You are on page 1of 456

ELEMENTARY

FINITE

ELEMENT

METHOD
C.S.

DESAI

B&l

D45e
2035801
Desai
element
finite
Lementary
15.624

jthod

Mi

Ou.

Q
library
public
AMD AU*N COUNT*,
SORT \WAYN6

INO.

ALLEN COUNTY PUBLIC LIBRARY

4JHB|||a|i|_.

ELEMENTARY
FINITE ELEMENT
METHOD

CHANDRAKANT S. DESAI
Professor, Virginia Polytechnic Institute

and State

University, Blacksburg, Virginia

PRENTICE-HALL,

INC., Englewood

Cliffs,

New Jersey 07632

Library of Congress Cataloging

Desai,

Chandrakant

Elementary

finite

(Civil engineering

Bibliography;

in

Publication Data

S.

element method.

and engineering mechanics

series)

p.

Includes index.
1.

Finite element method.

TA347.F5D47

I.

Title.

78-10389

624'. 171

ISBN 0-13-256636-2

Civil Engineering

N. M.

and Engineering Mechanics

Newmark and W.

J.

Series

Hall, Editors

w.

1979 by Prentice-Hall, Inc.,


Englewood Cliffs, N.J. 07632

All rights reserved.

may

No

part of this

book

be reproduced in any form or

by any means without permission


from the publisher.

in writing

Printed in the United States of America

10

987654321

Prentice-Hall
Prentice-Hall
Prentice-Hall
Prentice-Hall
Prentice-Hall
Prentice-Hall

International, Inc., London


of Australia Pty. Limited, Sydney
of Canada, Ltd., Toronto
of India Private Limited, New Delhi
of Japan, Inc., Tokyo
of Southeast Asia Pte. Ltd., Singapore

Whitehall Books Limited,

Wellington,

New Zealand

2G3S8Q1
To

My

Parents,

Maya and

Sanjay

Digitized by the Internet Archive


in

2012

http://archive.org/details/elenrientaryfiniteOOdesa

CONTENTS

PREFACE

1.

XI

INTRODUCTION
Basic Concept

Process of Discretization

and Laws 13
Cause and Effect 14
Review Assignments 14

Principles

References

2.

16

STEPS IN THE FINITE ELEMENT


Introduction

17

General Idea
References

34

17

METHOD

17

Contents

vi

3.

ONE-DIMENSIONAL STRESS DEFORMATION


Introduction

35

35

Explanation of Global and Local Coordinates

36

Local and Global Coordinate System for the

One-Dimensional Problem
Interpolation Functions

38

41

Relation Between Local and Global Coordinates

Requirements for Approximation Functions


Stress-Strain Relation

Principle of

Minimum

Expansion of Terms
Integration

46
Potential Energy

47

52

53

Approach (for assembly)


Method 58

Potential Energy

Direct Stiffness

Boundary Conditions 59
Strains and Stresses 64
Formulation by Galerkin's Method
Computer Implementation 76
Other Procedures for Formulation

76

Complementary Energy Approach


Mixed Approach 79
Bounds 83

76

References

55

67

Advantages of the Finite Element Method


Problems 85

4.

85

91

ONE-DIMENSIONAL FLOW
Problems

93

102

Bibliography

102

ONE-DIMENSIONAL TIME-DEPENDENT FLOW


(Introduction to Uncoupled and Coupled Problems)

Uncoupled Case 103


Time-Dependent Problems 108
One-Dimensional Consolidation
Computer Code 124
Problems 130
References

42

43

132

[8, 9]

121

103

Contents

6.

COMBINED COMPUTER CODE FOR ONEDIMENSIONAL DEFORMATION, FLOW, AND


TEMPERATURE/CONSOLIDATION
Philosophy of Codes
Stages

Listing

134

144

and Samples of Input/Output

Beam-Column

177
189

Other Procedures for Formulation


Problems 197
References

192

200

ONE-DIMENSIONAL MASS TRANSPORT


Introduction

References

Bibliography

202

202

Finite Element Formulation

202

209
209

ONE-DIMENSIONAL OVERLAND FLOW


Introduction

References

211

211

Approximation for Overland and Channel Flows


Finite Element Formulation
213

10.

172

172

Physical Models

9.

145

BEAM BENDING AND BEAM-COLUMN


Introduction

8.

133

134

Problems

7.

vii

213

222

ONE-DIMENSIONAL STRESS WAVE PROPAGATION


Introduction

224

Element Formulation
Problems 234
Finite

References

Bibliography

235
235

225

224

Contents

viii

11.

TORSION

237

Introduction

237

Triangular Finite Elements

238

Finite Element Formulation

240

Comparisons of Numerical Predictions and Closed

Form

Solutions

252

Approach 254
Review and Comments
Hybrid Approach 269
Mixed Approach 284
Stress

Static Condensation
Problems 293

References

12.

268

291

297

OTHER FIELD PROBLEMS: POTENTIAL, THERMAL,


AND FLUID FLOW
Introduction
Potential

299

299

Flow

300

Element Formulation 302


Stream Function Formulation 314
Thermal or Heat Flow Problem 321
Seepage 323
Finite

Electromagnetic Problems

327

Computer Code FIELD-2DFE


Problems
References

332

Bibliography

13.

328

328

332

TWO-DIMENSIONAL STRESS-DEFORMATION
ANALYSIS
Introduction

333
Plane Deformation
Finite

333

Element Formulation

Computer Code
Problems
References

368
371

354

338

333

Contents

14.

MULTICOMPONENT SYSTEMS: BUILDING FRAME


AND FOUNDATION
Introduction

Computer Code

373

382

Transformation of Coordinates
Problems 392

15.

389

392

PRELUDE TO ADVANCED STUDY AND APPLICATIONS


Theoretical Aspects

Bibliography

394

395

396

VARIOUS NUMERICAL PROCEDURES;


SOLUTION TO BEAM BENDING PROBLEM

399

SOLUTION OF SIMULTANEOUS
EQUATIONS

400

3.

PHYSICAL MODELS

419

4.

COMPUTER CODES

424

Appendix

1.

Appendix

2.

Appendix

Appendix

INDEX

372

372

Various Components

References

jx

429

PREFACE

The finite element method has gained tremendous attention and popularThe method is now taught at most universities and colleges, is researched
extensively, and is used by the practicing engineer, industry, and government
agencies. The teaching of the method has essentially been concentrated at
the postgraduate level. In view of the growth and wide use of the method,
however, it becomes highly desirable and necessary to teach it at the underity.

graduate

level.

There are a number of books and publications available on the finite


element method. It appears that almost all of them are suitable for the
advanced student and require a number of prerequisites such as theories of
constitutive or stress-strain laws, mechanics,

and variational

calculus.

Some

of the introductory treatments have presented the method as an extension of


matrix methods of structural analysis. This viewpoint
necessary, since the finite element

maturity and generality.

It

method has reached a

may no

longer be

significant level

of

has acquired a sound theoretical basis, and in

has been established as a general procedure relevant to engineering and


mathematical physics. These developments permit its teaching and use as a
general technique from which applications to topics such as mechanics,
itself

structures, geomechanics, hydraulics,

as special cases. It

is

and environmental engineering


method be treated

therefore essential that the

arise

as a

general procedure and taught as such.

This book
sufficiently

is

intended mainly for the undergraduate.

elementary so that

it

Its

approach

is

can be introduced with the background of


xi

Preface

xii

essentially

undergraduate subjects. At the same time, the treatment

enough so

that the reader or the teacher interested in various topics such as

broad

is

stress-deformation analysis, fluid and heat flow, overland flow, potential flow,

time-dependent problems, diffusion, torsion, and wave propagation can use

and teach from

it.

of the method
and provides a distinct and
element method at an elementary

The book brings out the

intrinsic nature

that permits confluence of various disciplines

rather novel approach for teaching the finite


level.

Although the book

is

intended mainly for the undergraduate,

it

can

be used for the fresh graduate and the beginner with no prior exposure to
the finite element method.

The

prerequisites for understanding the material

be undergraduate mathematics, strength of materials and undergraduate

will

courses in structures, hydraulics, geomechanics and matrix algebra. Intro-

ductory knowledge of computer programming

The

text

written in such a

is

way

that

is

desirable but not necessary.

no prior knowledge of variational

The derivations are presented through the use of


Over a period of the last five years or so, the author has
taught, based on these prerequisites, an undergraduate course and a course
for user groups composed of beginners. This experience has shown that
principles

is

necessary.

differential calculus.

undergraduates or beginners equipped with these prerequisites, available to

them

in the

undergraduate curricula at most academic institutions, can under-

stand and use the material presented in this book.

The

first

chapter presents a rather philosophical discussion of the

finite

element method and often defines various terms on the basis of eastern and
western concepts from antiquity. The second chapter gives a description of
the eight basic steps. Chapters 3-5 cover one-dimensional problems in stress-

deformation analysis and steady and time-dependent flow of heat and

The fundamental

mon

method

fluids.

by showing the comcharacteristics of the formulation for these topics and by indicating the

fact that their

generality of the

is

illustrated

governing equations are essentially similar. The generality

further established by including a

computer code

in

is

chapter 6 that can solve

the three problems in chapters 3-5.

