 °C
Caution: Previous type of check on nodal temperatures, based on heat
conducted through different elements, is not applicable here, since part of the
heat entering the fin through node-I is lost through convection between nodes 1
and 2 and so heat entering 2nd element at node-2 will be less. Similar inequality
exists between heat entering node-2 and node-3.

ww
Example 9.6
A metallic fin, with thermal conductivity 70 W/m oK, 1 cm radius and 5 cm
long extends from a plane wall whose temperature is 140°C. Determine the

w.E
temperature distribution along the fin if heat is transferred to ambient air a1
20°C with heat transfer coefficient of 5 W1m 2 OK. Take two elements along the
fin.

Solution
asy
En
Neglecting temperature variation across the cross section of the fin, conduction
can be considered along the length of the fin only while heat loss through

gi
convection takes place around the periphery of the round fin. Thus, using 1-0
model, heat conduction and heat convection matrices of the two elements are
obtained as
nee
I] 111
rin
2
3
- - - - = - - - x -----==----- Free
convection

[K T ]= \A [_; -;] g.n

[H]= Ph L
6
[2 1]
1 2
e t
{R}= Ph:ooL {;}

For each element, L = 0.05 = 0.025 m

~ = 70 >d 0-2 x 1t X (0.01)2 =8.8 X 10-3

L 0.025

PhL =1tx(0.01)x5xI0-4 x 0.025 =13.1xI0-8

6 6

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 273

PhT",L =21tx(0.0I)x5xI0-4 x20x 0.025 =785.7xI0-8

2 2
Assembling the element matrices, we get ([K T] + [H]) {T} = {R}
or
-I
(88 I0 1+{-1
X
3
I +I
-1

ww
w .Ea
Substituting the given boundary condition of constant temperature
TI = 140 DC and dividing throughout by 10-3, we get

17.6 -8.8]{T2 } { 1232}

[ - 8.8 17.6 T3 - 0.00786 syE
Solving these equations, we get
ngi
T2 = 93.3 DC and T3 = 46.67 DC.
nee
9.3 2-D HEAT CONDUCTION IN A PLATE
Heat conduction through a finite plate needs a 2-D model for estimating rin
temperatures at various points. Temperature variation across the thickness of a
thin plate at any location is negligible and hence 2-D heat transfer is assumed. g.n
Similar to 2-D structural analysis, we use iso-parametric triangular elements for
this analysis.
Temperature field within a triangular element is given by
et
T=NITI+N2T2+N3T3 = ~TI+11T2+(l-~-11)T3
and x=N I XI +N2X2+N3X3 = ~XI +11X2+(1-~-11)X3
Y = N I Yl + N2 Y2 + N3 Y3 = ~ Yl + 11 )'2 + (I - ~ -11) Y3

Of / a~} =[X13
{aT / Or! X23
Y13] {Of / ax} =[J1{Of / ax} where YIJ = Y,- YJ
Y23 aT / ay Of / ay and xlJ = X,- xJ

or {aT/ax}
Of/ay
= [Jjl {aT/a~}=_1 [Y23
Of/8TJ Det J - X23 -~:3 ][~ ~ =:]H:}

Downloaded From : www.EasyEngineering.net

274 FINITE ELEMENT ANAL YSIS

1 [Y23
= Oet J X 32
Y31
x\3
Y12]{~~}=[B]{Tel
X 21
T3

where, [B]=_I_[Y23 Y31 Y12]

OetJ x32 x13 x21
and [KT] = k A [B]T [B]

ww
w.E [H]= h L'_J [2 1]
asy
Convection along any edge I-J,
612

Example 9.7

En
Two dimensional simplex elements have been used for modelling a heated flat

gi
plate. The (x, y) coordinates of nodes i, j and k of an interior element are given

nee
by (5,4), (8,6) and (4,8) respectively. If the nodal temperatures are found to be
Ti = 110°C, Tj = 70°C and Tk = 130°C, find
(i) the temperature gradients inside the element and
(ii) the temperature at point P located at (xp, yp) = (6,5)
rin
Solution
g.n
e
A triangular element will have three natural or non-dimensional coordinates Nj,
NJ and Nk such that Nl + Nj + Nk = 1 or N k= 1- N l- NJ
(a) Temperature gradient inside the element is given by t
J{T}
~.
aT/Ox}=_I_ Yjk
{ aT/ay Oet J x kj [ Ylo
X ki
YiJ
xij T:

1 6-8 8-4 4-6 1l0}

8-5] {130
=-- 70
DetJ [4-8 5-4

Downloaded From : www.EasyEngineering.net

CHAPTER 9 STEADY STATE HEAT CONDUCTION 275

=~[-2
14 -4
4-2]
I 3 {
IIO}
70
130

since Det J = X,k YJk - XJk Y,k = (5 - 4) (6 - 8) - (8 - 4) (4 - 8) = 14

{~~: C~ ){- 2~~}

} =

ww
(b) To find out temperature at point (xp, Yp), the shape functions (N" N J, N k)
of that point are calculated .as given below.

w .Ea
Xp = N, x, + N J xJ + Nk Xk = N, x, + NJ xJ + (1 - N, - NJ) Xk
6 = 5 N, + 8 NJ + 4 (1 - N,- NJ)
= 4 + N, + 4 Nj or N, + 4 NJ = 2

syE
Similarly, yp = N, y, + NJ YJ + Nk Yk = N, y, + NJ YJ + (l -
5 = 4 N, + 6 NJ-+ 8 (I - N,- NJ)
,N,- NJ)Yk

ngi
= 8 - 4 N, - 2 NJ or 4 N, + 2 NJ = 3

which give N
,
=.±. N =-~.
7' J 14'
N
k
=~
14
nee
Then temperature at (xp, yp) is given by
T p = N, T, + NJ TJ +N k T k rin
=( ~ )(11O)+C~)(70)+C~)(130) g.n
(880+350+ 130)
14
et
9.4 SUMMARY

• Many engineering problems involve stresses due to change in

temperature of the component or surrounding fluid : 1-0 conduction
problem through composite walls and fins; and 2-D conduction problems
through plates.
• To solve for nodal temperature of a finite element model, element
conductivity matrix (due to conduction within the component) is calculated
for steady state problems, in the same way as element stiffness matrix.

Downloaded From : www.EasyEngineering.net

276 FINITE ELEMENT ANALYSIS

The problem may also involve convection on the boundary or periphery

with the ambient fluid.
• Thermal problems and stress analysis probiems can be solved with the
same finitt: dement model, same distribution of nodes and elements in
the compollt:nt; Thermal problem involves only one nodal unknown
temperature unlike I to 6 nodal OOF in structural analysis and hence size
of the assembled matrix is smaller.

ww
• Thermal problems deal with scalar temperature as the nodal unknowns
and, hence, do not require transformation of element matrices from
element (local) coordinate system to structure (global) coordinate system.

w.E
• Transient problems are solved in multiple time steps by iterative methods
and accuracy depends on the selected time step duration.

asy
OBJECTIVE QUESTIONS
1. Conductance matrix is the equivalent of stiffness matrix in
(a)
En
thermal analysis (b) dynamic analysis

2.
(c) fluid flow analysis

gin (d) static structural analysis

problem is solved through iterative method
(a)
(c)
transient thermal (b)
eer
steady state thermal
structure with thermal loads (d) static structural analysis

ing
3. No. ofOOF for a 4-noded quadrilateral thermal element is
(a) 4 (b) 8 (c) 12
4. No. of OOF for a 3-noded triangular thermal element is
(d) 16

.ne
(a) 3 (b) 6 (c) 9
5. No. of DOF for a 6-noded triangular thermal element is
(a) 3 (b) 6 (c) 9
(d)

(d)
12

12
t
6. No. of OOF for a 4-noded tetrahedran thermal element is
(a) 4 (b) 8 (c) 12 (d) 16
7. No. ofDOF for a 8-noded quadrilateral thermal element is
(a) 4 (b) 8 (c) 12 (d) 16
8. No. ofDOF per node in a triangular thermal element is
(a) (b) 2 (c) 3 (d) 4
9. No. ofOOF per node in a quadr, lateral thermal element is
(a) (b) 2 (c) 3 (d) 4

Downloaded From : www.EasyEngineering.net

C 1-1 APT E R 10
DESIGN VALIDA TION AND
OTHER TYPES OF
ww ANALYSIS
w.E
asy
10.1 COMPLIANCE WITH DESIGN CODESEn
gi
Every component may not be analysed for displacements and stresses under the

nee
influence of external applied loads. Standard components like bearings, springs,
bolts, .. are selected from design data books or manufacturer's catalogues.

rin
However, when components are to be manufactured to meet specific needs of
customers, they need to be analysed. If the equipment in operation pose danger
to the people working or living around those equipment, design of all

g
components have to satisfy prescribed design (safety) codes. Different countries
prescribe different design codes for equipment working in their countries. One
.ne
t
of the most popular codes is Boiler & Pressure Vessel Code (Section VIII) for
pressure vessels such as boiler drum, steam or gas turbine, valves, heat
exchanger, condenser,... formulated by American Society of Mechanical
Engineers (ASME). This code or its modified version is followed in many
countries.
Many theories have been proposed to explain failure of components when
SUbjected to external loads, based on maximum normal stress"maximum shear
stress, maximum strain, maximum strain energy, maximum distortion energy
etc .. ASME code is based on the most conservative theory i.e., maximum shear
stress theory. According to this, a component fails when the maximum shear
stress at any point exceeds prescribed limit for the particular material at the
temperature of operation.
Limiting values of all materials are usually based on the values obtained
from uni-axial tensile test. As explained in chapter-3, this specimen is subjected

Downloaded From : www.EasyEngineering.net

278 FIN ITE ELEMENT ANAL YSIS

to tensile stress ax along its length by the application of a tensile load. Its other
stress components a y and 'txy are equal to zero. By recording the elongation
corresponding to different values of load P, until the specimen breaks, yield
stress (Sy), maximum tensile stress (St), maximum elastic strain, maximum
plastic strain and modulus of elasticity (E) are calculated. These values are
tabulated in the code for each material, at different temperatures. To obtain
material property value at any other intermediate temperature, linear
interpolation is carried out from the values at the nearest lower temperature and

ww nearest higher temperature.

ASME code defines stress intensity, Sm = min (2S/3, S/3) for comparison
with the maximum shear stress in the component, with built-in factor of safety.

w.E
In the uni-axial tensile test, using Mohr's circle, we get maximum shear stress
'tmax = Sm/2. Maximum shear stress at any point in the component is equal to
half of the difference between algebraically largest (a]) and algebraically

asy
smallest (a2) principal normal stresses.
(a]-a2)/2<Sm/2 or Pm<Sm

En
Thus, stress intensity (P) is defined as twice the maximum shear stress.

gin
Different components of stress intensity are calculated, as detailed below, from
the three normal stresses and three shear stresses at each point of a component.
Assuming that the magnitudes of stress components vary across the cross

eer
section of any component, mean stress is defined as the membrane stress while
the varying part is defined as the bending stress. Stress intensity P calculated

ing
from the membrane part of the six stress components is called membrane stress
intensity Pm while the maximum stress intensity calculated from the bending
part (observed on the outermost layers) is called bending stress intensity Pb.
These are checked against different limits as given below,
.ne
Pm < Sm and Pm ± Pb < 1.5 Sm
The reason behind specitying two different limits for these components can
be easily understood from the following logic (Refer Fig.l 0.1). In the case of
membrane stress, equal stress at every point in the cross-section will lead to
t
simultaneous failure of the entire cross-section. In the case of bending stress,
stress varies linearly throughout the cross-section. If the maximum bending
stress exceeds the allowable limit, only the outer layer yields without failure of
the entire cross section. As the load increases, more and more outer layers start
yielding. Finally, at a particular load, the entire cross-section will reach the
same stress value, in opposite directions in the top and bottom layers. Thus, the
cross-section can withstand larger load in bending mode and hence the
allowable limit can be more.

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIUN VALIDATION AND OTHER TYPES OF ANALYSIS 279

,.-------/.. ~=======~-I.I.
7

1 1
1 1
1
1 1
/ 1
/
/ 1--'---.....1
/
1
/
/
Actual ,tress Memhrane t Memhrane Bending

ww
Bendlng,slre,s near failure
with 1'", + 1\ ~ () hendlng stress SIre,s (Pm) stre,S (Ph)

FIGURE 10.1 Varying influence of membrane and bending stresses across

w thIckness

.Ea
The membrane stress is again classified as general membrane stress (Pm)
applicable over a large area of the component and local membrane stress (Pd
applicable near the areas of structural discontinuity or stress concentration. Here
again local yielding results in readjustment of stresses without causing failure of
the entire component and thus a higher load can be sustained.
syE
So, PL < 1.5 Son
ngi
and
In some components, stress variation across the cross section may not be nee •••••(10.1)

rin
linear. In such cases, total stress is considered as the slim of membrane stress,
bending stress and peak stress (Q). Peak stress is not considered in design
checks. It influences only fatigue and creep damages.
Then, Pm ± Pb + Q < 3 Sm
g.n
and P L ± Pb + Q < 3 Son
All the above checks, in addition to checks related to fatigue failure or creep
failure wherever applicable, have to be satisfied for validating design of any
product. Output oranv FEM software includes stresses at various points ora
.....(10.2)

et
product while code check is based on stresses across some typical cross
sections or the product. Hence, the six components of stresses (3 normal
stresses and 3 shear stresses) obtained from Finite element analysis at every
node point in a continuum structure need to be categorised considering their
variation across each critical cross section into membrane (uniform),
bending(/inearly varying) and peak stress (non-linear part) components for
validating the design by such codes. Stress categorisation procedure varies
with relevant code. An example oftypical stress classification line (A-B) across
the thickness of a shell-nozzle junction and categorisation of total stress across
the thickness into membrane, bending and peak stress components is shown in
Fig.l0.2

Downloaded From : www.EasyEngineering.net

280 FINITE ELEMENT ANALYSIS

Nozzle axis A~---

ww
w.E
asy (J

E
FIGURE 10.2 Categorisation of Stresses across thickness

ngi
The procedure is summarised below.
•
nee
Each of the six stress components calculated by the software in Cartesian
coordinate system is categorised into membrane, bending and peak stress

•
values.

rin
Principal normal stresses O"J, 0"2 and 0"3 are calculated separately from the

•
bending and peak categories in a 3-D stress analysis. g
six stress components o"x, O"y, o"z, 't xy, 'tyz and 'tzx belonging to membrane,

.ne
Principal normal stresses O"J, 0"2 and 0"3 are calculated from the four stress

•
components·O"x, O"y, o"z (hoop) and 'txy in an axi-symmetric analysis of 3-D
component, with the hoop stress o"z equal to the principal stress 0"3
Principal normal stresses O"J and 0"2 are calculated from the three stress
t
components o"x, O"y and 't xy in a 2-D stress analysis (plane stress or plane
strain).
• Absolute maximum difference of any two principal stresses I O"J - 0"2 I ,
I 0"2 - 0"3 I or I 0"3 - O"J I is calculated for the membrane, bending and peak
categories and designated as Pm or PL, Pb and Q respectively.
Principal normal stresses and stress intensities (Pm or PL, Pb and Q) have to
be calculated separately for membrane, bending and peak stress components.
Depending on the type of analysis carried out using any general-purpose
software, these values have to be compared with the allowable values of the
material at the operating temperature.

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIGN VALIDATION AND OTIIER TYPES OF ANALYSIS 281

Note : This categorisation of stresses is not relevant for discrete structures

involving truss, beam, torsion or pipe elements, because stress is either constant
across their cross section (truss or pipe elements) or varies linearly across every
cross section and is maximum where bending moment or torque is maximum
(beam, torsion or pipe elements).

10.2 TRANSIENT HEAT CONDUCTION

ww
Time-dependent heat transfer problems are very common in engineering. A
brief explanation is given about this topic. Detailed presentation is not included

w.E
here, as this topic is outside the scope of syllabus in many universities. The
solution follows iterative method in multiple time steps and the accuracy
depends on the selected time step duration. The governing equation for transient

a
~; ~; ~:;s=(~)=[p~p
heat conduction problem is

yE )( ~)=(~)( ~) .....
+ + (10.3)

where, a = ngi
k/pC p is called thermal diffusivity

nee
k = Thermal conductivity
C p = Specific heat at constant pressure
and p = Mass density
rin
g.n
In solving this problem, we get an additional matrix, related to the capacity
of a material to absorb heat, called capacitance matrix [Kd

[Kc]=pC p

= pC p AL
ffftNf [N]dV e
[2 1]
for I-D element of uniform section
t
6 1 2
/\

= pCp 6
AL[ 1
-1
-I]
I
for I-D element of varying section

/\ =-'-------''-'-
A
(AI +AJ
where,
2
and [K T] {T} + [Kcl {T'} = {R}

where {T'}= [dTI/dt dT2 /dt ...... y

Downloaded From : www.EasyEngineering.net

282 FINITE ELEMENT ANALYSIS

10.3 BUCKLING OF COLUMNS

The truss element discussed in chapter 4 is assumed to be stable and extend or

shorten, due to the tensile or compressive load applied along its axis. This is not
alway,> true. Slender columns, ,>uojected to axial compressive load, are found to
bend. This phenomenon is called eltlstic insttlhility or elastic htu-kling
(implying that the bending vanishes when load is removed) and occurs due to
the difficulty of applying load exactly along the axis (without any ecccillricity)

ww or due to non-homogeneity of most practical materials (resulting in non-uniform

stress distribution across the cross section).

w A slender member AB of length 'L' and having hinged ends at A and B,

subjected to axial load 'P', bends as shown in Fig.IO.3. The bending behaviour

.Ea
depends on the end conditions of the member (free, hinged or fixed) as well as
its dimensions.

s~A~-------~--~t-~---------~B~---P
p____
y1--+E
ngi x

FIGURE 10.3 Buckling of a slender member due to axial compressive load

nee
Governing equation for this deformation is M I I = EI R

where, M = -

rin
P.y and I/R = d2y/dx 2

g
2
or d y + P.y =0 ..... (10.5)

H; .ne
dx 2 EI

t
The general solution is y = C 1 sin ax + C 2 cos ax where a =

The constants are evaluated from the end conditions.

y = 0 at x = 0 leads to C 2 = 0
while y = 0 at x = L leads to C 1 sin aL = 0

C1 = 0 is a trivial solution; while sin aL = 0 or L /P = n 1[ where 'n' is

fEl
any integer

For n = I, Pcr = 1[2

L
;1 is called critical load or Euler load.

This is an eigenvalue problem and the function y = C 1 sin ax is called an

eigen function. Discrete values of buckling load P = n 21[2EI/L2 are called

Downloaded From : www.EasyEngineering.net

CHAPTER 10 OESIGN VALIDATION AND OTHER TYPES OF ANALYSIS 283

eigenvalues and corresponding mode (or displacement) shapes are elgen

vectors, similar to natural modes of vibration, as shown in Fig. 10.4.

n=1
11=2 n=3

FIGURE 10.4 Buckling mode shapes

ww Corresponding buckling stress,

w.E a
cr
P
= cr
A
1{2 EI
-=--
L2A
1{2 E
(L/p)2
..... (10.6)

where p =
asy
~(~) is the radius of gyration.
E ngi
For loads smaller than Per. the deflection C, or 0 is zero, implying that the
column remains straight and shortens due to the applied compressive load.

nee
For a given material of modulus of elasticity 'E', the critical stress cr cr
increases as the slenderness ratio (Lip) decreases i.e., as the column becomes
shorter and thicker. Thus, below a particular value of' Lip', for which cr cr > cry,

rin
cr cr = 1{2E/(Llp)2 is not reached and the column .does not buckle, before the
column starts yielding. For mild steel, this limiting slenderness ratio is;:::: 100.

10.4 FATIGUE ANALYSIS g .ne

t
Fatigue life depends on fluctuating stress cycle, identified by stress amplitude
over a mean stress at a point in the component, and the number of such cycles
the product is designed for. All points of a component may not experience same
stress amplitude or same mean stress. Also, a single point may experience stress
cycles of different amplitude, for example due to cold start, warm start and hot
start or due to load change in a steam turbine casing.
Some components are subjected to more than 106 stress cycles, for example
due to flexural stress cycles of rotating components or stresses due to vibration
in static components. They fall in the category of high cycle fatigue and are
designed for infinite life by limiting stress amplitude below endurance limit
(SE)' Components subjected to pressure and temperature fluctuations, such as
due to different start-up procedures of power plant components, experience less

Downloaded From : www.EasyEngineering.net

284 FINITE ELEMENT ANALYSIS

than 106 stress cycles. Such components are designed for finite life, based on
the stress amplitude (>SE) at critical points of the component.
For example, a turbine casing or rotor experiences zero stress, when shut
down and very high stress during start-up, when it comes in contact with steam
at high temperature and pressure. Start-up procedure for a typical 110 MW
steam turbine, from cold condition, is given in Fig. 10.5. Change of stress from
stress-free cold condition to the high stress value during transient and back to

ww zero stress after shut-down forms one stress cycle, with stress range (St) and the
turbine is designed for a particular number of such cold starts (nt) during its
design life of about 40-50 years. Similarly other transients like warm start, hot

w.E
start, load fluctuation due to varying demand between day and night etc .. all
give different stress ranges (S2, S3, .. ) and each of them may occur for a different

a
number of cycles (n2, n3, .. ).

syE .,.
,. .:
,.i- - - - 360 Tlhr

"
ngi ~--;-I~-+-'- - 535°C
If
,..: ,"".f"---- 130 ata
": I

:"
. .
~....
. : ..
.,;,""
. -r nee
. ,. ,. . . . . 1'. . . . ". - ... 110 MW
Load

rin
/
-,,'/ : ." : - - - . Steam flow
360"C ............,. .. r
"" I. I
1 MWluiin - - - - . Pressure
35 ata
."".

~.
" ".;:
"..

- - / . .;
_.

30
!
75
I

I
:
I

:
I

110 130
g
---Temp

.ne
t
Time in min

FIGURE 10.5 Typical cold start diagram of a 110MW thermal power plant

The stress range will vary at different points of the structure, for the same
transient. Corresponding to each stress amplitude, number of cycles to cause
fatigue failure (Ni) is read from Stress amplitude (S) V s Number of cycles (N)
curve of the material (Refer Fig.lO.6, for a typical curve), assuming it acts
alone. Due to large scatter in the experimental data of S-N values from uni-axial
fluctuating load test, a factor of 2 on stress and 20 on number of cycles is used
in some codes, white preparing S-N curve of a material. Then, fatigue usage
fraction 'U' is obtained from the most widely used model (Palmgren-Miner
hypothesis, popularly known as Miner's rule), for validating safe design of a
component.

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIGN V ALIDA nON AND OTHER TYPES OF ANALYSIS 285

Log S

1
- - - - - - - -1- -
1 1
1
- - - - - - - -1- - . . , - -
1
S[ _________1_ 1 - -+1 - - ..,1- - lII.-_ _ __

ww Infinite life region: :

1
:

LogN

w.E FIGURE 10.6 Typical S-N curve of a material

as~[y~ii J
Fatigue usage fraction, U F = <1 .....(10.7)

E
The finite element analysis gives us 6 stress components (3 normal stresses

ngi
and 3 shear stresses) at each node point of a 3-D continuum structure or 4 stress
components (3 normal stresses along axial, radial and hoop directions and 1

nee
shear stress) in an axisymmetric analysis, while S-N curve of a material is
given for one particular stress. Usually vonMises stress or maximum principal

rin
stress difference is used with the S-N curve. This stress value is obtained at each
point and at each time step of the transient. Stress amplitude (Si), half of stress
range, is calculated over the complete transient, consisting of many time steps.
Corresponding to this stress amplitude, number of cycles to failure (Ni) is g.n
obtained from the S-N curve of the material and fatigue usage fraction
calculated to validate the design. et
10.5 CREEP ANALYSIS

Creep is a phenomenon in which a component, stressed well within its yield

point by the applied loads, yields when the load is applied for a prolonged
period. A constant stress state for a prolonged period is naturally feasible when
the equipment is operating in steady state condition. The time in which the
material starts yielding, at a particular stress level, is called rupture time (Refer
Fig. 10.7). Creep is prominently observed in equipment operating at elevated
temperatures.

Downloaded From : www.EasyEngineering.net

286 FINITE ELEMENT ANALYSIS

Stress

ww
w.E
Time to rupture, t,

FIGURE 10.7 Typical Stress vs. rupture time curve of a material

aIf a structure, during its life time, works for periods of time (tj) at different

syE I (t')
stress levels, life fraction rule (similar to Miner's rule) states that

ngi
Creep usage fraction, U c = < I ..... (10.8)
m -.!...
.=1 tn

nee
The finite element analysis gives us 6 stress components (3 normal stresses
and 3 shear stresses) at each node point of a 3-D continuum structure or 4 stress
components (3 normal stresses along axial, radial and hoop directions and I

rin
shear stress) in an axisymmetric analysis, while S-t, curve of a material is given
for one particular stress. Usually vonMises stress or maximum principal stress

calculated at each point during steady state operation of the component.

g
difference is used with the S-tr curve. This p..rolonged duration stress (Sj) is

Corresponding to this stress value and operating temperature, rupture time (tri)
.ne
is obtained for the particular material and the creep usage factor calculated to
validate the design.
Cumulative damage due to fatigue and creep
t
There are components, as in a hydro turbine, which are subjected to fatigue only
and creep at the operating temperature is insignificant. There are also
components, as in a steam turbine, where fatigue due to load transients and
creep due to prolonged. steady load operation at high operating temperature
are significant. It is well known that fatigue damage and creep damage are'inter-
related; but, the interaction effect is not well understood. Hence, some design
codes suggest calculation of cumulative usage as the algebraic sum of damages
due to fatigue and creep. Thus,

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIGN VALIDATION AND OTHER TYPES OF ANALYSIS 287

Cumulative usage fraction, U = I (n)

~ +I
III

,=1 n,
k
J=I
(t )
~
tT
••••• (10.9)

and it is usually limited to ::::: 0.5 to account for the unknown interaction
effects.

10.6 DAMPED FREE VIBRATION

ww
Every structure, when excited with some force, vibrates with a particular
frequency and amplitude. In most cases, amplitude of vibration reduces with
passage of time and finally reaches zero value. This is called damping -

w.E
inherent structural damping (usually considered as a function of mass and
stiffness) or external damping with spring and/or viscous dash pot.

asy
In a single degree of freedom system,
k u(t) + c u(t) + m U(t) = 0 ••••. (10.10)

h
were u, =
d-u .IS t h
dt E
' at that pomt
e veiOClty .

ngi
c = damping coefficient, which is usually a function of velocity

and ..
u=
du
dt
2
d u. I I' h nee
.
or dt 2 IS t le acce eratlon at t at pomt

A solution of the form u = aeA.t will satisfy the equation. rin

Then, m ,.} + c A, + k
equation and its roots are
= g
0 is the characteristic equation of the governing
.ne
t
The type of motion depends on the nature of the roots or on the value of
(c/2m)2 - (kim)

(i) If (2:

negative.
r (~)
> or c > 2 .Jkm , the roots A,\ and A,2 are real and

Then, u(t) decays as a function of time and no vibration occurs

Downloaded From : www.EasyEngineering.net

288 FINlTE ELEMENT ANALYSIS

(ii) If (2: r (~)

< or c < 2 .Jkm , the roots A.I and A.2 are a pair of

complex conjugate solutions. The real part represents exponential decay

of the amplitude of vibration and imaginary part represents oscillatory
part of motion with a damped free vibration frequency given by
~(k/m)- {c/2m)2

ww (iii) (2: r (~)

= or c = 2 .Jkm , is called critical damping

w Any linear combination ofthese roots is also a solution. Thus, in general,

.Ea u = al eA.lt + a2 eA2t is also a solution

By assuming sinusoidal function for u(t) and expressing 6(t) and ti(t) in

syE
terms ofu(t), the above equation can be rewritten as
(k + b c + b 2 m) u(t) = 0

ngi
In a multi-degree of freedom system, the stiffness, damping and mass terms
take the form of matrices and the governing equation can be expressed in the
matrix form as
2
([K] + b [C] + b [M] ) {u(t)}
nee
= {O} ..... (10.11)
where, structural damping is assumed to be a function of mass and stiffness,
given by
rin
[C] = a [M] + ~ [K] with a and ~ being real constants.
The elements C'j of the damping matrix [C] have a physical significance
g.n
analogous to elements k'J of the stiffness matrix [K]. Cjj is equal to the external
force required at node I in direction Uj to produce unit velocity at node J in
direction uJ with velocities at all other masses zero. Analogous to stiffness
matrix, C'J = Cj,.
et
Damping force {F}D = - [C] {Ii }.