Understanding and using the


the use of the computer.

teaching of the method

It is

may

finite

element method

is

closely linked with

the belief of the author that strictly theoretical

not give the student an idea of the details and

the ranges of applicability of the technique. Consequently this text endeavors

and simultaneously with the theoretical


and understanding of computer codes. The code in chapter 6 is thoroughly documented and detailed so that it can be used and understood without difficulty. Moreover, a number of rather simple codes are
introduced in the later chapters and their applications are included. Details
of these codes, designed for the beginner, are given in appendix 4. It is recommended that these or other available codes be used by the student while
to introduce the student, gradually

teaching, to the use

learning various topics in chapters 7-14.

Preface

xiii

Chapter 7 introduces the idea of higher-order approximation for the


problem of beam bending and beam-column. One-dimensional problems in
mass transport (diffusion-convection), overland flow due to rainfall, and
wave propagation are covered in chapters 8, 9 and 10, respectively. These
problems illustrate, by following the general procedure, formulations for
different categories of time-dependent problems.

Chapters 11-14 enter into the realm of two-dimensional problems. The


chapter on Torsion (chapter 11) and Other Flow Problems (chapter 12) have

been chosen because they involve only one degree-of-freedom

at a point.

Chapters 13 and 14 cover two-dimensional stress-deformation problems


involving two and higher degrees-of-freedom at a point.

The text presents the finite element method by using simple problems. It
must be understood, however, that it is for the sake of easy introduction that
we have used relatively simple problems. The main thrust of the method, on
the other hand, is for solving complex problems that cannot be easily solved
by the conventional procedures. In order to emphasize this and to show the
reader what kind of complex factors can be handled, chapter 1 5 includes a
rather qualitative description of the advanced study and applications of the
method. Here, a number of factors and aspects that are not covered in chapters 1-14 are stated and references are given for a detailed study.
For a thorough understanding of the finite element method, it is essential
that the student perform hand calculations. With this in mind, most chapters
include a number of problems to be solved by hand calculations. They
also include problems for home assignments and self-study.
The formulations have been presented by using both the energy and
residual procedures. In the former, the potential, complementary, hybrid,
and mixed procedures have been discussed. In the residual procedures, main
attention has been given to Galerkin's method. A number of other residual
methods are also becoming popular. They are described, therefore, in appendix I, which gives descriptions, solutions and comparisons for a problem by
using a number of methods: Closed form, Galerkin, collocation, subdomain,
least squares, Ritz, finite difference, and finite element.
Formulations by the finite element method usually result in algebraic
simultaneous equations. Detailed description of these methods is beyond the
scope of this book. Included in appendix
to the

commonly used

direct

and

iterative

2,

however, are brief introductions

procedures for solution of algebraic

simultaneous equations.
Physical models can help significantly in the understanding of various
concepts of the method. Appendix 3 gives the descriptions of some physical

models. Appendix 4 presents details of a number of computer codes relevant


to various topics in the text.

one or two undergraduate courses. The second


course may overlap with or be an introductory graduate course. Although a

The book can be used

for

Preface

xiv

number of topics have been covered

in the

book, a semester or quarter course

could include a selected number of topics. For instance, a quarter course can

and then one or two topics from the remaining chapters.


mechanics and stress-deformation analyses, the
topics can be Beam Bending and Beam-Column (chapter 7), and Two-Dimensional Stress Deformation (chapter 13). If time is available (in the case of a
semester course), chapter 10 on One-Dimensional Wave Propagation, chapter
11 on Torsion and/or chapter 14 on Multicomponent Systems can be added.
A class oriented toward field problems and hydraulics can choose one or
more of chapters 8, 9, 11, and 12 in addition to chapters 1-6.
Thanks are due to Y. Yamada, University of Tokyo; Peter Hoadley,
Vanderbilt University; William J. Hall, University of Illinois at Urbana;
E. L. Wilson, University of California at Berkeley; John F. Abel, Cornell
University and my colleagues S. Sture, T. Kuppusamy and D. N. Contractor
for reading the manuscript and for offering useful comments and suggestions.
A number of my students helped in solutions of some of the problems I
would like to express special appreciation to John Lightner for his assistance
in implementing some of the computer solutions.
I realize that it is not easy to write at an elementary level for the finite
element method with so many auxiliary disciplines. The judgment of this book
cover chapters

For a

1-6,

class interested in

is

better left to the reader.

All natural systems are essentially continuous or interconnected,

influenced by a large

number of parameters.

we must understand all


we make approximations, by

and are

In order to understand

such a system,

the parameters. Since this

possible

selecting only the significant of

is

not

them and neglecting the others. Such a procedure allows understanding


of the entire system by comprehending its components taken one at a
time. These approximations or models obviously involve errors, and we
strive continuously to improve the models and reduce the errors.

Chandrakant

S.

Desai

INTRODUCTION

BASIC CONCEPT
current form the finite element (FE)

method was formalized by civil


The method was proposed and formulated previously in different
manifestations by mathematicians and physicists.
The basic concept underlying the finite element method is not new The
principle of discretization is used in most forms of human endeavor. Perhaps
In

its

engineers.

the necessity of discretizing, or dividing a thing into smaller manageable

from a fundamental limitation of human beings in that they


cannot see or perceive things surrounding them in the universe in their entirety
things, arises

or totality. Even to see things immediately surrounding us,


several turns to obtain

we

a.

we must make

jointed mental picture of our surroundings. In other

around us into small segments, and the final


one that simulates the real continuous surroundings. Usually such jointed views contain an element of error.
In perhaps the first act toward a rational process of discretization, man

words,

discretize the space

assemblage that we visualize

is

divided the matter of the universe into five interconnected basic essences

(Panchmahabhutd), namely, sky or vacuum,

air,

added to them perhaps the most important of all,

water, earth, and


time,

by singing

Time created beings, sky, earth,


Time burns the sun and time will bring
What is to come. Time is the master
of everything

^he number

[l].

within brackets indicates references at the end of the chapter.

fire,

and

Chapter 1

Introduction

<**J-W

Figure 1-1 'Discretization' of universe, (a) 'Discretized' galaxy. (From


Smithsonian, Nov. 1976, reproduced by permission of Mr. Clifton

(From Smithsonian, Aug. 1976, by

Line.) (b) 'Discretized' earth.

permission of Dr. Athelstan Spilhaus.)

We
finite)

conceived the universe to be composed of an innumerable (perhaps

number of

solar systems, each system

composed of

its

own

stars,

our solar system we divided the planet


earth into interconnected continents and oceans. The plate of earth we live

planets,

on

is

and galaxies

[Fig. 1-1 (a)]. In

composed of interconnected

When man

finite plates [Fig. l-l(b)].

started counting, the

To compute

numeral system evolved.

drew
and outside

the circumference or area of a circle, early thinkers

polygons of progressively increasing and decreasing


the circle, respectively, and found the value of

size inside

to

a high degree of

made of

blocks or elements

accuracy.

In

(civil)

engineering

(Fig. 1-2). Figure 1-3

we

started buildings

shows how early constructions by man, made of mud,

naturally cracked into blocks or elements

building clue.

thus nature provided us with the

Chapter 1

Introduction

\_

IN!

_/

T_

.44

r
r

r
r

+-

r
r

Block
or element

r
r
r
r~

Figure 1-2 Building column composed of blocks or elements.

When engineers surveyed tracts of land, the track was divided into

smaller

was surveyed individually [Fig. l-4(a)]. The connecting of the individual surveys provided an "approximate" survey of the
entire tract. Depending on the accuracy of the survey performed, a "closing
error" would be involved. In aerial photography a survey of the total area is
obtained by matching or patching together a number of photographs. The
tracts,

and each small

tract

aberrations or discontinuities at the interfaces or junctions of the individual

photographs are shown in Fig.

For

stress analysis

l-4(b).

of modern framed structures in

civil

engineering clas-

methods such as slope deflection and moment distribution were used.


The structure was divided into component elements, each component was
examined separately, and (stiffness) properties were established (Fig. 1-5).
The parts were assembled so that the laws of equilibrium and physical condition of continuity at the junctions were enforced.
Although a system or a thing could be discretized in smaller systems,
components, or finite elements, we must realize that the original system itself
is indeed a whole. Our final aim is to combine the understandings of individual
components and obtain an understanding of the wholeness or continuous
sically,

nature of the system. In a general sense, as the

modern

scientific

thinking

Chapter 1

Introduction

Figure 1-3 'Discretized' hut due to cracking. {From Smithsonian,


Aug. 1976, by permission of Dr. Athe Is tan Spilhaus.)

recognizes,

which the Eastern philosophical and metaphysical concepts had

recognized in the past,

all

systems or things are but parts of the ultimate

continuity in the universe!

The foregoing

many

activities

of

abstract

man

and engineering examples make us aware of the


on discretization.

that are based

Chapter

Introduction

Closure error

Survey
of smaller
areas

(a)

^^5":

#
/>'

(b)
Figure 1-4 'Discretization' in surveying,

(a)

Closure error in survey

of subdivided plot, (b) Patching of aerial photographs to obtain

assemblage of a survey.