10.7 FORCED VIBRATION

When a structure is subjected to dynamic (time dependent) loads such as wind

load, the displacements, strains and stresses will also vary with time. The
solution can be obtained by marching in a series of time steps M and evaluating
accelerations, velocities and displacements at each step. Modal matrix can be
used in the mode superposition method to diagnose mass, damping and stiffness
matrices and thus uncouple the equations of motion.
([K] + b [C] + b2 [MD {u(t)} = {F(t)} ••••. (10.12)

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIGN VALIDATION AND OTHER TYPES OF ANAL YSIS 289

The calculation of nodal displacements V s time is called time history

response.
Sometimes, the dynamic load may not be known in the form of Force Vs
time, but as random vibration (such as earthquake) in the form of frequency Vs
acceleration. This frequency spectrum is treated as a linear combination of
individual acceleration (amplitude) Vs time of specific frequencies. Each
acceleration Vs time is equivalent to applying force Flit) at node I in Jlh DOF

ww
given by F,j = mass at node I x acceleration in Jlh DOF. This analysis is called
response spectrum analysis.

:1.0.8 w.E
TORSION OF A NON-CIRCULAR BAR

asy
A bar of circular cross section, subjected to pure torsion, is analysed using the
assumption that plane section before applying torsion remains plane after

E
applying the load. i.e. there is no warping of the section or all points on the

ngi
cross section will have either zero or equal displacements along the axis of the
bar. The same is not true with a bar of non-circular cross section. Displacements

nee
at different points of a bar of non-circular cross section subjected to torsion are
defined by
u = -9 zy; v = 9 zx and w = 9 ",(x,y)
rin
where ",(x,y) is the movement of cross section in the axial direction (z-axis) per
unit twist and is called warping function
and 9 is the angle of twist per unit length
g .ne
For pure torsion, ex = ey = ez = 'Yxy = 0
aw au aw
'Yxz = Ox + az = Ox -!:Jy
t
'Y
aw Ov aw
=-+-=-+9x
yz Oy az Oy

Applying Hooke's law, the stress components become ax = ay = az = 'txy = 0

'txz =?Y xz =G [ : - 9Y]

tyz =Gyyz =G[: +sx]

Downloaded From : www.EasyEngineering.net
Downloaded From : www.EasyEngineering.net

290 FIN IT EEL E MEN TAN A L Y SIS

The general equation of motion,

p. p + (J. =
I IJ,J
P (dUdt I
)

simplifies for the case of body force PI = 0 to

Ot xz + Ot yz =0 since displacement Ui is independent of time t

ax 8y

ww Substituting for the shear stresses, from the above, yields Laplace's equation

a2 w a2w
w.E --+--=0
ax2 8y2

a +a~ =
-t
2 2
Using 'V(x,y), we get
asy ax 8y
0

E
The stress-free condition of the periphery of the cross section requires that
the tangential stress be zero on the lateral surface of the prism. Thus, the
boundary condition is of Neumann type,
ngi
Ax + Tyz A.y = 0
nee
Txz

h
were, "I
II.
x = -dy an d "I
lI.
y
- dx
=- - are t he d"IrectJon cosmes
. 0 f t he tangent vector

rin
ds ds
Substituting the above, we get

g .ne
Integral of the variational problem is given by

1= U=(~) Jcr dV v
ij Eij
t
We use finite element idealisation over the cross section and interpolate the
unknown solution function 'V(x,y) by a polynomial over each element through
10dal values 'Vi. From variational principle,
81 = 0

or ~=O \f i=l,n
mvi

Downloaded From : www.EasyEngineering.net

CHAPTER 10 DESIGN VALIOATION AND OTHER TYPES OF ANALYSIS 291

This condition leads to a system of linear algebraic equations to be solved

for '1'. in the usual manner.
Alternatively, a hybrid finite element model can be developed by introducing
stress function cp(x,y) interpolated over each element (satisfying equilibrium
conditions) and warping function 'I'(x,y) specified along inter element boundary
(satisfying compatibility conditions). This model gives more accurate stress
solutions, particularly in the vicinity of external boundary. The stress

ww
components, in terms of the stress function, are given by

't xz
8cp
=- ; 't yz
c3<p
=-

w.E dy dx

Complete restraint of warping of a member, of non-circular cross section and

asy
subjected to torque, reduces rotation at the end but introduces axial normal
stresses much larger than the torsional shear stress. Even in a simple portal

En
frame, shown below, warping is at least partially restrained at nodes A, 8, C
and D. Such restraint influences response due to loads normal to plane of the

gi
frame. Full or partial restraint of warping is experienced at every joint of a
frame and can be included with an additional OOF (rate of twist, d9 x/dx) at each

nee
node. Most commercial software do not include this 7th OOF at a node and
warping restraint as well as associated axial stress are ignored.

c rin
B

g.n
A D e t

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
En Left Blank"
"This page is Intentionally
gin
eer
ing
.ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 11
COMPUTATIONAL FLUID
DYNAMICS
ww
w.E
1L1 INTRODUCTION asy
En
The physics of almost every fluid flow and heat transfer phenomenon is
governed by three fundamental principles - mass conservation, momentum

gin
conservation (or Newton's second law) and energy conservation - taken
together with appropriate initial or boundary conditions. These three principles

eer
may be expressed mathematically in most cases through integral or partial
differential equations (POE) whose closed form solutions rarely exist. The
ability to seek numerical solutions of these governing equations has led to the
development of Computational Fluid Dynamics (CFO).
ing
To obtain numerical solution to the physical variables of the fluid field,
various techniques are employed:
.ne
• manipulating the defining equations
• dividing the fluid domain into a large number of small cells or control
volumes (also called mesh or grid)
t
• transforming partial derivatives into discrete algebraic forms and
• solving the sets of linear algebraic equations at the grid points
Modem CFO can handle fluid flow associated with other phenomenon such
as chemical reactions, multi-phase or free surface problems, phase change
. (melting, boiling, freezing), mass transfer (dissolution) and radiation heat
transfer. A CFO code has three basic components - preprocessor, solver and
postprocessor. The solver is the heart of a CFO code and is usually treated as a
'black box' while the other two components provide user/computer interface.
Solver is based on one of the three major discrete methods - finite difference
method (FOM), finite element method (FEM) or finite volume method (FVM).

Downloaded From : www.EasyEngineering.net

294 FINITE ELEMENT ANALYSIS

Over 90% of CFD codes are based on FDM or FVM. FVM is now very well
established and is used in most commercial CFD packages like FLUENT,
FLOW-3D.
FEM was initially developed for structural analysis but has been extended to
fluid flow problems as it offers the advantage of non-regular grid, capable of
simulating complex boundary geometries. Also, the methodology used for
describing flow conditions within each cell, though more complex than FDM,
have a higher degree of accuracy. FVM draws together best attributes of FDM

ww and FEM. It is capable of simulating complex boundary geometries while

utilising relatively straight forward finite difference relationships to represent
the governing differential equations. Complete presentation of FDM and FVM

w.E
are not in the scope of this book. FDM is detailed below as some of these ideas
are also used in FVM.

11.2
a syE
GOVERNING EQUATIONS

Partial differential equation is an equation involving one or more partial

derivatives of an unknown function of two or more independent variables.

ngi
Order of the highest derivative in the equation is the order of the equation. The
partial differential equation is termed linear if the dependent variable (unknown

nee
function) and its partial derivatives are of first degree. The equation is termed
homogentous if each term of the equation contains either the dependent
variable or one of its derivatives.

rin
Some important linear partial differential equations in fluid dynamics are:

1. I-D wave equation

a2u 2 a2u
--=c - -
at? &2 g .ne
2. I-D heat conduction equation
k
where a = - -
(pCp) t
3. 2-D Laplace equation

4. 2-D Poisson equation

5 .. 3-D Laplace equation

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 295

A single partial differential equation can have more than one solution.
Unique solution of a partial differential equation, corresponding to a given
physical problem, depends on additional information like:
(a) Boundary comlitions: values of the required solution on the boundary
of some domain
(b) Initial conditions: values of the solution, when time t is an independent

ww variable, at time = o.
Superposition : For a homogeneous linear partial differential equation with

w
known solutions UI and U2, any linear combination of these solutions is also a
solution.
.Ea
i.e.
where syE
u = CI UI + C2 U2 is also a solution,
CI and C2 are constants.

Ex: ngi
Any PDE which is linear in the highest derivative is termed as quasilinear.
A Uxx + B uxy + C Uyy = F(x, y, u, ux, Uy) ..... (11.1)

where
au
u =- . u =_. u
x Ox' y Oy' xy
au =--
a2 u
Ox Oy nee
rin
Using

g.n
and
= uxx dx + uxy dy
et
= Uxy dx + Uyy dy

eq. (11.1) can be rewritten as l~ B

dy
dx
~Jr:}~F}
dy u yy du y

r- ~~
For a non-trivial solution,

A (dy)2 - B (dx)(dy) + C (dxi = 0 or

A( ~~ B( J+ C =0

Downloaded From : www.EasyEngineering.net

296 FINITE ELEMENT ANALYSIS

This can be expressed in matrix form as

ABC
dx dy 0 =0 ..... (11.2)
o dx dy

d [B±~B2 -4AC]
which gives ~ = = - - - - - - - =

ww
dx 2A
Characteristic curves, represented by this equation, can be real or imaginary
depending on the value of (B 2 - 4AC).

w Partial differential equations (PDEs) (in particular, governing equations in

.Ea
fluid dynamics) are classified into three categories:
• Elliptic, when B2 - 4AC < 0

syE . a2u a2 u
[Laplace equation, - 2 + -2 = 0
ax ay

ngi 02U 02U

Poisson equation, - 2 +-2 =f(x,y)]

• Parabolic, when B2 - 4AC = 0

ox
nee
oy

[Heat equation, ' : ~ a( ~~ ) rin

and a > 0]

• Hyperbolic, when B2 - 4AC > 0

g.n
[2 nd order Wave equation,
et
11..3 FINITE DIFFERENCE METHOD (FDM)
Basic idea of FDM is that derivatives in differential equations are written in
terms of discrete quantities of dependent and independent variables, resulting in
simultaneous algebraic equations with all unknowns prescribed at discrete mesh
points covering the entire domain. Appropriate types of differencing schemes
and suitable methods of. solution are chosen, depending on the particular
physics of the flows which may include:
• inviscid or viscous flow
• incompressible or compressible flow
• irrotational or rotational flow

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 297

• laminar or turbulent flow

• subsonic or transonic or supersonic or hypersonic flow etc.
FDM utilises a time distance grid of nodes and a truncated Taylor series
approach to determine the conditions at any particular node one time-step in the
future based on the conditions at adjacent nodes at the current time.
There are two schemes of using Finite Difference type approximations to
convert the governing Partial Differential Equation into an algebraic format -

ww
Explicit and Implicit. In the explicit method, required value of variable at one
time-step in the future is calculated from known current values. It requires
selection of a very small time-step based on the grid size and hence takes more

w.E
computer time. Implicit methods allow arbitrarily large time-step thus reducing
computer time required for solution. However, these methods do not give direct
solution and require iterative solution which sometimes may lead to

a
convergence problems (for control volumes with large aspect ratios). Also,

syE
these methods are not accurate for convective processes.

:1.1.4
ngi
ELLIPTIC EQUATIONS (OR BOUNDARY VALUE PROBLEMS)

First boundary value (Dirichlet) problem -

nee
u is prescribed on the boundary curve C of region R
Second boundary value (Neumann) problem-

Un = rin
au (normal derivative of u) is prescribed on the boundary curve C of

region R
an
Third boundary value (Mixed) problem -
g .ne
u is prescribed on a part of the boundary curve C and
part of C of region R
Un on the remaining

C is usually a closed curve or sometimes consists of two or more such

t
curves.
To obtain numerical solution, partial derivatives are replaced by difference
quotients. Using Taylor's expansion,

~
u(x + h,y) u(x,y) + hu.(x,y) + (~ )Uu (x,y) + (~ )u."" (x,y) + ... ..(1.1.3)

Similarly,

u(x - h,y) = u(x,y) - hux(x,y) + (h22)uxx (x,y) - (~ )uxxx (x,y) + ... ..(11.4)

Downloaded From : www.EasyEngineering.net

298 FINITE ELEMENT ANALYSIS

Subtracting eq. (4) from eq. (3) and neglecting higher order terms in h,
[u(x+h,y)-u(x-h,y)]
u" ( x,y ) ~ 2h ••••• (11.5)

[u(x,y+k)-u(x,y-k)]
Similarly, uy( x,y ) ~ 2k .••••(11.6)

These are called central difference formulae for the first derivative.

ww On the same lines, two other forms can also be derived based on Taylor
series expansion at some other points

w.E
u x ( x, y ) ~
[u(x + y,h)- u(x,y)]
h
or Forward difference formula ••••• (11.7)

Ux
a syE
( x,y ) ~ [u(x,y)-u(x-h,y)] or Backwar"
h
''.Ierence fiormula..... (118)
·.ldif.J1"1: .

ngi
Second derivative can also be expressed, in a similar way, as

uxx ( x,y ) ~
~[ux(x+h,y)-u(x-h,y)] [ux(x+h,y)-ux(x,y)]
2h or h

~
nee
[u(x + h,y) - 2u(x,y) + u(x - h,y)]
.••••(11.9)

rin
h2

[uy(x,y+ k)- uy(x,y - k)] [u y(x,y + k) - uy(x,y)]

Uyy(x,y)~ 2k or

[u(x,y + k) - 2u(x,y)+ u(x,y + k)]

g k

.ne
t
~ k2 •••.• (11.10)

These are called central difference formulae for the second derivative.
Substituting eq. 11.9 and eq.l1.1 0 in Laplace equation and using h = k,
a2 u a2 u
V 2 u = - 2 +-2 = u(x + h, y) + u (x- h, y)
ax 8y
+ u (x, y + k) + U (x ,y - k) - 4u (x, y) = 0 .....(11.11)
or u at (x, y) equals the mean of values of u at the four neighbouring mesh
points.
This is called 5-point regular operator

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 299

Laplace equation can also be expressed using some other difference

formulae as
V 2U = U(x + h,y + k) + u(x - h, y + k) + u(x + h,y - k) +
u(x - h, y - k) - 4u(x, y) = 0 ..... (U.12)
This is called 5-point shift operator

ww
Combining eq.ll.ll & -eq.ll.12, we get 9-point formula, given by

u (x + h, y) + U (x - h, y) + u (x, y + k) + u (x, y + k) +

w.E
u (x + h, y + k) + u (x - h, y + k) + u (x + h, y - k) +
u (x - h, y - k) - 8u(x, y) = 0 ••••. (11.13)

a syE
The coverage of 5-point operators and 9-point operator are shown below on
the model,

x, y + k
1
x - h,y + k
I
ngi
x+ h, Y + k
1 I I I

1
x, h + «
·4
(x, y)
1
x h. y
.-4
x,y
nee 1 8
x,y
1

x, h-·k I
x h,y k
I I
x + h. y. k
I

rin
I 1

5-point regular operator 5-point shift operator 9-point formula

g .ne
Similarly, Poisson equation can be rewritten in central difference form as

u (x + h, y) + u(x - h, y) + u(x, y + k) + u(x, y - k)

- 4u(x, y) = h2 f(x, y) .....(U.14)
t
Use of polar coordinates (r-9) may be more convenient with some
geometries. For those situations, Laplace equation in polar coordinates can be
obtained by substituting

cos 9 = x; sin 9 = r; tan 9 = Y and r2 = x2 + l

r r x

in •••••(11.15)

Downloaded From : www.EasyEngineering.net

300 FINITE ELEMENT ANALYSIS

where

ww i+2

w.E va
Vr

asy i+3
Vr
i+I

En i+4

gi
(a) Numerical solution of Dirichlet problem

nee
A grid consisting of equidistant horizontal and vertical straight lines of
distance h is introduced in the given region R. Then, the unknown
values of u at each of the grid intersection points or mesh points are
related to the neighbouring points. This yields a system of linear
rin
algebraic equations. The coefficients of the system form a sparse matrix.
It can be changed to a band matrix by numbering mesh points in such a
way that all non-zero elements are arranged around the principal g.n
diagonal. The system of linear algebraic equations can' be solved by
Gauss elimination method or Gauss-Siedel iteration method.
Alternating Direction Implicit (ADI) method uses Equation (11.9) and
e t
rewritten as
u(x + h, y) - 4u(x, y) + u(x - h, y) = - u(x, y + k) - u(x, y - k)
or Ui+l,j - 4ul ,j + Ui_l,j = - UI,j+) - Ui,j_) ••••• (11.1.8)

and u(x, y + k) - 4u(x, y) + u(x, y - k) = - u(x + h, y) - u(x - h, y)

Ui,j+l - 4Ui,j + UI,j_) = - Ui+), j - u l _), j ••••• (~1.!)

Solution is obtained by iteration method from an arbitrary starting v~lue

Uij at all mesh points. Eq.(11.l1) for a fixed row). gives a system of.t,

linear equations, corresponding to n columns, in n unknowns which'can

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 301.

be solved for U'j by Gauss elimination. This can be repeated, row by

row, for all the rows. In the next step, Eq.( 11.12) for a fixed column i,
gives a system of m linear equations, corresponding to m rows, in m
unknowns which can be solved for U'j by Gauss elimination. This can
be repeated, column by column, for all the columns.
Convergence can be improved by moditying eqs. 11.11 and 11.12 as
U,+l,j - (2 + p)u',J + Ui-l,J = - U',j+l + (2 - p)u',j - U"j-l ...••(11.18)

ww U',j+l - (2+P)Ui,j + U',j-l = - U,+I,J + (2-p) U',j - U,-l,j ...•.(11.19)

w.E
where p is a positive number

Example 11.1

asy
A rectangular plate is subjected to temperatures on the boundary as shown. Find
the temperature distribution inside the plate.
Solution
E
Let us divide the plate into a 3 x
ng\.i
3 mesh, for ease of calculation, as shown
T~OuC

1,3
A
I

I
I
n eer
ing
I
1,2 I 2,2 3, 2

~ I
I ,/
T= 100 "c

.ne
I
0,1 3,
I
1,1 I 2, I
I

1,0
B
I

I
I
2,0
I
T= 100 "c
t
(a) Gauss elimination method
Applying eq.(11.9) at (1,1) -4 Tl,l + T 2,l + T l,2 = - (Tl,o + TO,I) = -200
Applying eq.(11.9) at (2, I) Tl,1 - 4 T 2,l + T 2,2 = - (T 2,o + T3,1) = -200
Applying eq.(11.9) at (1,2) Tl,l - 4T l,2 + T 2,2 = - (T l,3 + To,2) = -I 00
'/ Applyingeq.(11.9) at (2,2) T 2,l +Tl,2 -4T 2,2 =-(T3,2+ T2,3)=-IOO
Solving these 4 simultaneous equations, we get
Tl,l = T2,l = 87.5 °C; T 1,2 = T 2,2 = 62.5 °C
Note that the results also satisfY symmetry about the center line AB

Downloaded From : www.EasyEngineering.net

302 FINITE ELEMENT ANALYSIS

(b) Gauss-Siedel iteration method or Liebmann's method

Starting with an initial approximation of TI,I = T2 ,1 = T I,2 = T 2,2 = 100,
and substituting the most recent values on the right side of the modified
equations, we get improved values for the unknowns on the left side,
after each iteration.
-4 TI,I + T 2,1 + T I,2 + 4T2 ,2 = -200 or TI,I = (T 2,1 + T I,2 + 200) 14

ww TI,I -4 T 2,1
TI,I
+ T 2,2 = -200 or T 2,1 = (TI,I + T 2,2 + 200) 14
- 4T I,2 + T 2,2 =-100 or T I,2 = (TI,I + T2,2 + 100) I 4

w.E Iteration
T 2,1 + TI,2 - 4T2,2 = -100 or T 2,2 = (T2,I+T I,2+100)/4
T],1 T2,] T],2 T2,2

asy 1
2
100
93.75
100
90.62
75
65.62
68.75
64.06

E
3,
4
5
89.06

ngi
87.89
87.72
88.28
87.94
87.61
63.28
62.94
62.61
62.89
62.72
62.55
6
7
87.55
87.51
87.52
87.50 nee 62.52
62.50
62.51
62.50
8 87.50 87.50 62.50
rin62.50
(c) Alternating Direction Implicit (AD/) method
Let us start with initial approximate solution of
TI,I = T 2,1 = T I,2 = T 2,2 = 100
g .ne
Iteration number is not ipdicated in the following equations for the
terms whose values are specified (on the boundary and hence, constant).
t
Step 1 : Using Ti+l,j - 4Ti,J + Ti-I,j = - T.,j+1 - T',J-I with different rows
and columns,
For j = 1 and for i = 1, T 2,1 (I) - 4T 1,1 (I) + T 0,1 -- - T 1,0 - T 1,2(0)
for i = 2, T 3,1 - 4T2,1 (I) + T 1,1 (I) -- - T 2,0 - T 2,2(0)
Solution of these two simultaneous equations in two unknowns give,
TI,I =T2,1 = 100
For j = 2 and for i=l, T 2,2(I) - 4T 1,2(I) + T 0,2 -- - T 1,1 (0) - T 1,3
for i = 2, T 3,2 - 4T . 2,2(1) + T 1,2(1) -- - T 2,1 (0) - T 2,2

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 303

Solution of these two simultaneous equations in two unknowns give,

TI,2 = T 2,2 = 66.67

Step 2: Using T'J+1 - 4T'J + T'J 1 = - TltlJ - T'1J with different rows
and columns,
For i = 1

ww for j = 2, T 1,3 - 4T 1,2

Solution of these two simultaneous equations give,
(2)
+ Tl,l
(2)
= - TO,2 - T 2,2
(1)

w.E
For i=2
T 1,1 =

and forj=l,
91.11; T 1,2 = 64.44
T2/)-4T2,1(2)+T2,O=-T1,1(1)-T3,1

asy for j = 2, T 2,3 - 4T 2,2(2) + T 2,1(2) = - T 1,2(1) - T 3 ,2

T 2,1 = 91.11; En
Solution of these two simultaneous equations give,
T 2,2 = 64.44

gin
Steps I & 2 are repeated until reasonably accurate solution for the
unknowns is obtained
(b) Numerical solution of Neumann problem and Mixed problem
eer
In order to take into effect the values of nonnal derivatives specified on
the entire (or part of the) boundary, the region is extended to some ing
imaginary mesh points outside the given region and central difference
formula for the derivative on the boundary is used to express the .ne
unknown value at the imaginary mesh point in tenns of the values on
the actual geometry. The simultaneous equations so obtained are solved
by one of the standard numerical method for computing the required
t
solution.

Example 11.2

Solve the equation

a2 u + -a22u
-2 = V 2 u = f(x, y) = 12 xy for a rectangular plate
ax ay
of 1cm x 1.5 cm with boundary conditions u = O.on x = 0; u = 0 on y=
o ; u = 3i on x = 1.5 and Un = 6x on y = 1.

Downloaded From : www.EasyEngineering.net

304 FINITE ELEMENT ANALYSIS

Solution:
y
1,3: : 2,3
Un - 6x 1 1
1,21 12,2
0--)
,- 3,2

)
u=o u=3y I, I 2, I
0, I 3, I

ww u=o
x 0,0
1,0 2.0 3,0

w.E Plate of I cm x 1.5 cm is divided into a mesh of 2 x 3. So each cell is of

0.5 em x 0.5 em.

or
asy 2 12xy
h = 0.5 and h f(x, y) = - -
4
= 3 xy

E
From the given boundary condition u =
ngi 3i, tl30=
,
0; U31 = ~ = 0.375; U32 = 3
'8 '

From the given boundary condition Un = 6x,

at UI,2 = 6 x 0.5 = 3 nee
and at U2,2 = 6 x I =6
rin
or UI,3
8U
ay
UI, I + 3
(U -U II )
Using central difference formula, - -l2' = I3'
2h
, =3
g .ne
..... (11.20)

t
=

8U (U 23 - u 2 I)
Similarly, - -22' = ' '=6
ay 2h

or ..... (11.21)
Using 5-point regular operator and retaining only unknowns on the left side,
at (l, I), tl2,1 + UI,2 - 4 UI,I = 3 xy - UI,O - UO,I
=3 x 0.5 x 0.5-0-0=0.75 ..... (11.22)

and at (2,1), til,! + U2,2 - 4 U2,1 = 3 xy - U2,O - U3,! = 3 x 1 x 0.5 - 0 - 0.375

= 1.125 ..... (11.23)

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 305

To use nonnal derivative as the boundary condition, 5-point operator with

imaginary points in the extended region is applied at (1,2) and (2,2) as follows:

At (1,2), Ut,t + U2,2 + Ut,3 - 4 Ut,2 = 3 xy - UO,2 =3 x 0.5 x I - 0 = 1.5

or substituting Ut,3 = Ut,t + 3 from eq.II.22,

..... (11.24)

ww At (2,2), Ut,2 + U2,t + U2,3 - 4 U2,2 = 3 xy - U3,2 =3 x I x I - 3 =0

w.E
or substituting

2
U2,3

U2,1
= U2,1 + 6 from eq.II.23,
+ UI,2 - 4 U2,2 = 0 - 6 =-6 ..... (11.25)

a
Eqns 11.22, 11.23, 11.24 and 11.25 can also be written in matrix form as

-4
-4
syE I 0 U1,1 0.75

ngi
0 U 2,1 1.125
-4
= -1.5
2 0 U 1,2

By solving them, we get

2 -4 U 2,2

= 0.077;
-6

nee
= 0.866; 0.191 and

rin
UI,I UI,2 U2,1

U2,2 = 1.812

11..5 FINITE VOLUME METHOD (FVM) g .ne

This method utilises control volumes and control surfaces. Finite Volume
fonnulations can be obtained either by a finite difference basis or by a finite
element basis. The control volume, in a 1-0 problem covering nodes i-I, i and
t
i + 1, for node i covers Lix to the right of node i and Lix to the left of node
2 2
i with the control surface CSt and CS 2 being located at i -! and i +! .
2 2
I I
i-- i+-
2 2
i-I i+ I
x x x x x
L6X
~4I
I CSl cv
6x
CS21
~4I
6X~
2 2

Downloaded From : www.EasyEngineering.net

306 FINITE ELEMENT ANALYSIS

(a) FVM via FDM

2
C onsl·der t he ear I·ler examp Ie d u - 2 -- 0 fior 0<x< 1
dx 2
with Dirichlet Boundary conditions u = 0 at x = 0 & at x = 1

1
~
d2U
- 2 -2 dx=O O<x<l
o dx

ww Integrating,
d ) I
(~ I - f2 dx = 0
I

w.E
dx 0 0

I (L\u) -
CSI,CS 2 L\x
Icv 2 L\x =0 at i =2

asy
Diffusion flux (::) is conserved between i-I and i through control

En
surface at i -~ or CS I and between i and i + I through control surface

at i +~ or CS 2
2
gi 2

nee
rin
This can also be written, in terms of finite differences, as

( Ul+~: u i ) _ ( u i ~:l-I ) = 2L\x

or Ui + I - 2 Ui + Ui _ I = 2 L\x
2
g.n
(b)
This is identical to the equation obtained by FDM.
FVM via FEM e t
In the same example, let u = NI UI + [1- +[:J
N2 U2 :}I U2

dU) =_(U2-UI); (dU) =_(U 3 -U 2 )

(
dx CSI h dx CS2 h

d'u
2
-2~ (~L, -(~L, -2~ (u, -2u, + u,)
dx h h2
or U3 - 2U2 + UI = 2 h 2
This is also identical to the equation obtained by FDM.