Figure 1-5 'Discretization' of engineering structure, (a) Actual structure, (b) 'Discretized' structure, (c) Idealized

(a)

(b)

one-dimensional model.

(c)

PROCESS OF DISCRETIZATION
Discretization implies approximation of the real

and the continuous.

We

use

a number of terms to process the scheme of discretization such as subdivision,


continuity,

potential,

compatibility,

minimum

convergence, upper and lower bounds, stationary

residual,

and

error.

As we

shall see later,

although these

terms have specific meanings in engineering applications, their conception

we discuss some of these


and aspects have been adopted from Russell [2].

has deep roots in man's thinking. In the following


terms; a

number of

figures

Subdivision

Zeno argued that space is finite and infinitely divisible and that for things
must have magnitudes. Figure l-6(a) shows the concept of finite

to exist they

Figure 1-6 Finiteness and


nitely divisible triangle.

divisibility,

{From Ref.

[2],

(a) Infinite space,

Limited, London.)

(a)

(b) Infi-

by permission of Aldus Books

(b)

Chapter 1

Introduction

space If the earth were contained in space, what contained the space in turn
:

[2]? Figure l-6(b) illustrates this idea for the divisibility

number of component

of a triangle into a

triangles.

Continuity
is made up of divisible elements.
between any two points in a line, and
there exist other moments between two moments in a period of time. Therefore, space and time are continuous and infinitely divisible [2], and things

Aristotle said that a continuous quantity

For

instance, there exist other points

are consecutive, contiguous,

and continuous

(Fig. 1-7).

3
(b)

Figure 1-7 Concepts of continuity, (a) Consecutive, (b) Contiguous,


(c)

Continuous. {From Ref.

[2],

by permission of Aldus Books

Limited, London.)

These ideas of finiteness, divisibility, and continuity allow us to divide


continuous things into smaller components, units, or elements.
Convergence

For evaluating the approximate value of n, or the area of a circle, we


can draw polygons within [Fig. l-8(a)] and around [Fig. l-8(b)] the circle.
As we make a polygon, say, the outside one, smaller and smaller, with a

number of sides, we approach the circumference or the area of the


This process of successively moving toward the exact or correct solution can be termed convergence.

greater
circle.

Chapter 1

The

Introduction

idea

is

analogous to what Eudoxus and Archimedes called the method

of exhaustion. This concept was used to find areas bounded by curves; the
available space was filled with simpler figures whose areas could be easily

Archimedes employed the method of exhaustion for the parabola


by inscribing an infinite sequence of smaller and smaller
one can find the exact numerical formula for the parabola. Indeed,
practitioner of the finite element method soon discovers that the

calculated.

(Fig. 1-9); here


triangles,

an active

pursuit of convergence of a numerical procedure

is

indeed fraught with

exhaustion
Figure 1-8 Convergence and bounds for approximate area of circle,
(a)

Polygons inside

circle, (b)

gence for approximate area of

Polygons outside

circle, (c)

Conver-

circle,

A(5) = 1.322

A(4) = 1.125 units

A(6) = 1.462

A(7) = 1.540

A(8) = 1.591

A(9) = 1.628

(a)


Chapter 1

Introduction

10

'^^

^X

l\
1

J\

\
:>^-

\l

A(4) = 2.250

A(5) = 2.043

A(6) = 1.945

A(7) = 1.896

A(9) = 1.843

A(8) = 1.864

(b)

Figure 1-8 (Continued)

In the case of the circle (Fig.


or outside polygon

is

1-8),

convergence implies that as the inside


number of sides, we

assigned an increasingly greater

approach or converge to the area of the circle. Figure l-8(c) shows the plots
of successive improvement in the values of the area of the circle from the two
procedures polygons of greater sides drawn inside and outside. We can see
that as the number of sides of the polygons is increased, the approximate
:

areas converge or approach or tend toward the exact area.

11

Inside polygons

1.0

0.0

Number

10

11

12

13

of sides

(c)

Figure 1-8 (Continued)

Physical models:

The student can prepare

board or

models to

plastic)

illustrate

pictorial or physical (card-

convergence from the example of

the area of the circle.

Bounds

Depending on the course of action that we take from within or from


without, we approach the exact solution of the area of the circle. However,
the value from each method will be different. Figure l-8(c) shows that convergence from within the circle gives a value lower than the exact, while that
from outside gives a value higher. The former yields the lower bound and the
latter,

the upper bound. Figure 1-10 depicts the process of convergence to

upper and lower bound solutions.

12

Chapter 1

Introduction

Parabola

Figure 1-9 Concept of convergence or exhaustion. {From Ref.

by permission of Aldus Books Limited, London.)

Figure 1-10 Concept of bounds.

Upper
bound

Exact solution-

Lower
bound

[2],

Error
It

should be apparent that discretization involves approximation. Consewhat we obtain is not the exact solution but an approximation to

quently,

that solution.

The amount by which we

differ

can be termed the

error.

For

example, the areas (or perimeters) of the polygons inscribed in the circle (Fig.
1-8) are

always

less

than the area (or perimeter) of the

circle,

and the areas

(or perimeters) of the circumscribed polygons are always greater than the

area (or perimeter) of the circle.

and the exact perimeter


the

number of

is

sides of the

The

difference between the approximation

the error, which becomes smaller and smaller as

polygons increases.

We

can express error in the

area as

A*

= the exact area, A =

where A*

- A = e,

(1-1)

approximate area, and e

= error.

PRINCIPLES AND LAWS

To

describe the behavior of things or systems around us,

laws based on principles.

we need

to establish

A law can be a statement or can be expressed by a

mathematical formula. Principles have often been proposed by intuition,


hypothesized, and then proved.

Newton's second law

states

F= ma

(l-2a)

F=mp

d-2b)

or

where

F=

force,

m=

mass, and a

displacement u with respect to time

body is

in

t.

= acceleration
The

principle

or second derivative of

is

that at a given time the

dynamic equilibrium and a measure of energy contained

in the

body

assumes a stationary value.

simple statically loaded linear elastic column [Fig. 1-1

1(a)]

follows the

principle that at equilibrium, under a given load and boundary constraints,


the potential energy of the system assumes a stationary (minimum) value and

that the equation governing displacement

where v

= length

is

= ^>

displacement in the vertical y direction,

of the column,

A =

(1-3)

AE

cross-sectional area,

P = applied load, L
E = modulus of

and

elasticity.

13

14

Chapter

Introduction

= area of

cross section

4TP

tU^

(b)

(a)

Figure 1-11 Structures subjected to loads (causes), (a) Column,


(b)

Body

or structure.

CAUSE AND EFFECT


The essence of all investigations is the examination and understanding of
causes and their effects. The effect of work is tiredness and that of too much
work is fatigue or stress. The effect of load on a structure [Fig. 1-1 1(b)] is to
cause deformations, strains, and stresses too much load causes fatigue and
;

failure.

When
effect

studying

finite

element methods, our main concern

is

the cause

and

of forcing functions (loads) on engineering systems.

The foregoing

offers a rather abstract description

process of discretization, inherent in almost

all

of ideas underlying the

human

endeavors. Compre-

hending these ideas significantly helps us to understand and extend the


element concept to engineering; that is the goal of this text.

finite

IMPORTANT COMMENT
Although we have presented the descriptions in this book by using simple
problems, we should keep in mind that the finite element method is powerful
and popular because it allows solution of complex problems in engineering
and mathematical physics. The complexities arise due to factors such as
irregular geometries, nonhomogeneities, nonlinear behavior, and arbitrary
loading conditions. Hence, after learning the method through simple examples, computations, and derivations, our ultimate goal will be to apply it to
complex and challenging problems for which conventional procedures are
not available or are very

difficult.

REVIEW ASSIGNMENTS
In the beginning stages of the study,
student

homework

it

may prove

that requires review of

very useful to assign the

some of the

basic laws, principles,

::

Chapter

Introduction

and equations. This

will facilitate

understanding the method and also reduce

the necessity of reviews by the teacher.

assignments that cover topics from

Home
1.

Assignment

15

The following

are

many undergraduate

two suggested home


curricula.

Define:
(a)

Stress at a point

(b) Strain
(c)

2.

Hooke's law

Define
(a) Principal stresses

and

(b) Invariants of stresses


3.

(a)

strains

and

strains

Define potential energy as a

sum

of strain energy and potential of applied

loads,
(b)

Give examples for analysis in

(civil)

engineering in which you have used

the concept of potential energy.


4.

Define
(a)

Darcy's law and coefficient of permeability

(b) Coefficient of thermal conductivity

and

coefficient of thermal

expansion

5.

Derive the Laplace equation for steady-state flow.

6.

Derive the fourth-order differential equation governing beam bending,

wA EI ir*)= p - k w

where

w =

displacement,

spring constant,

coordinate along

E=

axis.

= support reaction, k =
= moment of inertia, and x =

applied load, and k s w

modulus of

beam

>

elasticity,

See Fig. 1-12.