Downloaded From : www.EasyEngineering.net

CHAPTER 11 COMPUTATIONAL FLUID DYNAMICS 307

U.6 FDM vs. FEM

S.No. Finite Difference Method Finite Element Method

I. Finite difference approximation Interpolation functions and
from Taylor series expansion is polynomi~1 expansion are used
used
2. Equations are written for structured Equations are written for grids, not
grids and less complicated. Hence, necessarily structured, with nodes

ww need lesser computer time irregularly connected around the

entire domain resulting in a large
sparse matrix system for solution
3.
w.E
Treatment of governing equation
and different boundary conditions is
not uniform. The solution method is
Variational formulation employed not
only for the governing equation but
for all constraint conditions - useful

asy
tailor-made for each situation for solution stability and accuracy

En
gin
eer
ing
.ne
t

Downloaded From : www.EasyEngineering.net

ww
w.E
a
"This page s
isy
En
Intentionally Left Blank"

gin
eer
ing
.ne
t

Downloaded From : www.EasyEngineering.net

CHAPTER 12
PRACTICAL ANALYSIS
USING A SOFTWARE
ww
w.E
12.1.
asy
USING A GENERAL PURPOSE SOFTWARE

En
Many general purpose software are readily available for analysis of mechanical,
civil and aircraft structures based on FEM. Even though actual commands may

under the following three categories.

(i) Pre-processor Phase
gi
vary from one software to another, the general features can be broadly classified

nee
In this phase, data is input by the user regarding
(a) Idealised I-D, 2-D or 3-D geometric model consisting of:rin
g
• element type (discrete structure with truss, 2-D beam, 3-D beam
or pipe elements; continuum with 2-D plane stress, plane strain,
.ne
thin shell, 3-D solid or thick shell elements)
• appropriate nodal coordinates
• element attributes and element connectivity.
t
In some large components, it is also possible to create a large 2-D
or 3-D model using key points and Boolean operations on areas or
volumes. These areas or volumes can be meshed by the software
into many elements of equal or different sizes depending on user's
choice as per the expected stress distribution. The software makes
sure that the generated elements satisi)' aspect ratio norms.
(b) Properties of materials such as
• modulus of elasticity, Poisson's ratio, mass density, coefficient
of linear thermal expansion etc. for structural analysis

Downloaded From : www.EasyEngineering.net

310 FIN ITE ELEMENT ANAL YSIS

• thermal conductivity, specific heat etc. for a thermal analysis;

with options of :
- isotropic or orthotropic
- constant or temperature-dependent material data.
Some softwares have standard material properties in their database.
User need to specifY the material type only. Care must be taken to
see that the units in the material database correspond to the units of
the other data input by the user.

ww (c) Section properties like:

• area for a truss element

w.E • moment of inertia and section depth in a beam element

• thickness in a 2-D plate etc.
Some software include the facility of choosing a standard section

asy
shape such as C, I, L, H, from its database with their dimensions
specified by the user and then calculate values like area, moment of

E
inertia etc.

ngi
Material and section properties may be identified with material
numbers and section numbers so that each element can be associated
with a particular material number and a particular section number

nee
(called element attributes). In this way, material data and section
data, common to many elements, need not be input repeatedly,

rin
saving considerable time and effort in data preparation.
(d) Load particulars such as:
• distributed loads due to self weight, wind load
• concentrated or point loads g .ne
• steady-state or transient temperature distribution over the entire
model, for a structural analysis
• free-stream temperature, constant temperature on some part of
boundary for a thermal analysis etc.
t
It is also possible to analyse the same structure for different sets of
loads (defined in some software as load steps or load cases)
(e) Boundary conditions or restraints for translation or rotation OOF
at various nodes (including restraints on rigid body motion),
indication of symmetry for a structural analysis or insulated wall
for a thermal analysis etc.
Most software use consistent dimensions without any conversion of
units inside the program. It is, therefore, user's responsibility to input
data and interpret the output in appropriate units. For example, if
coordinates are input in 'mm' and loads in 'Newton', then area of cross
section should be in 'sq. mm', moment of inertia should be in 'mm4' ancl

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 311

modulus of elasticity should be in 'N/sq. mm', Correspondingly, the

displacements calculated at nodes will be in 'mm' and stresses will be in
'N/sq.mm'.
Most programs also check for validity as well as sufficiency of the
given data either in the pre-processor phase or in the solution phase
before proceeding with the actual solution to save computer time as well
as user's time. The pre-processor phase generates a data file in the
sequence required by the solution phase

ww (ii) Solution Phase

In this phase, the program uses the data file generated by the pre-

w.E
processor stage and carries out desired analysis.
Different options usually available are:
(a) static structural analysis, which calculates nodal displacements
(b)
a syE
dynamic structural analysis, which calculates natural frequencies
and mode shapes or time history response (corresponding to load
Vs time data) or response to earthquake (corresponding to

ngi
thermal analysis, which calculates nodal temperatures due to

(iii)
or convection boundary conditions.
Post-processor phase
nee
thermal conduction in a solid body with specified temperature and/

rin
The output of solution phase is a large set of nodal displacement or

g
temperature values. The post processor phase reads these values as well

.ne
as geometry data of pre-processor phase and presents in a more easily
readable form such as iso-stress or iso-temperature contours, plots of
deformed shape etc. Some software also has the facility of presenting
output for a specific combination of different loads (or load steps).
Many general purpose software, such as ANSYS, ADINA,
NASTRAN, PAFEC, NISA, PAFEC, STRUDL, etc., are commercially
t
available in the market. The specific format and sequence of input data
may vary between them. Also, modelling options as well as loads and
boundary conditions that each of the software can handle may also vary.
But, data to be input generally remains the same.

12.2 SOME EXAMPLES WITH ANSYS

An attempt is made, through simple exercises, to make the students understand

various features of a general purpose finite element software. ANSYS software
has been chosen for this purpose. The input commands or their format may
change in different versions of this software or between different software. So,

Downloaded From : www.EasyEngineering.net

3:1.2 FINITE ELEMENT ANAL YSIS

this should only be taken as a model. Till the students understand proper
method of giving necessary data and using appropriate commands, they are
advised to cross-check the results obtained using any general purpose
software with those calculate{1 by conventional strength of materials
approach. This will ensure that the data input by them is interpreted by the
software in the way they desired.
ANSYS is a general purpose software developed by Swanson Analysis
Systems Inc, USA for analysis of many different engineering problems. Input

ww sequence for solving the following problems are given here. Theoretical results
from conventional methods are also given here for verifying the output of
ANSYS and thus ensure that the features of this software are properly utilised.

w.E
These problems are not exhaustive of the features of the software but only
meant to link the theory covered so far with some typical problems.

asy
ANSYS software has been used here. for explaining some examples, with
their explicit permission.

DATA:
E
Example 12.1
Simple truss with concentrated loads
A = 25 cm 2 ngi
L I- 2 = L 2- 3 = 100 cm L2-4 = 60 cm
E = 2 x 107 N/cm 2
n
P = IOOOON
eer DIX = DIY = D3X = 0

ing
4

L-~e~_____2~______~e~3 ~~______~______~O~3
.ne
p
Case - 1
Case - 2
t
ANSYSmodel
The truss is assumed to be in X-V plane with X in the horizontal direction, Y in
the vertical direction and origin at node 1. The truss is analysed for two steps
(or load cases), with a concentrated load P acting along -Y direction at node 2
(case-I) and same load acting at node 4 (case-2). In both the cases, node 1 is
fixed in all DOF while node 3 is fixed in Y direction only (roller support). To
explain different input commands, this problem is solved 2 times with one load
acting at a time while problem 3 is solved in a single run with loads acting
separately in 2 different load steps.

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 3:1.3

The method of solving as multiple load cases gives results for the these load
cases separately, while saving computer time in the calculation of element
stiffness matrices in element local coordinate system, transformation of these
matrices to the common global coordinate system, assembling these element
matrices, and applying boundary conditions on the assembled stiffness matrix.
Input Data In ANSYS

Preferences - Structural

ww
Preprocessor -

w.E
Element type - Add - Structural link - 2D spar 1
Real constants - Add - Set No.1; Area 25
Material props - Constant Isotropic - Material No I; EX 2e7

a syE
Modeling create - Nodes - on Working plane-
(0,0),( 100,0),(200,0),( 100,60)
Elements - Thm Nodes - (1,2 ),(2,3 ),( 1,4),(2,4),(3,4)

ngi
Loads - Loads Apply - Structural Displacement - on Nodes - 1 FX, FY

Case-l Structural Force/Moment - on Nodes - 2nee -3 FY

** Case-2 **
FY Constant value
Structural Force/Moment - on Nodes - 4 rin -10000

FY Constant value
Solution - Analysis type - New Analysis - Static g
-10000

.ne
Solve current LS - Solution is done - Close
General Postproc - Plot results - Deformed shape - Def + undeformed
List results - Nodal solution - DOF solution - All DOFs
t
Node UX UY
1,2,3,4,5
Element solution - Line Elem results - Structural ELEM
EL MFORX SAXL
1,2,3,4,5
Reaction solution - All items
Node FX FY
1,3

Downloaded From : www.EasyEngineering.net

314 FINITE ELEMENT ANALYSIS

Results Obtained: Consistent with units of input data, the displacements are
read in 'cm' and the member forces/reactions are read in 'N'.

Case 1 Case 2
<hy = -0.0083834 8 2y = -0.0083834
84 = -0.0071834 84 = -0.0071834
Rly = R3Y = 5000 Rly = R3Y = 5000

ww FI2 = F2-3 = 8333.3 FI2 = F2-3 = 8333.3

w.E FI-4 = F 3-4 = -9718.3

F24 = 10000
FI -4

F2- 4 = 0
= F3-4 = -9718.3

aL sy L
Check of Results: Solving by the method of joints and applying the conditions

Fx = 0 and Fy = 0 at each joint or node.

At node 2
ECase 1

ngi
F2--4 = 10000 N
Case 2
F2--4 = 0 N
At node 4
= 9718.25 N n
F 1--4 = F3-4 = F2--4 / 2 sin 8

eer
FI--4 = F3-4 = P / sin 8
= 9718.25 N
At node 3 F2-3 = F3--4 cos 8 = 8333.3 N
R3Y = F3--4 sin 8 ing
F2-3 = F3--4 cos 8 = 8333.3 N
R3Y = F3--4 sin e
= 5000 N = 5000N
.ne
At node 2
At node 1
F I- 2 = F2-3 = 8333.3 N
R ly = FI--4 sin 8 = 5000 N
Rlx = FI--4 cos 8 _ F I_2
F I- 2 = F2- 3 = 8333.3 N
Rly = FI--4 sin 8 = 5000 N
R lx = FI--4 cos 8 - FJ-2
t
=ON =ON

Example 12.2

Stepped shaft subjected to temperature change.

DATA: Element 1 - A = 24 cm 2 ; a = 20 x 10-6 / DC; E = 1 X 107 N/cm 2
Element 2 - A = 18 cm 2 ,• a = 12 x 100-6/ DC·, E = 2 x 107 N/cm 2
Element3- A=12cm 2 ,·a=12x 1O-6 / DC ·E=2x , 107N/cm2

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 315

~T=80°C fromAtoD

A B C D
CD CD CD

ww
w.E
IlOcm
~'m+40'm->
ANSYS Model

a syE
R:!, M:!
CD 3
R" M2
CD 4 R - Real constant set

ngi
x - - - - x M _ Material properties set

Input Data In ANSYS

Preferences - Structural
Preprocessor - nee
Element type - Add - Structural link - 20 spar
Real constants - Add - Set No. I; Area 24
rin
Set No. 2; Area 18
Set No.3; Area 12
Material props - Constant Isotropic - Material No 1; EX I e7 ; ALPX 20e-6
g .ne
Material No 2 ; EX 2e7 ; ALPX 12e-6
Modeling create - Nodes - on Working plane - (0,0),(80,0),(140,0),(180,0)
Elements -Elem attributes - Real const. Set no. I , MatI No. I
t
Thru Nodes - (1,2)
Elem attributes - Real COlIst. Set no.2, MatI No.2
Thru Nodes - (2,3)
Elem attributes - Real const. Set no. 1, Mati No.1
Thru Nodes - (3,4)
Loads - Loads Apply - Structural Displacement - on Nodes - 1,4 FX
Structural Temperature - on Nodes - Pick ALL
Temp Constant value 80
Solution - Analysis type - New Analysis - Static

Downloaded From : www.EasyEngineering.net

316 FINITE ELEMENT ANALYSIS

Solve current LS - Solution is done - Close

General Postproc - Plot results - Defonned shape - Def + undefonned
List results - Nodal solution - DOF solution - Translation UX
Node UX
1,2,3,4
Element solution - Line Elem results - Structural ELEM

ww EL
1,2,3
MFORX SAXL

w.E Reaction solution - Structural force FX

Node
1,4
FX

asy
Results Obtained: OB = 0.016
RA = - Ro = 336000
Oc = 0.0176
FI =F2=F3=-336000

E
Example 12.3
ngi
Beam With Concentrated & Distributed Loads
DATA: A = 20 cm
1= 50 cm 4
2

n
L I-2 = L2- 3 = 100 cm
h = 5 cm
eer E=2 x 107N/cm2
P = 10000 N (Case-I) p = 60 N/cm (Case-2)

21
p
i n g
J~l-.L-l 1 ---x-l~1~.
Oly = 03y = 0

i ne
3 2
-=.....L..- 3

i t T
Case - 1

Input Data In ANSYS

Preferences - Structural
Case- 2

t
Preprocessor -
Element type - Add - Structural Beam - 2D Elastic 3
Real constants - Add - Set No.1; Area (A) 25 ; Moment ofInertia(IX) 50
Height of section(h) 5
Material props - Constant Isotropic - Material No 1; EX 2e7
Modeling create - Nodes - on Working plane - (0,0), (100,0), (200,0)
Elements - Thru Nodes - (1,2),(2,3)
Loads - Loads Apply - Structural Displacement - on Nodes - 1,3 FY

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 3:1.7

Structural Force/Moment - on Nodes - 2

FY Constant value - 10000
Write LS file - LS file No. I
Loads - Delete - Structural Force/Moment - on Nodes - 2 - All
Apply - Pressure - on Beams - 1,2 - Face No. I - value = - 60
Write LS file - LS file No.2

ww
Solution - Analysis type - New Analysis - Static
Solve from LS files - Start file No I; End file No.2; Increment I

w.E
General Postproc - Read First set -
Plot results - Deformed shape - Oef + undeformed
List results - Nodal solution - OOF solution - All OOFs

asy Node UX UY

E EL ngi
Element solution - Line Elem results - Structural ELEM
MFORX SAXL

Reaction solution - All items nee

Node FX FY
rin
g
Plot Ctrls - Animate - Deformed shape - Def + undeformed - Play
Read Next set - .ne
Plot results - Deformed shape - Def + undeformed
List results - Nodal solution - OOF solution - All DOFs
Node· UX UY
t
Element solution - Line Elem results - Structural ELEM
EL MFORX SAXL