1
Figure 1-12

Home
1.

Assignment 2

Define:
(a)

(b)
(c)

Determinant
Row, column, and rectangular matrices
Matrix addition and subtraction

Chapter 1

Introduction

16

(d)

Matrix multiplication

(e)

Inverse of a matrix

Transpose of a matrix
(g) Symmetric matrix
(h) Sparsely populated and banded matrices
(f )

2.

(a)

Define a

set

of algebraic simultaneous equations, (b) Describe Gaussian

elimination with respect to the following equations,

2*i

4*!

and
3.

find the value of

x and x 2
x

+
+

3* 2

5* 2

=
=

14,
10,

Define:
(a)

Total derivative

(b) Partial derivative for

one variable and two variables and the chain rule of

differentiation

REFERENCES
[1]

[2]

The Upnishads Praise of Time. See available translations of The Upnishads


from Sanskrit to English, e.g., The Upnishads by Swami Nikhilananda, Harper
Torchbooks, New York, 1964.
Russell, Bertrand, Wisdom of the West, Crescent Books,

Books

Ltd.,

London, 1959.

Inc.,

Rathbone

THE FINITE
ELEMENT METHOD
STEPS IN

INTRODUCTION
Formulation and application of the

finite

element method are considered to

consist of eight basic steps. These steps are stated in this chapter in a very

The main aim of this general description is to prepare for


complete and detailed consideration of each of these steps in this and the
subsequent chapters. At this stage, the reader may find the very general
general sense.

description of the basic steps in this chapter

However, when these

somewhat overwhelming.

steps are followed in detail with simple illustrations in

the subsequent chapters, the ideas

and concepts

will

become

clear.

GENERAL IDEA
Engineers are interested in evaluating effects such as deformations, stresses,
temperature, fluid pressure, and fluid velocities caused by forces such as

The nature of disbody depends on the characteristics of the force system and of the body itself. Our aim is to find this
distribution of the effects. For convenience, we shall often use displacements
or deformations u (Fig. 2-1) in place of effects. Subsequently, when other
problems such as heat and fluid flow are discussed they will involve distribution of temperature and fluid heads and their gradients.
applied loads or pressures and thermal and fluid fluxes.

tribution of the effects (deformations) in a

17

18

Steps

in the Finite

Element Method

Chapter 2

Distribution of u(x, y)
for entire body

u e (x, y)

Nodal

line

Additional

node
Finite element

Corner or
primary node

(b)

(a)

Figure 2-1 Distribution of displacement

head

(p.

(a) Discretization

tion of Ue over a generic element

We assume that
tional

it is

u,

temperature T, or fluid

of two-dimensional body, (b) Distribue.

difficult to find the distribution

methods and decide to use the

finite

the concept of discretization, as explained in Chapter


into a

number of

of u by using conven-

element method, which


1

We

based on

body

smaller regions [Fig. 2- 1(a)] called finite elements [1,2].

consequence of such subdivision

is

that the distribution of displacement

also discretized into corresponding subzones [Fig. 2- 1(b)].

elements are

is

divide the

now

easier to

distribution of u over

examine as compared to the

A
is

The subdivided
entire body and

it.

For stress-deformation

analysis of a

body

in equilibrium

under external

loading, the examination of the elements involves derivation of the stiffness-

load relationship.

To

derive such a relationship,

we make

use of the laws and

principles governing the behavior of the body. Since our primary concern
to find the distribution of u,

terms of

u.

we

contrive to express the laws and

is

principles in

We do this by making an advance choice of the pattern,

shape, or

outline of the distribution of u over an element. In choosing a shape,

we

follow certain rules dictated by the laws and principles. For example, one law
says that the loaded body, to be reliable

breaks anywhere in
tinuous. Let us

now

its

and

functional, cannot experience

regime. In other words, the

body must remain con-

describe in detail various steps involved in the foregoing

qualitative statements.

Step

1.

Discretize and Select Element Configuration

This step involves subdividing the body into a suitable number of "small"
bodies, called finite elements.

The

intersections of the sides of the elements

and the

are called nodes or nodal points,

interfaces between the elements are


and nodal planes. Often we may need to introduce additional
node points along the nodal lines and planes [Fig. 2- 1(b)].
An immediate question that arises is, How small should the elements
chosen be? In other words, how many elements would approximate the
continuous medium as closely as possible? This depends on a number of
factors, which we shall discuss.
What type of element should be used ? This will depend on the characteristics of the continuum and the idealization that we may choose to use. For
instance, if a structure or a body is idealized as a one-dimensional line, the
element we use is a "line" element [Fig. 2-2(a)]. For two-dimensional bodies,
we use triangles and quadrilaterals [Fig. 2-2(b)]; for three-dimensional
idealization, a hexahedron with different specializations [Fig. 2-2(c)] can be

called nodal lines

used.

Although we could subdivide the body into regular-shaped elements in


we may have to make special provisions if the boundary
irregular. For many cases, the irregular boundary can be approximated by

the interior (Fig. 2-3),


is

a number of straight lines (Fig.


problems,

it

may be

2-3).

On

the other hand, for

many

other

necessary to use mathematical functions of sufficient

order to approximate the boundary. For example,

if

the boundary shape

is

approximate that

we can use a second-order quadratic function to


boundary. The concept of isoparametric elements that we

shall discuss later

makes use of

similar to a parabolic curve,

this idea. It

may

be noted that inclusion of

irregular boundaries in a finite element formulation poses

Step

2.

Select Approximation

no great

difficulty.

Models or Functions

we choose

a pattern or shape for the distribution (Fig. 2-1)


quantity that can be a displacement and/or stress for stressdeformation problems, temperature in heat flow problems, fluid pressures

In this step,

of the

unknown

and/or velocity for fluid flow problems, and both temperature (fluid pressure)
and displacement for coupled problems involving effects of both flow and
deformation.

The nodal points of the element provide strategic points for writing
mathematical functions to describe the shape of the distribution of the
unknown quantity over the domain of the element. A number of mathematical functions such as polynomials and trigonometric series can be used for
this purpose, especially

polynomials because of the ease and simplification


19

20

Steps

in the Finite

Element Method

Chapter 2

'

'

One-dimensional body

Line elements

(a)

unknowns
each node

at

Quadrilateral

Triangle

Two-dimensional body

Quadrilateral and
triangular elements

(b)

Hexahedron element

Three-dimensional body

(0
Figure 2-2 Different types of elements, (a) One-dimensional ele-

ments,

(b)

elements.

Two-dimensional

elements,

(c)

Three-dimensional

Chapter 2

Steps

in the Finite

Element Method

21

body

Original

Discretized

body

Figure 2-3 Discretization for irregular boundary.

they provide in the

finite

element formulation. If we denote u as the unknown,

the polynomial interpolation function can be expressed as

= N lUl + N u + N

Here u u u 2l u 3 ,
and u 2

N N

we

u3

+ Nm u m

,um are the values of the unknowns

Nm are the

shall give details

(2-1)

at the nodal points

interpolation functions; in subsequent chapters

of these functions. For example, in the case of the

we can have u and u 2 as


unknowns or degrees of freedom and for the triangle [Fig. 2-2(b)] we can
have u u u2
u 6 as unknowns or degrees offreedom if we are dealing with
line

element with two end nodes [Fig. 2-2(a)]

, . . . ,

a plane deformation problem where there are two displacements at each


node.

A degree of freedom can be defined as an independent (unknown) displacement that can occur at a point. For instance, for the problem of onedimensional deformation in a column [Fig. 2-2(a)], there is only one way in
which a point is free to move, that is, in the uniaxial direction. Then a point
has one degree of freedom. For a two-dimensional problem [Fig. 2-2(b)], if
deformations can occur only in the plane of the body (and bending effects are
ignored), a point is free to move only in two independent coordinate directions; thus a point has two degrees of freedom. In Chapter 6, when bending
considered, it will be necessary to consider rotations or slopes as independent degrees of freedom.
We note here that after all the steps of the finite element method are

is

accomplished,

we

shall find the solution as the values of the

unknowns u

at

u m To initiate action toward obtaining the


u l9 u2>
solution, however, we have assumed a priori or in advance a shape or pattern
that we hope will satisfy the conditions, laws, and principles of the problem

all

the nodes, that

is,

at hand.

The reader should realize that the solution obtained


unknowns only at the nodal points. This is one of
discretization process. Figure 2-4

shows that the

will

be

in

terms of the

the outcomes of the

final solution is

a combina-

22

Steps

in the Finite

Element Method

Chapter 2

u(x, y)

u 3 (x, y)

Common

boundary

(b)

Finite element
approximation

Exact

Common

boundary

(0
Figure 2-4 Approximate solution as patchwork of solutions over
elements, (a) Assemblage, (b) Neighboring elements, (c) Section

along A-A.

tion of solutions in each element patched together at the

This

is

common boundaries.

further illustrated by sketching a cross section along A-A. It can be

seen that the computed solution

is

not necessarily the same as the exact

continuous solution shown by the solid curve. The statement in Chapter


that discretization yields approximate solutions can be visualized

schematic representation (Fig.


solution; that

Step

3.

is,

2-4).