Reaction solution - All items

Node FX FY

Plot Ctrls - Animate - Deformed shape - Def + undeformed - Play

Downloaded From : www.EasyEngineering.net

318 FINITE ELEMENT ANALYSIS

Results Obtained:
Case 1 Case 2
~>z = - 1.6667 D2 = - 1.25
8 1 = - 8 3 = - 0.025 8 1 =-8 3 =-0.02
R 1y = R3Y = 5000 R1y = R3Y = 6000
Check of Results:

ww D
max
PL3
=--
48 E I
= 1.6667 cm
D
max
5 P L4
=--'--
384 E I
1.25 cm

w.E
=

8
PL2
=-- 8 =--
pIJ
max 16EI max 24 E I

asy = 0.025
P
= 0.02
pL 200

E
R 1y =R 3y =-=5000N R 1y =R 3y =-60x-=6000N
2 2 2
r~~~~~~
ngi
D'fO~i~:~=:-:::::::::,,,--,.....,"'o:::::::\::::> --
~~~~<I

as plotted nee True shape of

rin deformed plot

Note: Deformed plot of the beam in both the cases will show straight lines

g.
between nodes 1 & 2 as well as between nodes 2 & 3, eventhough a cubic
displacement polynomial is used while calculating stiffness matrix of each
beam element. This is not reflective of the formulation used but a limitation
of tire post processor to represent displacement distribution between any net
two points. Mathematically, with the displacement data at two nodes, only a
straight line can be fit:

Example 12.4
Natural frequencies and mode shapes of a cantilever
DATA: A = 20 cm2 1= 50 cm4 h = 5 cm
L I - 2 = ~-3 = L3-4 = L4- 5 = 25 cm
E = 2 X 107 N/cm 2 p = 8 x 10-3 kg/cm 3 D1X = DIY = 8 = 0

~;---------;---------!---------~
Downloaded From : www.EasyEngineering.net
Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 319

Input Data In ANSYS

Preferences - Structural
Preprocessor -
Element type - Add - Structural Beam - 20 Elastic 3
Real constants - Add - Set No. I; Area(A) 20 Moment ofInertia(Ix) 50
Height of section(h) 5
Material props - Constant Isotropic - Material No I; EX 2e7; Density Se-3

ww
Modeling create-Nodes-on Working plane-
(0,0),(25,0),(50,0),(75,0),( I 00,0)
Elements - Thru Nodes - (1,2),(2,3),(3,4),(4,5)

w.E
Loads - Loads Apply - Structural Displacement - on Nodes - I ALL
Solution - Analysis type - New Analysis - Modal

asy
Analysis options - Subspace; No. of modes to extract - 4 ;
No. of modes to expand - 4
Expansion pass - on

En
Solve current LS - Solution is done - Close

gi
General Postproc - Results summary- Freq I to 4
Read First set - Plot Ctrls - Animate - Mode shape - Play!
nee
Repeat for all modes
Read Next set - Plot Ctrls - Animate - Mode shape - Play!
Results Obtained: Natural frequencies -
rin
4.4215,27.645, 77.476, 125.S0
Check of Results: In this problem, number of frequencies calculated and the
accuracy of results depend on the number of elements used in the model. g.n
where
Least frequency, (01
1
=(2 1t) ~
() = max deflection of the cantilever
e t
L4
= -p- and p = p A g is the distributed load
SEI
Then, (01 = 3.557 rad/sec
\

'Example 12.5
1-0 heat conduction through a composite wall
DATA: L I =30 cm L2=15 cm L 3=15 cm
KI = 20 W/m °c K2 = 30 W/m °c K3 = 50W/moC
7 2 2
E =2 X 10 N/cm h = 25 W/m °c T I = SOO °C ; T 5 = 20°C

Downloaded From : www.EasyEngineering.net

320 FINITE ELEMENT ANALYSIS

ANSYS Model

Fluid at Wall of Wallot Wallnt

r~ =ROO 0(' Material 1 Matcnal2 Matenal,
T - 20 "c

L - 10 em LI,M I Ll ,M 1 1.3' M3
2 3 4 5
x x x

r-
x
"
Convecllon
element
Conductioll elemcnt~

---1
ww Real constants - Area of cross section, A = I cm2. for all the 4 elements
M - Material properties set

w.E
Input Data In ANSYS

asy
Preferences - Structural
Preprocessor - Element type - Add - Thermal link - Convection;

E ngi
Real constants - Add - Set No.1; Area 25
20 conduction

Material props - Constant Isotropic - Material No I ; HF 25

nee Material No 2; KX 20
Material No 3; KX 30

Modeling create - Nodes - on Working plane- rin

Material No 4; KX 50

(0,0),(0.1,0),(0.4,0),(0.55,0),(0.7,0)
Elements - Elem attributes - Elem type I ; Mati No. I g .ne
Thru Nodes - (1,2)
Elem attributes - Elem type 2 ; Mati No.2
Thru Nodes - (2,3)
t
Elem attributes - Elem type 2 ; Mati No.3
Thru Nodes - (3,4)
Elem attributes - Elem type 2 ; Mati No.4
Thru Nodes - (4,5)
Loads - Loads Apply - Temperature - on Nodes - I 800; 5 20
Solution - Analysis type - New Analysis - Steady state
Solve current LS - Solution is done - Close

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 321

.
General Postproc ~ Plot results - Nodal solution - OOF solution - Temperatures
List results - Nodal solution - OOF solution - Temperatures
Node Temperature

Element solution - Line Elem results - Heat flow

EL Heat flow

ww Reaction solution - All items

w.E Node Heat flow

Other option
asy
E
Preprocessor - Element type - Add - Thermal link

ngi
Convection - Option - K3 - SFE command - 20 conduction
Loads - Loads Apply - Temperature - on Nodes - I 800; 5 20

Results Obtained: T2 = 304.76 °C ; T3 nee

Convection - on elem - 1 - HF 25 ; TBulk 800
= 119.05 °C ; T4 = 57.14 °C
Heat flow = 12380.95 W
rin
Check of Results

Ovemll thennal resistance, U = [ I r 15.873

g .ne
Heat flow,
1 L) L2
-+--+--+-
h K) K2 K3
L3

Q = U (T) - Ts) = 15.873 (800 - 20) = 12380.95

t
Q =h (T)- T 2)
=K) (T2 - T3 ) => T3 = 119.05 °C
L)
=K (T3 -T4 )
2
L2

=K3 (T4 -TJ

Downloaded From : www.EasyEngineering.net

322 FINITE EL E MENT ANALYSIS

Results with scaling correction factors of the program

T2 = 798.31 °C
When SFE command option is used for the convection element, effective
film coefficient, h/ ff = TB hf (where, TB is the Bulk temperature value input in
SFE command and h f is the film coefficient value input in SFE command) is
used. This results in a higher temperature drop across wall thickness and
consequently in higher thermal stresses. Design based on these temperatures

ww will be conservative.

w.E
Example 12.6

Stress concentration factor in a plate with hole

a
DATA: L = 160 cm

P = 10240 N syE
Hole dia = 20 cm E = 2
H = 100 cm
X
7
10 N /cm 2
Plate thickness, t = 0.8 cm
Poisson's ratio = 0.3

T ngi
H p+-_ _~ nee +p

rin
L
g~I .ne
ANSYS Model: Since the geometry as well as loads are symmetric about the
two major dimensions of the plate, a quarter plate can be modeled for analysis.
t
To ensure uniform loading along the small side, the load P is applied as uniform
P
pressure. p = -
Ht

I
I
I
I
I
I
I
I

~ - -------------------

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 323

Note: In the case of continuum analysis, unlike in the case of discrete

structures, accuracy of results obtained by Finite Element Method improves
in general by the use of more number of elements as well as by the use of
higher order elements such 8-noded quadrilateral or 6-noded triangle. Due to
the limitations of number of DOF in the educational version of ANSYS,
refinement of solution is not attempted here.
Input Data In ANSYS

ww
Preferences - Structural
Preprocessor -

w.E
Element type - Add - Structural solid - Quad 4 node
- option - Plane stress wlthk

asy
Real constants - Add - Set No.1; Thickness 0.8
Material props - Constant Isotropic - Material No 1; EX 2e7; NUXY 0.3

En
Modeling create - Rectangle - By 2 corners - X,Y, L, H 0,0,80,50

gi
Circle - Solid circle - X,Y, Radius
Operate - Boolean subtract - Areas
nee
0,0, 10

- Base area; Area to be subtracted

rin
Loads - Apply - Structural Displacement - Symmetry B.C. - on lines

g
Pressure - on line - constant value; 80
Meshing - Size cntrls - Global size - Element edge length 3
Mesh - Areas Free .ne
Solution - Solve current LS - Solution is done - close
General postproc - Plot results - Deformed shape - Def + Undeformed shape
Plot results - Nodal solution - DOF solution - Translation UX
t
Element solution - Stress - X-direction SX
Sorted listing - Sort Nodes - Descending order - Stress X-direction
List results - Element solution - Stress - X-direction SX
Max value 238.63 N/cm 2
Sorted listing - Sort Nodes - Descending order - Stress Y-direction
List results - Element solution - Stress - Y-direction SY
Max value 29.147 N/cm 2

Downloaded From : www.EasyEngineering.net

324 FINITE ELEMENT ANALYSIS

Alternative method of creating model

Preprocessor - Modeling create - Lines - Arcs - By Cent & Radius - (0,0),
(10,0)
Arc length in degrees 90
Key points - On Working plane - (80,0), (80,50), (0,50)
Lines - Straight line - By key points
Area - Arbitrary - By lines

ww Results Obtained:
Normal stress along X-axis (SX) Max value = 238.63 N/CI112

w .Ea
Max normal stress, in the absence of stress concentration
P 10240 x 0.8 == 160 N/cm 2
= (H -d)t
syE
Stress concentration factor
(100-20)

= 238.63 =1.4914375

ngi 160
Check of Results: For D/H = 0.2, Stress Conc. factor for a rectangular plate

nee
with circular hole = 2.51 (from Mechanical design Handbooks).

Example 12.7
Centrifugal stresses in an axisymmetric solid
E = 2 x 107 N/cm 2 rin
Poisson's ratio = 0.3 Mass density = 8 gm/cm3

g.n
Data:
Speed, N = 3000 rpm All dimensions are in mm
ANSYS Model: Since the geometry and loads are axi-symmetric, anyone
section in axis-radius plane can be modeled. Also, since the geometry as well as
loads is symmetric about the mid plane along the axis, half the flywheel can be
modeled for analysis with symmetry boundary conditions applied on the plane
et
of symmetry. ANSYS program assumes X-axis to be along the radius while Y-
axis represents the axis of symmetry. Also, the program requires that the
model be input in the right handed coordinate system (151 quadrant of X-Y
plane is more convenient). For this purpose, a quarter of the component on the
bottom side of the right half is modeled as shown. This quarter section is rotated
by 90° to represent this in the familiar way with horizontal x-axis and vertical y-
21tN
axis. Angular velocity ro =- - = 314 rad/sec is input about the axis of
60
revolution (Y-axis).

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 325

T
300 100

1
'-'-'-'-'-'-'-'-'-'-'-'
I 1 40 . f rotatIOn
._._·_.·AxISO .

ww
w.E 25 25

asy
Y:2 3
t
En
4
gi nee
5

; 1 --------------------------------------]'------..x
rin
Input Data In ANSYS
Preferences - Structural
Preprocessor - Element type - Add - Structural solid - Quad 4 node
g .ne
- option - Axisymmetric
Material props - Constant Isotropic - Material No 1
EX 2e5; NUXY 0.3; DENS 8e-3
t
Modeling - create - Key points - On Working plane
(20,0),( 150,0),( 150,25),(50,25),(50,50),(20,50)
Lines - Straight line - By key points -
(1,2),(2,3),(3,4 ),(4,5),(5,6),(6,1)
Area - Arbitrary - By lines - Pick lines 1,2,3,4,5,6
Meshing - Size cntrl- Global size - Element edge length 5
Mesh - Areas - Free - Pick area 1
Loads - Apply - Structural Displacement - Symmetry B.C. - on lines - 1
Others - Angular velocity - OMEGY about Y-axis 314

Downloaded From : www.EasyEngineering.net

326 FINITE ELEMENT ANALYSIS

Solution - Solve current LS - Solution is done - close

General postproc - Plot results - Deformed shape - Def + Undeformed shape
Nodal solution - DOF solution - Translation UX
Translation UY
Element solution - Stress - X-direction SX
V-direction SY

ww Z-direction SZ
Sorted listing - Sort Nodes - Descending order - Stress Xrdirection SX
Y -direction SY

w.E
Results Obtained: Max
Z-direction SZ
Min

asy UX(mm)
UY(mm)
7.767
0
1.477
-5.054

E SX (N/mm 2)

ngi
SY (N/mm 2)
SZ (N/mm 2)
0.337e3
0.121e3
0.543e3
-0. 11ge2
-O.145e3
0.6005e2

OBJECTIVE QUESTIONS n eer

I. ANSYS uses
(a) frontal solution (b) ing
banded matrix solution

2.
(c) 'Cramer's rule (d)
.ne
Cholesky decomposition
A single analysis with 3 similar load steps takes __ time compared to
3 analyses with single load case in each
(a)
(c)
3 times more
same
(b)
(d)
< 3 times less
not related
t
3. An analysis with 1 load step takes __ time compared to analysis with
3 similar load cases
(a) 113 times less (b) > 1/3 times less
(c) same (d) not related
4. Consistent loads are based on
(a) stress equilibrium (b) displacement continuity
(c) energy equivalence (d) force balance

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 327

5. Within elastic limit, results due to a combination of loads is same as

linear superposition of results by each of those loads
(a) always true (b) always false