Obviously,

computed solution
the error is a minimum.

tion to be such that the

is

we would

from

this

like the discretiza-

as close as possible to the exact

Define Strain (Gradient)-Displacement (Unknown)

and Stress-Strain (Constitutive) Relationships

To proceed to the next step, which uses a principle (say, the principle of
minimum potential energy) for deriving equations for the element, we must
define appropriate quantities that appear in the principle.

For

stress-defor-

Chapter 2

Steps

Element Method

in the Finite

23

mation problems one such quantity is the strain (or gradient) of displacement. For instance, in the case of deformation occurring only in one
direction

assumed to be

[Fig. 2-5(a)], the strain e ,


y

small,

is

given by

dv
y

where v

is

the deformation in the

direction, such a relation

is

(2-2)

dy

For the case of fluid flow

direction.

the gradient g x of fluid head [Fig. 2-5(b)]


gx

in

one

(2-3)

dx

Here cp is the fluid head or potential and g x is the gradient of <p, that
change of q> with respect to the x coordinate.

Variation of

is,

rate of

g x gradient or slope

v?

of

<

I
e

gradient

or slope of v
x,

v?

(b)

Variation

of v

Figure 2-5 Problems idealized as one-dimensional, (a) One-dimensional stress-deformation, (b) One-dimensional flow.

In addition to the strain or gradient,

we must

quantity, the stress or velocity; usually, this

ship with the strain. Such a relation


alized sense,

it

is

is

is

also define an additional


done by expressing its relation-

called a stress-strain law. In a gener-

a constitutive law and describes the response or effect

(displacement, strain) in a system due to applied cause (force).


strain

law

is

one of the most

vital parts

The

stress-

of finite element analysis. Unless

it is

defined to reflect precisely the behavior of the material or the system, the
results

from the analysis can be of very little significance. As an elementary


Hooke's law, which defines the relationship of stress to
in a solid body:

illustration, consider

strain

ay
where a y

elasticity. If

= Eye

(2-4a)

y,

stress in the vertical direction

we

substitute e y

from Eq.

and

Ey =

Young's modulus of
we have the

(2-2) into Eq. (2-4a),

expression for stress in terms of displacements as

~ dv
dy

(2-4b)

24

Steps

One of

in the Finite

Element Method

the other simple linear constitutive laws

is

Chapter 2

Darcy's law for fluid

flow through porous media


vx

= coefficient of permeability,

where k x

electrical engineering the

Step

= -k x gx

4.

(2-4c)

vx

= velocity,

corresponding law

is

Ohm's

and gx

= gradient.

In

law.

Derive Element Equations

By invoking

and principles, we obtain equations governing


The equations here are obtained in general terms

available laws

the behavior of the element.

and hence can be used for

all

elements in the discretized body.

number of alternatives are possible for the derivation of element


equations. The two most commonly used are the energy methods and the
residual methods.

Use of the energy procedures

At

this stage

of our study of the

requires knowledge of variational calculus.


finite

element method, we shall postpone

detailed consideration of variational calculus

manner introduce

and

in a

somewhat

less

rigorous

the ideas simply through the use of differential calculus.

ENERGY METHODS
These procedures are based on the idea of finding consistent

states

of

bodies or structures associated with stationary values of a scalar quantity

assumed by the loaded bodies. In engineering, usually this quantity is a


measure of energy or work. The process of finding stationary values of
energy requires use of the mathematical disciplines called calculus of variations
involving use of variational principles. In this introductory book, it is not
considered necessary to elaborate on this subject.

As noted above, we

shall

introduce and use the energy methods through the familiar topic of differential calculus.

Within the realm of energy methods, there are a number of methods and
variational principles, e.g., the principle of stationary potential

and com-

plementary energies, Reissner's mixed principle, and hybrid formulations,

which are commonly used

in finite element applications [3-6].

STATIONARY VALUE

maximum, minimum, or
Under certain conditions, the

In simple words, the term stationary can imply a


saddle point of a function F(x) (Fig. 2-6).

function

may

minimum or a maximum value. To


we equate the derivative of F to zero

simply assume a

point of a stationary value,

find the

Chapter 2

Steps

in the Finite

Element Method

25

F(x)

Maximum
Neutral

Minimum

Figure 2-6 Stationary values of a function.

POTENTIAL ENERGY
In the case of stress-deformation analysis, the function

F is

often repre-

sented by one of the energy functions stated previously. For instance,


define

Fto

simple column under the given support conditions [Fig. 2-5(a)],


elastic

and

minimum

we can

be the potential energy in a body under load. If the body, say, a


if it is in

equilibrium,

it

To comply

potential energy.

is

can be shown that the column


with the

np

denote potential energy by the symbol

and
assume

linear

will

commonly used notation we

where the subscript denotes

potential energy.

The

potential energy

is

defined as the

sum of the

internal strain energy

and the potential of the external loads


the latter term denotes the
p
capacity of load P to perform work through a deformation v of the column.
\

Therefore,
II P

When we

= U+

apply the principle of

take the derivative (or variation) of

(2-6a)

minimum

potential energy,

U p and equate

it

to zero.

the load remains constant while taking the derivative

m
The symbol 3 denotes

= 8u

sw

we can

essentially

then

= o.

(2-6b)

variation of the potential energy n^.

subsequently in Eq. (2-9),

we

We assume that

interpret the variation or

As

indicated

change as com-

posed of a series of partial differentiation of 11^. Here we use the relation


between the variation in potential of external loads and in work done by the
loads as

dW
Note

5W

(2-6c)

that the negative sign in Eqs. (2-6b)

potential of the external loads in Eq. (2-6a)

the external loads.

is

and

(2-6c) arises because the

lost

through the work done by

Steps

26

The

Element Method

Chapter 2

fact that for linear, elastic bodies in equilibrium the value of

minimum can be
of n^,

in the Finite

verified

greater than zero

is

S2

Up

is

by showing that the second derivative or variation


that

is,

U P =3 U-S WP
2

>0.

(2-7)

Proof of Eq. (2-7) can be found in advance treatments on energy methods


and is not included in this text. The symbol S is a compact symbol used to
denote variation or a series of partial differentiations. For our purpose, we
shall interpret it simply as a symbol that denotes derivatives of 11^ with
respect to the independent coordinates or
expressed.

For example,

np
where u ,u 2
1

3T1 P

unknowns

in terms of

which

it is

if

=np(u

w are the total

19

u2 ,...,uj,

(2-8)

number of unknowns

then

(at the nodes),

implies

*L=o,
L=o,
du
2

(2-9)

Here n

total

number of unknowns.
we shall

In subsequent chapters
stationary energy

and other energy

illustrate the

use of the principle of

principles for finite element formulations

of various problems.

METHOD OF WEIGHTED RESIDUALS


One of the two major alternatives
method

is

are employed under the


least squares,

for formulating the finite element

(MWR). A number of schemes


among which are collocation, subdomain,
methods [3, 7]. For many problems with certain

method of weighted

the

residuals

MWR,

and Galerkin's

later chapters), Galerkin's method


from variational procedures and is closely
related to them. Galerkin's method has been the most commonly used
residual method for finite element applications.
The
is based on minimization of the residual left after an approximate or trial solution is substituted into the differential equations governing a

mathematical characteristics (discussed in


yields results identical to those

MWR

problem. As a simple

illustration, let us consider the following differential

equation

-='<*>

<2

10a >

Chapter 2

where w*

Steps in the Finite Element

unknown, x

the

is

is

Method

the coordinate,

27

the time,

t is

forcing function. In mathematical notation, Eq. (2-10a)

is

and f(x)

is

the

written as

Lu* =/,

(2-10b)

where

- dx
is

dt

the differential operator.

We

are seeking an approximate solution to Eq. (2-10) and denote an

approximate or

function u for u* as

trial

ft

cp

(2-11)

a^j

<x 2 <p 2

a n <p H

Here p t q> 2
<p are known functions chosen in such a way as to satisfy
the homogeneous boundary conditions; cp Q is chosen to satisfy the essential,
geometric, or forced boundary conditions; and a, are parameters or constants to be determined. Categories of boundary conditions are explained
subsequently and discussed in Chapter 3. Often, for convenience, u in
,

Eq. (2-11)

written as

is

u
in

which a

and

into Eq. (2-10),

we

<p

are

left

(p Q

=*
If the

(2-12)

t <Pt

approximate solution u

is

R(x)=Lu-f

(2-13)

which

is

zero

In the

if

u*.

method of weighted

residuals, the

aim

to find an approximate

is

solution u for w* such that the residual R{x) in Eq. (2-13)


possible or

is

substituted

with a residual

is

made

as small as

minimized. In other words the error between the approximate

and the exact solution u is minimized. A number of schemes


aim of minimization of R(x) details of some of
the major schemes collocation, subdomain, least squares, and Galerkin
methods are given in Appendix 1. Moreover, in the subsequent chapters,
Galerkin's method is used to derive finite element equations for a number of
simple problems. For the sake of completeness, only brief and general statements for these methods are given below.
solution w*

are possible to achieve the

Mathematically the idea of minimization can be expressed as

R(x)W (x)dx
t

= 0,

1, 2,

n,

(2-14)

where D denotes the domain of a structure or body under consideration.