(c) sometimes true (d) needs repeated analysis
6. A truss element in ANSYS is identified as
(a) line element (b) spar element

ww7.
(c) truss element (d) beam element
A uniformly distributed load on a beam is indicated in ANSYS as

w.E
(a)
(c)
udl
pressure
(b)
(d)
uvl
equivalent nodal loads
8.
asy
Uniform pressure along an edge of a plate element is specified
ANSYS as
In

(a)
(c)
E
pressure on element
pressure at each node
ngi
(b)
(d)
pressure along edge
same pressure at all nodes
9.
the mid-point appears in ANSYS as
(a) circular arc
n
Deformed shape of a simply supported beam with concentrated load at

eer
(b) triangle with max displacement at mid-point
ing
(c) parabolic arc
(d) straight line .ne
10. Deformed shape in ANSYS is drawn with
(a) actual nodal displacements
t
(b) normalised nodal displacements
(c) magnified nodal displacements
(d) reduced nodal displacements
11. Loads command in ANSYS includes
(a) loads & displacements (b) loads & stresses
(c) loads only (d) loads or displacements
12. As a default option, mesh is refined in ANSYS using
(a) g-method (b) h-method (c) p-method (d) r-method

Downloaded From : www.EasyEngineering.net

328 FIN IT EEL E MEN TAN A L Y SIS

13. Real constants in ANSYS indicate

(a) material properties (b) section properties
(c) thermal properties (d) nodal loads
14. "Attributes" in ANSYS refer to
(a) section & material properties (b) section properties
(c) material properties (d) applied loads

ww 15. Basic shapes of area I volume in ANSYS modelling are called

(a) Basics (b) Primitives

w.E
16.
( c) Primaries (d) Areas and volumes
Most FEM software reduce computer memory requirement by storing
(a)
asy
half of symmetric stiffness matrix
(b) half of banded matrix
(c)
En
stiffness matrix as a column vector
(d) complete stiffness matrix
17.
(a)
(c)
gi
Most FEM software use
displacement method
stress method nee
(b)
(d)
force method
hybrid method
18. Stresses in most FEM software are given in
rin
(a)
(c)
N/mm2
units based on input data
(b)
(d)
Pascal

g.n
user specified units
19. Most FEM software analyse a structure using
(a)
(c)
displacement method
force method
(b)
(d)
stress method
mixed method
e t
20. Displacements in most FEM software are given in
(a) mm (b) m
(c) units based on input data (d) user specified units
21. Distributed load along an edge of a plane stress element is usually
specified as
(a) pressure at nodes along the edge
(b) pressure along the edge
(c) equivalent nodal loads at the nodes on the edge
(d) different values of pressure applied at all nodes ofthe element

Downloaded From : www.EasyEngineering.net

CHAPTER 12 PRACTICAL ANALYSIS USING A SOFTWARE 329

22. A tensile distributed load along an edge of a plane stress element is

represented by _ _ at the nodes
(a) +ve pressure (b) -ve pressure
(c) +ve nodal loads (d) -ve nodal loads
23. Generalised load means
(a) load (b) displacement

ww
24.
(c) load or displacement
Attributes in ANSYS refer to
(d) temperature
for the elements

w (a)

.Ea
(b) section property set number
material & section property set numbers

25.
(d) load set number
syE
ANSYS accepts section properties set based on_
(a) element type
(c) type of load
(b)
(d) ngi
element size
type of material

nee
rin
g.n
et

Downloaded From : www.EasyEngineering.net

ww
w.E
asy
En Left Blank"
"This page is Intentionally
gin
eer
ing
. net

Downloaded From : www.EasyEngineering.net

ANSWERS

ww
CHAPTER-l

w
1.
7.
(b)
(a) .Ea
2.
8.
(c)
(b)
3.
9.
(a)
(a)
4.
10.
(a)
(a)
5.
11.
(a)
(a)
6.
12.
(c)
(c)

CHAPTER-4 syE
1.
7.
(c)
(b)
2.
8.
(d)
(c)
3.
9.
(a)
(b) ngi
4.
10.
(b)
(d)
5.
11.
(a)
(d)
6.
12.
(d)
(b)

CHAPTER-5 nee
1. (c) 2. (d) 3. (b) 4. (b) 5. (a)
rin
6. (b)
7.
13.
(b)
(c)
8.
14.
(b)
(c)
9.
15.
(a)
(d)
10.
16.
(c)
(b)
11.
17.
(a)
(d)
12.
18.
(b)
(b) g.n
19.
25.
(a)
(d)
20.
26.
(b)
(d)
21.
27.
. (c)
(d)
22.
28.
(b)
(c)
23.
29.
(a)
(a)
24.
30.
(c)
(b)
et
31. (c) 32. (c) 33. (b) 34. (a) 35. (b) 36. (a)
37. (a) 38. (d) 39. (a) 40. (a) 41. (a) 42. (b)
43. (a) 44. (b) 45. (c) 46. (b) 47. (a) 48. (b)
49. (c) 50. (c) 51. (b) 52. (c) 53. (c) 54. (b)

CHAPTER-6

1. (d) 2. (a) 3. (a) 4. (b) 5. (c) 6. (d)

7. (b) 8. (c) 9. (d) 10. (b)

Downloaded From : www.EasyEngineering.net

332 FINITE ELEMENT ANALYSIS

CHAPTER-7

1. (d) 2. (b) 3. (b) 4. (b) 5. (d) 6. (a)

7. (a) 8. (c) 9. (a) 10. (a) 11. (d) 12. (c)
13. (d) 14. (b) 15. (d) 16. (a) 17. (c) 18. (d)
19. (c) 20. (b) 21. (c) 22. (a) 23. (b) 24. (b)
25. (c) 26. (d) 27. (d) 28. (a) 29. (c)

ww CHAPTER-8

w 1.
7.
(d)

.Ea
(c)
2.
8.
(a)
(c)
3.
9.
(b)
(b)
4.
10.
(a)
(b)
5. (b) 6. (a)

CHAPTER-9
syE
1.
7.
(a)
(b)
2.
8.
(a)
(a)
3.
9. ngi
(a)
-(a)
4. (a) 5. (b) 6. (a)

CHAPTER-12 nee
1. (b) 2. (b) 3. (b) 4. (c)
rin5. (a) 6. (b)
7.
13.
(c)
(b)
8.
14.
(c)
(a)
9.
15.
(b)
(b)
10.
16.
(c)
(b)
11.
17.
(a)
(a)
12.
18. g.n
(b)
(c)
19.
25.
(a)
(a)
20. (c) 21. (a) 22. (b) 23. (c) 24. (c)
et

Downloaded From : www.EasyEngineering.net

REFERENCES FOR
ADDITIONAL READING
ww
w.E
I. asy
"Introduction to Finite Element Analysis - Theory and application" -

2. E
Harold C Martin & Graham F Carey - McGraw-Hili Book Company.

ngi
"Theory of Matrix structural analysis" - przemieniecki J S - McGraw-
Hill Book Company.
3.
n eer
"The Finite element method By Zienkiewicz 0 C" - Tata McGraw-Hili
Publishing Company Ltd.
4.
K L - Prentice Hall of India.
ing
"Numerical methods in Finite element analysis" - Bathe K J & Wilson

5. "Introduction to the Finite element method" - Desai C S & Abel J F -

Van Nostrand.
.ne
6.

7.
"The finite element method in engineering" - Singiresu S Rao -
Butterworth Heinemann.
"Introduction to Finite elements in engineering" - Tirupathi R
Chandrupatla & Ashok D Belegundu - Prentice Hall of India.
t
8. "Finite element analysis - Theory and programming" - C S Krishna
Moorthy - Tata McGraw-Hili Publishing Company Limited.
9. "An Introduction to the Finite element method" - Reddy J N - McGraw-
Hill Book Company.
10. "Concepts and applications of Finite Element Analaysis" - Robert Cook
etal - John Wiley & Sons.
II. "Fundamentals of Finite Element Analysis" - David Hutton - Tata
McGraw-Hili.

Downloaded From : www.EasyEngineering.net

334 FINITE ELEMENT ANALYSIS

12. . "A first course in the Finite Element Method" - Daryl L. Logan -
Thomson, Brooks, cole.
13. "Theory of elasticity" - Timoshenko S & Goodier J N - McGraw-Hili
Book Company.
14. "Dynamics of structures" - Walter C Hurty & Moshe F Rubinstein -
Prentice Hall of India.
15. "Computational Fluid Dynamics" - T.J .Chung - Oxford University
Press.

ww 16. "Classification of FiJ1ite Element stresses according to ASME Section

III stress categories" - W.C.Kroenke - Pressure vessels and piping;

w.E 17.
analysis and computers, ASME, June 1974.
"A Computational approach for the classification of FEM axisymmetric

asy
stresses as per ASME code" - G.Lakshmi Narasaiah et al - AS ME
Pressure Vessel & Piping Conference, 19-23 June 1988, Pittsburgh,
USA.
18.
E
"Design Check of Pressure Vessel Analysis by FEM" - G.Lakshmi

ngi
Narasaiah - National Conference on State of the art technologies in
Mechanical Engineering (NCSAME-2006) held at JNTU College of
Engineering, Hyderabad.

nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

INDEX
ww
w.E
a
Approximate method 4
A
syE Buckling
- elastic buckling 282

ngi
Bulk modulus 73
Area coordinates 175
c
Aspect ratio 138
Attributes 309 nee
Capacitance matrix 281
Assembling element matrices 94, 209
- banded matrix 209 Collocation method 8rin
Characteristic equation 239

- skyline method 212 - point collocation 8

g
- sub-domain collocation 8
.ne
Bandwidth
-
B

half-bandwith 210
Compatibility 78, 134, 139
Conductivity matrix 258
Constitutive equations
t
- minimization 211 - stress-strain relations 65
Boundary conditions 96,224 Continuum structures 27,62
- cyclic symmetry 227
Convergence 27, 133
- elimination method 96,263
Coordinate system
- multi-point constraints 98
global I structure 90, 92
- penalty approach 97,263
local I element 90, 92
- sector symmetry 227
natural/non-dimensional 169
- symmetry 226
transformation matrix 92

Downloaded From : www.EasyEngineering.net

336 Index

Critical load 282 Elements 27

elastic buckling 282 - complex 128
- constant strain triangle (CST) 129
D - linear strain triangle (LST) 167
- multiplex 128
D' Alembert's principle 237
- quadratic strain triangle (QST) 167
Damped free vibration 287 - simplex 128

ww
Damping 287 - transition 228
Elements - shapes

w.E
Deformed plots 231
- hexahedron 143, 168, 181
Degree of fixity 2,27
- quadrilateral 178, 180, 185, 187

- uncoupled 101 asy

Degrees of freedom (DOF) 61, 85 - tetrahedron 143,168,187
- triangular element 129, 175, 177, 184,273
Design codes 277
Discrete structures 3, 26, 62 E ngi
Elements - types
- 3-Dsolid 143
- axisymmetric 144
Displacement field 12,90
Displacement method 31, 32 nee
- beam 87,98, 173
- general beam 100
Distortion energy 72 - pipe 103
rin
g
Distributed loads 201 - plate bending 141
consistent 204
equivalent 201
- thick shell 143
- thin shell 143 .ne
Dynamic coupling 241
- torsion 86
- truss 86,90, 159, 170, 172
Equilibrium equations 79
t
E
Equivalent stress
Eigen values & Eigen vectors - vonMises stress 71
- matrices 35
Errors 30
Elastic Buckling 282 - discritisation 31
- critical load 282 - modelling 3 1
- mode shapes 283 - numerical 31
Elastic instability 282 Exact method 4
Extensive property 89

Downloaded From : www.EasyEngineering.net

Imlex 337

F I
Fatigue 283 Interpolation functions

Finite difference method 293 - hermite 169

- lagrange 169
Finite element method (FEM) 4, 15
Isoparametric elements 169, 170

ww
- displacement method 31
- force method 32
J
-
-
w.E
hybrid method 32
mixed method 32 Jacobian 131, 182

Finite volume method 293

Flexibility 64 asy L

E
Lagrange elements 137
Forced vibration 288
- time-history response 289
ngi Laplace equation 257,290
Least squares method 10
- response spectrum analysis 289
Fourier's law 256,258 nee
Linear analysis 63

Frequencies
Lower bounds 27

rin
- natural frequencies 311

G Mass matrix 238,240

M
g .ne
Galerkin method 5
Gaussian points 216
-
-
consistent 243
lumped 241
Mathematical model I, 2, 4
t
Geometric Isotropy 134, 136
Matrices-operations
- characteristic equation 55, 59
H - cholesky method 51
Helmholtz equation 257 - cramer's rule 45
Higher order elements 167 - direct method 44
- eigen values 55,238
Hybrid method 32
- eigen vectors 55,238
Hyper cube 135
- normalisation 239

Downloaded From : www.EasyEngineering.net

338 Index

- orthogonal 240 - mode shapes 237,239

- gauss elimination method 47 spectral matrix 239
- gausss Jordan method 43, 46, 53
Nodes 27
iterative method 53, 58
- LU factorization 48 Non-dimensional coordinates 169

- method of cofactors 42 Non-linear analysis 63

ww solution of simultaneous equations 41

Matrices - properties 36
-
-
geometric non-linearity 64
material non-linearity 64

w
-
-
-
determinant 39
inversion 41
.Ea
multiplication 37
Normalisation 239

Numerical integration 216

-
-
positive definite 55
quadratic form 38
syE o
- transpose 36,37
ngi
Optimum design 1,33

nee
Matrices-types 35 Orthotropic material 77
Matrix method 26
p
Mesh generation 214
- h-method 215 rin
g.n
Parametric elements 169
- optimum mesh 215
- p-method 2 I 5 Pascal tetrahedron 135

- r-method 215
Miner's rule 284
Pascal triangle 134

Plane strain 72
et
Mixed method 32 Plane stress 72
Modelling 220,240
Poisson's ratio 72
Modulus of elasticity 64
Postprocessor 3 II
Mohr's circle 68
Potential energy 12, 14
N Preprocessor 309
Natural coordinates 170 Principal stress 68
Natural frequencies 237 - stress at a point 79
- modal matrix 239

Downloaded From : www.EasyEngineering.net

Index 339

R - pure shear 70
- shear 66
Rayleigh-Ritz method 12
- vonMises 71
Response spectrum analysis 289
Stress categorisation 279
- forced vibration 288
- bending 278
Rigid body motion 62 - classification line 279

ww
Rigidity modulus 63
Rupture time 285
-
-
membrane 278
peak 279

w.Es Stress function 291