For the one-dimensional column problem, the domain is simply the linear
extent of the column.

28

Steps in the Finite Element

Method

Chapter 2

In Eq. (2-14) the


denote weighting functions. Various residual schemes
such as collocation, subdomain, and Galerkin use different weighting
t

For

functions.

instance, in the case of the collocation

method,

W=

1.

The

expression in Eq. (2-14) implies that the weighted value of R(x) over the

domain of a

structure vanishes. Figure 2-7 shows a schematic representation


of Eq. (2-14). The shaded areas in Fig. 2-7(b) denote error between the

approximate and the exact solution u*

R(x) over

u over the domain D.

[Fig. 2-7(a)] is related to the error

integral of R(x) over

u*

w,

The residual
and the sum or the

minimized.

is

R(x)= Lu-f

(a)

Figure 2-7 Schematic representation of residual, (a) Integration of

over D. (b) Error u*

As a simple

u over D.

illustration, let us consider the following

second-order

differ-

ential equation that governs the problems of one-dimensional stress deforma-

column and flow

tion in a

in

Chapters 3 and

d 2v
dy

where v*

the

is

unknown

4, respectively

(2-1 5a)

f,

(deformation), y

is

the coordinate axis, c denotes a

material property, and /is the forcing function. For a column, c


the applied load,
area.

Assume

E is

that

the

= EA, /is

modulus of elasticity, and A is the cross-sectional


10 units of load; then Eq. (2-1 5a)
1
and

/=

EA =

specializes to

d 2 v*
dy

An

(2-15b)

10.

approximate solution for v* can be written as a special case of

Eq. (2-11) as
v

where

<p

satisfy the

= ai +

cc 2

(x 3

= 2 *t9t,

(2

"

16)

2
This function should be chosen so as to
1, <p 2 = y, and <p 3 = y
boundary conditions of the problem. Then the residual R(y) is
.

Chapter 2

Steps in the Finite Element

Method

29

given by

R(y)

According to the

MWR,
f*

W~
1

I0

(2 " 17a)

'

Eq. (2-14) leads to

Riy)W {y)dy
)

= 0,

(2-17b)

1, 2, 3,

Jo

or

^(^-10)w,(y)dy =
or

j^d^y w

(y)dy

0,

(2-17C)

(2 . 17d)

Here L

is the length of the column which represents the domain D. Now, we


can substitute for the second derivative in Eq. (2-17) by differentiating v in
Eq. (2-16) twice. The final results will yield three simultaneous equations in

au

cc 2

and a 3

as

[[^^-^]w

[[^^y

(y)dy

[ -^-10]w
d

which are solved for % u

<x 2

substituted into Eq. (2-16),

At

this stage,

to proceed
in the

it is

from Eq.

and a 3

we

y)dy

i(

0,

= 0,

(y)dy

When

(2-18)

0,

these values of

a a 2 and a 3
,

are

obtain the approximate solution for v*.

not necessary to go into the details of the steps required


(2-17) to Eq. (2-18); they are given in

Appendix

and

subsequent chapters.

ELEMENT EQUATIONS
Use of either of the two foregoing methods will lead to equations
ing the behavior of an element, which are commonly expressed as
[k]{q]

= {Q},

describ-

(2-19)

vector of unknowns at the


[k]
element property matrix, {q}
element nodes, and {Q}
vector of element nodal forcing parameters.
Equation (2-19) is expressed in a general sense; for the specific problem of
vector of nodal displacements,
stress analysis, [k]
stiffness matrix, {q}

where

and {Q}

=
=
= vector of nodal forces. Details of the matrices in Eq. (2-19) will be

developed and described fully in subsequent chapters.

Step

Assemble Element Equations to Obtain Global

5.

or Assemblage Equations and Introduce Boundary

Conditions

Our

final

aim

to obtain equations for the entire

is

body

approximately the behavior of the entire body or structure. In

that define

fact, as will

discussed in various chapters, use of the variational or residual procedure

body

relevant to the entire

it is

for simplicity that

we view

be
is

the procedure in

Step 4 as having been applied to a single element.

Once
element,

the element equations, Eq. (2-19), are established for a generic

we

are ready to generate equations recursively for other elements by

using Eq. (2-19) again and again. Then


equations. This assembling process
continuity (Chapter

1). It

is

we add them

together to find global

based on the law of compatibility or

requires that the

body remain continuous; that

is,

the neighboring points should remain in the neighborhood of each other after

the load

is

applied (Fig. 2-4). In other words, the displacements of two

adjacent or consecutive points must have identical values [Fig. 2-8(a)].

Depending on the type and nature of the problem, we may need to enforce
the continuity conditions more severely. For instance, for deformations
occurring in a plane,

it

may be

sufficient to enforce continuity

of the dis-

Figure 2-8 Interelement compatibility, (a) Compatibility for plane

problems, (b) Compatibility for bending-type problems.


Slopes

may not

be

equal

Elements

Equal displacements
(a)

Slope

Equal slopes or
gradients

Equal displacements
(b)

30

Chapter 2

Steps

On

placements only.

in the Finite

Element Method

31

the other hand, for bending problems, the physical

properties of the deformed

body under the load

the continuity of displacements

we

requires that in addition to

ensure that the slopes or the

first

deriva-

of displacements are also continuous or compatible at adjacent nodes


[Fig. 2-8(b)]. Often it may become necessary to satisfy compatibility of the
curvatures or the second derivative also.

tive

Finally,

we

obtain the assemblage equations, which are expressed in

matrix notation as
[Ktfr}

= {R},

(2-20)

where [K] = assemblage property matrix, {r} = assemblage vector of nodal


unknowns, and {Q} = assemblage vector of nodal forcing parameters.
For stress-deformation problems, these quantities are the assemblage stiffness matrix, nodal displacement vector, and nodal load vector, respectively.

BOUNDARY CONDITIONS
Until now we have considered
Equation (2-20)

[Fig. 2-9(a)].

only the properties of a body or structure

tells

us about the capabilities of the body to

is an engineer. How he
depend on the surroundings and the
problems he faces; these aspects can be called constraints. In the case of
engineering bodies, the surroundings or the constraints are the boundary
conditions. Only when we introduce these conditions can we decide how the
body will perform.
Boundary conditions, then, are the physical constraints or supports
[Fig. 2-9(b)] that must exist so that the structure or body can stand in space
uniquely. These conditions are commonly specified in terms of known values
of the unknowns on a part of the surface or boundary S and/or gradients or
derivatives of the unknowns on S 2 Figure 2- 10(a) depicts a beam. In the case
of the simply supported beam, the boundary S is the two end points where

withstand applied forces.


will

perform

It is just like

saying that one

his engineering duties will

the displacements are given. This type of constraint expressed in terms of

displacements

is

often called the essential, forced, or geometric boundary

conditions.

At

the simple supports of the beam, the bending

moment

zero

that

is,

the second derivative of displacement vanishes. This type of constraint

is

is

often called a natural boundary condition.

Figure 2- 10(b) shows a cylinder through which fluid or temperature


On the boundary S^ temperature or fluid head is known; this is the
essential boundary condition. The right end is impervious to water, or

flows.

insulated against heat; then the

heat flux, which

is

boundary condition

proportional to the

first

is

specified as fluid or

derivative of fluid

head or tem-

perature. This is the natural boundary condition.


To reflect the boundary conditions in the finite element approximation of
the body represented by Eq. (2-20), it is usually necessary to modify these

[K]{r}= {R}

(a)

[K](7}=

{R]

Constraints

(b)

Figure 2-9 Boundary conditions or constraints, (a)


constraints (b)

Body with

Figure 2-10 Examples

boundary conditions,

Body without

constraints.

of boundary conditions,
(b) Pipe flow with

(a)

s,

"TT
(a)

\
Flow
(b)

Beam

with

boundary conditions.

S,

Chapter 2

Steps

in the Finite

Element Method

33

equations only for the geometric boundary conditions. Further details and
procedure for such modification are given in subsequent chapters. The final

modified assemblage equations are expressed by inserting overbars as


[K]{f}

Step

6.

Solve for the Primary

Equation (2-21)

is

set

= {R}.

(2-21)

Unknowns

of linear (or nonlinear) simultaneous algebraic

equations, which can be written in standard familiar form as

K
K

ll

rl

2l r t

+ K 12 r +
+ K22 r +

+K
+K

ln r n

2n rn

= Ru
= Ri>
(2-22)

K s + Km2 r +
n

+K

nn rn

=R

These equations can be solved by using the well-known Gaussian elimination


or iterative methods. Detailed coverage of these methods is beyond the scope
of this book; however, later when individual problems are considered, some
of these methods will be illustrated. Moreover, brief descriptions of these
in Appendix 2. At the end of this step, we have
unknowns (displacements) r l9 r2 ,
rn These are called
primary unknowns because they appear as the first quantities sought in the
basic Eq. (2-22). The designation of the word primary will change depending
on the unknown quantity that appears in Eq. (2-22). For instance, if the

methods are given

solution

solved for the

problem

is

formulated by using stresses as unknowns, the stresses

called the primary quantities.

be the

fluid or velocity

Step

7.

will

be

For the flow problem the primary quantity can

head or potential.

Solve for Derived or Secondary Quantities

Very often additional or secondary quantities must be computed from the


primary quantities. In the case of stress-deformation problems such quantities
can be

strains, stresses,

moments, and shear forces for the flow problem they


;

can be velocities and discharges. It is relatively straightforward to find the


secondary quantities once the primary quantities are known, since we can
make use of the relations between the strain and displacement and stress

and

strain that are defined in Step 3.

Step

8.

Interpretation of Results

final and the important aim is to reduce the results from the use of the
element procedure to a form that can be readily used for analysis and
design. The results are usually obtained in the form of printed output from

The

finite

the computer.

We then

select critical sections

of the body and plot the values

Steps

34

in the Finite

Element Method

of displacement and stresses along them, or


often very convenient

and

less

we can

Chapter 2

tabulate the results. It

is

time consuming to use (available) routines and

ask the computer to plot or tabulate the results.

REFERENCES
[1]

Turner, M.

J.,

Clough, R. W., Martin, H. C, and Topp, L. C, "Stiffness and


Complex Structures," /. Aero. Sci., Vol. 23, No. 9,

Deflection Analysis of
Sept. 1956.
[2]

Clough, R. W., "The Finite Element Method in Plane Stress Analysis,"


Proc. 2nd Conf. on Electronic Computation, ASCE, Pittsburgh, Sept. 1960.

[3]

Crandall,

[4]

Argyris,

S.

J.

H., Engineering Analysis, McGraw-Hill,

New

in

York, 1956.

H., Energy Theorems and Structural Analysis, Butterworth's,

London, 1960.
[5]

[6]

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Element Method,

Vah

Pian, T. H.*H and Tong, P., "Finite Element Methods in Continuum


.

Mechanic^"
York, 1972.
[7]

to the Finite

Finlayson,

Academic

in

Advances

Press,

in

Applied Mechanics, Vol.

12,

Academic

Press,

New

The Method of Weighted Residuals and Variational Principles,


York, 1972.

New

ONE-DIMENSIONAL
STRESS DEFORMATION

203SQ01
INTRODUCTION
From

we shall consider engineering problems idealThe main motive for treating these simple problems

here through Chapter 10

ized as one-dimensional.

to introduce the reader to the details of various steps so that basic concepts
can be understood and assimilated thoroughly without undue complex and
is

lengthy derivations. An advantage to this approach is that hand calculations


can be performed for two- and three-dimensional problems this can become
;

increasingly difficult.

Although simple problems are treated (Chapters 3-10), we introduce


concepts and terms that are general and relevant to advanced theory
and applications. These concepts are explained and defined in easy terms,
often with intuitive and physical explanations. It may be mentioned that

many

one-dimensional idealizations permit not only simple derivations but often


provide satisfactory solutions for

As our

first

many

practical problems.

problem we consider the case of a column,

strut,

or bar of

uniform cross section subjected to purely axial loading [Fig. 3- 1(a)]. Under
these conditions we can assume that the deformations will occur only in one,
vertical direction. Consequently, we can further assume that the column can
be replaced by a
[Fig. 3- 1(b)].

line

Now we

with the axial

stiffness

EA lumped

consider derivations of the

the step-by-step procedure as described in Chapter

finite

at the centerline

element method in

2.

35

One- Dimensional Stress Deformation

36

Chapter 3

ic

s
Nodes

Finite

elements

S EA
zMx^

777Z

(0

(b)

(a)

Figure 3-1 Axially loaded column, (a) Actual column, (b) One-

dimensional idealization,

Step

1.

(c)

Discretization.

Discretization and Choice of Element Configuration

Before we proceed further it is necessary to describe the coordinates or


geometry of the column by using a convenient coordinate system. In the onedimensional approximation, it is necessary to use only one coordinate along
the vertical direction.
is

We

y axis. Because this coordinate system


column (or structure), it can be called the

call this the

defined to describe the entire

global coordinate system.

We now

discretize the

column

into

an arbitrary number of smaller units

that are called finite elements [Fig. 3-l(c)].

The

intersections of elements are

called nodes or nodal points.

At

this stage, before Step 2,

it is

useful to introduce the concept of a local

or element coordinate system. There are a

number of advantages

local system for deriving element equations (Steps 2-4)

for multidimensional problems,

makes the required

tions extremely simple to handle. Indeed,

it is

its

derivatives

and integra-

possible to obtain

by using the global system, but use of the local system


subsequently) facilitates the derivatives and is economical.

tions

to using a

use, particularly

(as

all

we

deriva-

shall see

EXPLANATION OF GLOBAL AND LOCAL COORDINATES


For a simple explanation, let us consider an example, Fig. 3-2(a); here we
need to define or survey a plot of land and relate it to a standard or global
point or benchmark A. Surveying the plot could entail locating each point
and establishing its distance from point A. Let us assume that point A
away from the plot and that it is difficult to establish a direct relation
with A. A local point B is available, and its distance from A is known, howin
is

it

far

Chapter 3

One- Dimensional Stress Deformation

37

Plot of land

"^S B

Local station

"Global" station

(a)

r
>
1

>s =

.L-+1-

A(Dfo

-1
s

Y2

y2
Y3
Vi

Vi
I

Global y

v/^W
s

Global y

= Local coordinate = y//

^^

77777,

L =

(y-y 3 )/(//2)

(b)

(c)

Figure 3-2 Global and local coordinates, (a) Concept of global and
local coordinate systems, (b) Local coordinate

point

ever. Point

1. (c)

B is

measured from node

Local coordinate measured from midnode

accessible

from

all

points in the plot.

distances of each point in the plot

3.

We

can

first

define the

from the point B; then, knowing the

distance x AB from point A to point B, it is possible to define the distances of


the points in the plot with respect to point A. For example, if the distance

of any point P from

B is x B

then

*a

its

distance from point

= x* +

*ab-

A is
(3-la)

Here we can call the measurements with respect to point B local coordiand those from point A, global coordinates. We can see that the mea-

nates,

One- Dimensional Stress Deformation

38

surements with respect to point B, which


of the plot

much

is

in the vicinity,

Chapter 3

make

the definition

simpler than those with respect to point A, which

is

more

There can be a number of possible points like B with


respect to which we can define the local coordinate system such choices will
depend on the nature of the problem and the convenience and ease of
difficult to reach.

measurements.

The basic idea of the use of local coordinate system(s)


ment method is very similar to the foregoing concept.

in the finite ele-

LOCAL AND GLOBAL COORDINATE SYSTEM


FOR THE ONE-DIMENSIONAL PROBLEM
As mentioned above,

there are a

number of ways

in

which we can define local

coordinate systems for the one-dimensional problem (Fig.


here two local coordinate systems, Figs. 3-2(b) and
In the

first

case,

we measure

3-1).

We

consider

(c).

from the node point 1


is measured from
analogous to point B and that

the local coordinate

of a generic element e [Fig. 3-2(b)]. The global coordinate


the base of the column.
the base

is

Note

that

analogous to point

node

is

in Fig. 3-2(a).

y then the global coordinate of any point


;

y=y+y

We

call the local

in the element

coordinate

is

(3-lb)

hence

=y- yi-

y
Often

it is

0- lc)

more convenient to express the local coordinate as a nondi-

mensional number; such a procedure can considerably facilitate the integrations and differentiations involved in the subsequent computation. Here we
nondimensionalize by dividing^ by the length of the element; thus
o

y -y\ _
yi-yx

(3-ld)

where s = nondimensionalized local coordinate, / = length of the element,


and Vi and y 2 = global coordinates of nodes 1 and 2, respectively. Note that
because we nondimensionalized the coordinate by dividing by the length of
the element, I = y 2 J>i, the value of s varies from zero at node 1 to unity
at node 2.
In the second alternative, we can attach the origin of the local system at
an intermediate point in the element, say, at the midpoint [Fig. 3-2(c)]. Here
the local coordinate

is

written as

L
The values of L range from

^-

at point

to

(3-2)

at point 3 to

at point 2.

One- Dimensional Stress Deformation

Chapter 3

An

important property of these local coordinates

39

is

that they are in

nondimensionalized form, and their values are expressed as numbers and often
lie

between zero and unity.

property that imparts simplicity to the

It is this

subsequent derivations.
Step

2.