Stress intensity 278
S-l\: curve 284
fatigue 284 asy Stress-strain relations 74
- plane strain 76
S-tr curve 286
- creep 285
Serendipity elements 136
E ngi
- plane stress 75, 78
Structural analysis 32

n
Substructuring 230
Solution phase 311
Static condensation 230 eer
Super elements 230

Stiffness 14, 64
- infinitelzero 2 ing
Thermal analysis 32
T

- relative 3 Thermal stress 73

.ne
Strain at a point 67
-
-
hoop 145
normal 67
Time history response 289,311
- forced vibration 288
Transformation matrix 92
t
- shear 67
Transition element 228
Strain-displacement relations 73, 184
Strain energy 22 u
Strength of materials approach 2, 3 Upper bounds 27
Stress at a point 65 Usage fraction 285
- complementary shear 70 - creep 286
- max shear 71 - cumulative 286
- normal 66 - fatigue 284
- principal 68 - miner's rule 284

Downloaded From : www.EasyEngineering.net

340 Index

- cumulative 286
- fatigue 284
w
Weighted residual methods 4
- miner's rule 284
Warping function 289
v y
Variational method 4, 12

ww Young's modulus
- modulus of elasticity 63,64

w.E
asy
E ngi
nee
rin
g .ne
t

Downloaded From : www.EasyEngineering.net

Philosophies of Structural Safety and Reliability
100% (1)
Philosophies of Structural Safety and Reliability
287 pages
100 Python Programming Challenges
50% (2)
100 Python Programming Challenges
61 pages
Cook, Malkus, Plesha, Witt - Concepts & Appls of Finite Element Anal - 4a. Ed. - J. Wiley - 2002 - 733p
No ratings yet
Cook, Malkus, Plesha, Witt - Concepts & Appls of Finite Element Anal - 4a. Ed. - J. Wiley - 2002 - 733p
733 pages
Unit 5 (Trigonometric Identities)
No ratings yet
Unit 5 (Trigonometric Identities)
34 pages
Course Syllabus CE463
100% (1)
Course Syllabus CE463
2 pages
Optimisation of Strength Reduction Finite Element Method Codes For Slope Stability Analysis
No ratings yet
Optimisation of Strength Reduction Finite Element Method Codes For Slope Stability Analysis
13 pages
Finite Element Method
No ratings yet
Finite Element Method
172 pages
Geotechnical Engineering Made Easy Handwritten Classroom Notes Part-2
No ratings yet
Geotechnical Engineering Made Easy Handwritten Classroom Notes Part-2
58 pages
Application of Finite Element and Constitutive Models
100% (1)
Application of Finite Element and Constitutive Models
116 pages
05-Seismic Performance
No ratings yet
05-Seismic Performance
281 pages
2D Frame Analysis Software in Matlab - Roopak PDF
No ratings yet
2D Frame Analysis Software in Matlab - Roopak PDF
17 pages
Sdof
No ratings yet
Sdof
160 pages
Finite Element Method: U. Saravanan
No ratings yet
Finite Element Method: U. Saravanan
210 pages
MATLAB Linear Algebra
No ratings yet
MATLAB Linear Algebra
39 pages
Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay
No ratings yet
Prof. B V S Viswanadham, Department of Civil Engineering, IIT Bombay
53 pages
CH09 Variational Formulation
No ratings yet
CH09 Variational Formulation
49 pages
School of Engineering and Mathematical Sciences: by Umar Draz Ahmad
No ratings yet
School of Engineering and Mathematical Sciences: by Umar Draz Ahmad
116 pages
Settle3D: Settlement and Consolidation Analysis
No ratings yet
Settle3D: Settlement and Consolidation Analysis
56 pages
Staad Pro
No ratings yet
Staad Pro
2 pages
Fem Formulae PDF
No ratings yet
Fem Formulae PDF
12 pages
Instant Access to Wind Loading of Structures 3rd Edition John D. Holmes ebook Full Chapters
100% (10)
Instant Access to Wind Loading of Structures 3rd Edition John D. Holmes ebook Full Chapters
67 pages
Introduction Finite Element Method
No ratings yet
Introduction Finite Element Method
31 pages
Structural Dynamics Syallbus Nptel
No ratings yet
Structural Dynamics Syallbus Nptel
3 pages
Theory of Structures (2nd Edition)
No ratings yet
Theory of Structures (2nd Edition)
638 pages
CH 17
100% (1)
CH 17
21 pages
Duhamel Integral
100% (1)
Duhamel Integral
11 pages
Discontinuum Mechanics Using Finite and Discrete Elements
100% (1)
Discontinuum Mechanics Using Finite and Discrete Elements
163 pages
Foundation Ch.8
No ratings yet
Foundation Ch.8
42 pages
Solution Report For: Home My Test Bookmarks My Profile
No ratings yet
Solution Report For: Home My Test Bookmarks My Profile
13 pages
An Introduction To Matrix Structural Analysis and Finite Element Methods
No ratings yet
An Introduction To Matrix Structural Analysis and Finite Element Methods
8 pages
ST7201-Finite Element Method
No ratings yet
ST7201-Finite Element Method
14 pages
Forebay
No ratings yet
Forebay
10 pages
Pre - PHD Civil Syllabus
No ratings yet
Pre - PHD Civil Syllabus
54 pages
Lecture 1
No ratings yet
Lecture 1
69 pages
CH04 Function Approximation
No ratings yet
CH04 Function Approximation
30 pages
Introduction to Probability Second Edition Joseph K. Blitzstein 2024 scribd download
100% (2)
Introduction to Probability Second Edition Joseph K. Blitzstein 2024 scribd download
55 pages
Chapter 2 NPTEL Wave Propagation
100% (1)
Chapter 2 NPTEL Wave Propagation
32 pages
Magneto-Rheological Dampers: Presented By:-P.Mukesh 13071A0339
No ratings yet
Magneto-Rheological Dampers: Presented By:-P.Mukesh 13071A0339
27 pages
Development of A Structural Health Monitoring System For Bridges
No ratings yet
Development of A Structural Health Monitoring System For Bridges
2 pages
Chapter 5. Two-Dimensional Finite Elements: Plane Stress (Thin Members) Out-Of-Plane Normal and Shear Stress Are Zero X y
No ratings yet
Chapter 5. Two-Dimensional Finite Elements: Plane Stress (Thin Members) Out-Of-Plane Normal and Shear Stress Are Zero X y
16 pages
3CL1201 Finite Element Method For Structural Engineering
No ratings yet
3CL1201 Finite Element Method For Structural Engineering
38 pages
Nonlinear FE Analysis 2009 1
100% (1)
Nonlinear FE Analysis 2009 1
40 pages
Finite Element Method (10ME64)
No ratings yet
Finite Element Method (10ME64)
72 pages
LSPP eZSetup1
No ratings yet
LSPP eZSetup1
249 pages
Mathcad Functions
No ratings yet
Mathcad Functions
3 pages
Finite Element Method
From Everand
Finite Element Method
Gouri Dhatt
1/5 (1)
Towards A Unified Soil Mechanics Theory: The Use of Effective Stresses in Unsaturated Soils, Revised Edition
From Everand
Towards A Unified Soil Mechanics Theory: The Use of Effective Stresses in Unsaturated Soils, Revised Edition
Eduardo Rojas
No ratings yet
Basic Finite Element Method as Applied to Injury Biomechanics Readable Ebook Download
100% (4)
Basic Finite Element Method as Applied to Injury Biomechanics Readable Ebook Download
17 pages
Structural Analysis TH PDF
No ratings yet
Structural Analysis TH PDF
22 pages
Greenhouse Technology Management 2nd Edition Kandukuri Radha Manohar 2024 Scribd Download
100% (1)
Greenhouse Technology Management 2nd Edition Kandukuri Radha Manohar 2024 Scribd Download
67 pages
TOM
No ratings yet
TOM
15 pages
Structural Ananysis
No ratings yet
Structural Ananysis
16 pages
Greenhouse Technology Management 2nd Edition Kandukuri Radha Manohar - The ebook in PDF/DOCX format is ready for download now
100% (2)
Greenhouse Technology Management 2nd Edition Kandukuri Radha Manohar - The ebook in PDF/DOCX format is ready for download now
57 pages
Finite Element Analysis G. Lakshmi Narasaiah
No ratings yet
Finite Element Analysis G. Lakshmi Narasaiah
349 pages
CS502 Handouts
No ratings yet
CS502 Handouts
184 pages
Gopal_Krishna_dis-Buckling
No ratings yet
Gopal_Krishna_dis-Buckling
138 pages
Operation and Control in Power Systems PDF
100% (1)
Operation and Control in Power Systems PDF
421 pages
CS502 Highlighted Handout
No ratings yet
CS502 Highlighted Handout
184 pages
Prestandard For: Performance-Based Design
No ratings yet
Prestandard For: Performance-Based Design
122 pages
Active Learning
100% (3)
Active Learning
116 pages
Compliant Mechanisms 1st Edition pdf download
100% (1)
Compliant Mechanisms 1st Edition pdf download
50 pages
Hand Out
50% (2)
Hand Out
162 pages
Leadership PDF
No ratings yet
Leadership PDF
1 page
What Is %G?: What Is Acceleration? Peak Acceleration? Peak Ground Acceleration (PGA) ? What Is Spectral Acceleration (SA) ?
No ratings yet
What Is %G?: What Is Acceleration? Peak Acceleration? Peak Ground Acceleration (PGA) ? What Is Spectral Acceleration (SA) ?
9 pages
Concrete Bridge Construction 5-393.350
No ratings yet
Concrete Bridge Construction 5-393.350
42 pages
Eciv 720 A Advanced Structural Mechanics and Analysis: Isoparametric CST
No ratings yet
Eciv 720 A Advanced Structural Mechanics and Analysis: Isoparametric CST
84 pages
Fundamentals of Finite Element Methods: Pdhonline Course S272 (4 PDH)
No ratings yet
Fundamentals of Finite Element Methods: Pdhonline Course S272 (4 PDH)
18 pages
4 - Precast Concrete Bridges PDF
No ratings yet
4 - Precast Concrete Bridges PDF
30 pages
24年12月亚太数学校对
No ratings yet
24年12月亚太数学校对
36 pages
Density Based Clustering
No ratings yet
Density Based Clustering
22 pages
Mathematics For Junior High Schools
100% (1)
Mathematics For Junior High Schools
34 pages
Python For Data Science
No ratings yet
Python For Data Science
45 pages
20METM054 Mohammad Azam Shahab - ENDSEM_20_21
No ratings yet
20METM054 Mohammad Azam Shahab - ENDSEM_20_21
2 pages
Graphical Solution - Mohr'S Stress Circle
No ratings yet
Graphical Solution - Mohr'S Stress Circle
4 pages
DSTL Quantum PDF by Motivation Bank
100% (1)
DSTL Quantum PDF by Motivation Bank
259 pages
Recurrent Neural Networks and Robust Time Series Prediction
No ratings yet
Recurrent Neural Networks and Robust Time Series Prediction
15 pages
Po-Jui Huang and Duan-Yu Chen Department of Electrical Engineering, Yuan Ze University, Chung-Li, Taiwan Dychen@saturn - Yzu.edu - TW, S970561@mail - Yzu.edu - TW
No ratings yet
Po-Jui Huang and Duan-Yu Chen Department of Electrical Engineering, Yuan Ze University, Chung-Li, Taiwan Dychen@saturn - Yzu.edu - TW, S970561@mail - Yzu.edu - TW
5 pages
Industrial Engineering Article
No ratings yet
Industrial Engineering Article
2 pages
Tutorials: CS140 - Computer Programming 1
No ratings yet
Tutorials: CS140 - Computer Programming 1
33 pages
CL 8 Ws Rational Numbers
No ratings yet
CL 8 Ws Rational Numbers
2 pages
One Million Digits of Pi On One Page! - (Plus Guides and Information) - Pi Day
No ratings yet
One Million Digits of Pi On One Page! - (Plus Guides and Information) - Pi Day
151 pages
Examples3 PDF
No ratings yet
Examples3 PDF
2 pages
Sample Cover Letter
No ratings yet
Sample Cover Letter
2 pages
Greedy Solution To The Fractional Knapsack Prob
No ratings yet
Greedy Solution To The Fractional Knapsack Prob
3 pages
Lesson Plan Mathematics
100% (1)
Lesson Plan Mathematics
5 pages
College of Liberal Arts & Sciences: SOCIAL STATISTICS: Learning Module No. 4
No ratings yet
College of Liberal Arts & Sciences: SOCIAL STATISTICS: Learning Module No. 4
6 pages
6-1 Solving One-Step Inequalities Part 2
No ratings yet
6-1 Solving One-Step Inequalities Part 2
13 pages
Session 1-Set Theory
No ratings yet
Session 1-Set Theory
49 pages
Read The Instructions/Directions Carefully. Use Another Sheet of Paper For Rough Draft Solutions. Use Black or Blue Ballpen Only
No ratings yet
Read The Instructions/Directions Carefully. Use Another Sheet of Paper For Rough Draft Solutions. Use Black or Blue Ballpen Only
3 pages
The LA Theorem
No ratings yet
The LA Theorem
4 pages
CS 332: Algorithms: Final Exam
No ratings yet
CS 332: Algorithms: Final Exam
26 pages
Full download Foundations of Complex Analysis in Non Locally Convex Spaces Function Theory Without Convexity Condition 1st Edition Aboubakr Bayoumi (Eds.) pdf docx
100% (2)
Full download Foundations of Complex Analysis in Non Locally Convex Spaces Function Theory Without Convexity Condition 1st Edition Aboubakr Bayoumi (Eds.) pdf docx
71 pages
Model Paper 7 em
No ratings yet
Model Paper 7 em
3 pages
Legends in Maths
100% (1)
Legends in Maths
6 pages
GE 312 Lecture 6a
No ratings yet
GE 312 Lecture 6a
47 pages
L01 Your Turn - Evaluating Algebraic Expressions
No ratings yet
L01 Your Turn - Evaluating Algebraic Expressions
4 pages