Select Approximation
for the

Unknown

One of the main


grasp at this stage

Model

or Function

(Displacement)

ideas in the finite element


is

method

that the reader should

the a priori or in advance selection of mathematical

functions to represent the deformed shape of an element under loading. This


implies that since

it is

a closed form or exact solution,

difficult to find

we

guess a solution shape or distribution of displacement by using an appropriate


this function, we must follow the laws,
and constraints or boundary conditions inherent in the problem.
The most common functions used are polynomials. In the initial stages
of the finite element method, the polynomials used were expressed in terms
of generalized coordinates however, now most finite element work is done
by using interpolation functions, which can often be considered as transformed generalized coordinate functions.

mathematical function. In choosing


principles,

Generalized Coordinates

The

we can use

simplest polynomial that

is

the one that gives linear

variation of displacements within the element [Fig. 3-3(b)],

=a +

oc 2

(3-3a)

y,

or in matrix notation.
l

[1

y]\

<;)

(3-3b)

or

W=
where o^ and a 2
in the element,

y and y2 To show
x

- 3c

= generalized coordinates, y = the coordinate of any point


and v = displacement at any point in the element. The a's

contain the displacements at nodes v


.

(3

[+]{}.

this,

we

first

and v 2 and the coordinates of nodes


1 and 2 by substituting

evaluate v for points

fory,
Vl

=,

+a y

v2

<%!

l9

a 2j 2

(3-4a)

(3-4b)

or in matrix notation,

N=P "IN

(3-40

= [A]{}.

(3-4d)

or
{q}

One- Dimensional Stress Deformation

40

y3

Yi

Y2

'1

Chapter 3

(a)

V =

OLy

OL

(b)

N, =/ 2 //=1-sorN 1 = 1(1

L)

Equal

A A
(e)

Figure 3-3 Linear interpolation functions and interelement compatibility.

Here

{q}

[v x

v2]

is

the vector of nodal displacements. Second,

we

solve

for {a} as

{a}

= [A]-i{q},

where

(3-5a)

[A]-

-yi~

yi

_-i

l/l

~
1

yi

(3-5b)

yi

-y\ _-i

-yC
l

(3-5c)

Chapter 3

Here

One- Dimensional Stress Deformation

= Jacobian

|/|

element.

=y y =

determinant

41

equals the length of the

Thus

\[y

fai]

(2

-y

(3-5d)
1

(2

1)

2)

(2

v2 \
x 1)

Therefore

(3-5e)

which shows that

a's are functions of

and made up of y u y 2i v l9 and v 2 Note


.

that a's are related to but are not explicit functions of nodal displacements.

This

is

one of the reasons we call them generalized coordinates.


can substitute {a} into Eq. (3-3) to express v in terms of the

Now we

nodal displacements:
v

y 2v

-y,v 2

-+*>
(

(3-6a)

,,

-(Z ^)*.-(ZL f 2 )^
=

fv,

tf,*,

+ fv =
l

(l

y.) Vl

(3

j-v 2

+Nv
= s, N = y/l = s,

6b >

(3-6c)

(3-6d)

2i

"

where iVj = 1 y/l


1
and y === /, [Figs. 3-2(b) and
2
3-3]. Equation (3-6d) leads us to the concept of interpolation function models.
In this equation, N and N2 are called interpolation, shape, or fozs/s functions.
The displacement v at any point in the element can now be expressed as
x

v=[N,

N ]H

(3-7a)

= [N]{q},
where [N]

is

(3-7b)

called the matrix of interpolation, shape, or basis functions.

property of the interpolation functions

example, in the above,

N +N =
t

is

that their

sum

equals unity. For

INTERPOLATION FUNCTIONS
Since our aim in the finite element analysis

is

to find nodal displacements

v x and v 2 we can see the advantage of using approximation models of the


type in Eq. (3-7). In contrast to Eq. (3-3a), here the displacement v is
expressed directly in terms of nodal displacements. Also, the use of interpola,

42

One- Dimensional Stress Deformation

tion functions

makes

it

Chapter 3

quite easy to perform the differentiations

and

integra-

tions required in the finite element formulations.

In simple words, an interpolation function

of unity at the node point to which

it

is

pertains

a function that bears a value

and a value of zero

nodes. For example, Figs. 3-3(c) and (d) show distributions of

along the element; the function


point

JV,

pertains to point

1,

and

at other

and

pertains to

2.

For a given element it is possible to devise and use different types of local
coordinates and interpolation functions N
Let us consider the second
coordinate system [Fig. 3-2(c)] described in Step 2 and in Eq. (3-2), and
t

express the v as

v=i(l-L) Vl +i(l+L)v

= N v, + N v
2

(3-8a)

(3-8b)

= [tfi N ]l Vl \

(3-8c)

(3-8d)

Note

that

(i

1,

[N]{q].

2) in

Eq. (3-8b) are different from

(i

1,

2) in Eq.

same linear variation within the element.


Alternative coordinate systems based on local measurement from other points
on the element are possible; for example, see Prob. 3-1.
(3-6d); however, they yield the

RELATION BETWEEN LOCAL AND GLOBAL COORDINATES

An

important point to consider is that a one-to-one correspondence exists


between the local coordinate s or L and the global coordinate y of a point in
the element. For example, for L in Eq. (3-2) we have

= i(l - L)y + ^(1 + L)y


= #iJ>i + N y
= [N]{y],
t

where

T
{y]

[y

y2 ]

is

(3-9a)

(3-9b)

(3-9c)

the vector of nodal coordinates.

In fact an explanation of the concept of isoparametric elements, which

is

can be given at this stage. A comparison of Eq. (3-8) and (3-9) shows that both the displacement v and the
coordinate y at a point in the element are expressed by using the same (iso)
the most

common

procedure

interpolation functions.

An

now

in use,

element formulation where

similar) functions for describing the deformations in

(or geometry) of an element

This

is

is

we use the same (or


and the coordinates

called the isoparametric element concept

rather an elementary example;

we

and more general isoparametric elements.

shall subsequently

[1].

look at other

VARIATION OF ELEMENT PROPERTIES


We often tacitly assume that the material properties such as cross-sectional area A and the elastic modulus E are constant within the element. It
is not necessary to assume that they are constant. We can introduce required
variation, linear or higher order, for these quantities.

For instance, they can

be expressed as linear functions:

E = N,E + N E = [NtfEJ,
A = N A + N A = [N]{A},
E and {A} r = [A A are the
x

where [E n }
of E and A

=
at

[E

(3-10)

(3-11)

2]

nodes

2]

and

vectors of nodal values

2, respectively.

REQUIREMENTS FOR APPROXIMATION FUNCTIONS


As we have

stated before, the choice of an approximation function is guided


by laws and principles governing a given problem. Thus an approximation
function should satisfy certain requirements in order to be acceptable. For
general use these requirements are expressed in mathematical language.
However, in this introductory treatment, we shall discuss them in rather

simple words.

An

approximation function should be continuous within an element. The

linear function for v [Eqs. (3-7)

and (3-8)]

is

indeed continuous. In other words,

it

does not yield a discontinuous value of v but rather a smooth variation of

v,

and the variation does not involve openings, overlaps, or jumps.

up

The approximation function should provide interelement compatibility


by the problem. For instance, for the column prob-

to a degree required

lem involving

axial deformations,

it

is

necessary to ensure interelement

compatibility at least for displacements of adjacent nodes. That

is,

the

approximation function should be such that the nodal displacements between


adjacent nodes are the same. This is shown in Fig. 3-3(e). Note that for this
case, the higher derivatives such as the first derivatives may not be compatible.

The displacement

at node 2 of element 1 should be equal to the displacement


node 1 of element 2. For the case of the one-dimensional element, the linear
approximation function satisfies this condition automatically.
As indicated in Chapter 2 (Fig. 2-8), satisfaction of displacement compatibility by the linear function does not necessarily fulfill compatibility of first
derivative of displacement, that is, slope. For axial deformations, however,
if we provide for the compatibility up to only the displacement, we can still
expect to obtain reliable and convergent solutions. Often, this condition is
at

tied in with the highest order of derivative in the energy function such as the

potential energy.
derivative dv/dy

For example,

= ,

is

in

Eq. (3-21) below, the highest order of

hence, the interelement compatibility should


43

One- Dimensional Stress Deformation

44

include order of v at least

up

to

(zero), that

is,

Chapter 3

displacement

v.

In general,

should provide interelement compatibility up to order

the formulation

n
1, where n is the highest order of derivative in the energy function.
Approximation functions that satisfy the condition of compatibility can be
called conformable.

The

other and important requirement

is

that the approximation function

should be complete; fulfillment of this requirement will assure monotonic


convergence. Monotonic convergence can be explained in simple terms as a
process in which the successive approximate solutions approach the exact
solution consistently without changing sign or direction.

For

instance, in

approximate areas approach the exact area in such a way that


each successive value of the area is smaller or greater than the previous value
of area for upper and lower bound solutions, respectively.
Completeness can be defined in a number of ways. One of the ways is to
Fig. l-8(c) the

relate

it

to the characteristics of the chosen approximation function. If the

function for displacement approximation allows for rigid body displacements

(motions) and constant states of strains (gradients), then the function can be

considered to be complete.

mode
For

rigid

body motion represents a displacement

that the element can experience without development of stresses in

it.

instance, consider the general polynomial for v as

=a +
2

cc 2

In Eqs. (3-3a) and (3-6)

a 3y 2

a 4 j> 3

we have chosen

the general polynomial as

n
cc n+

(3-3d)

iy

a linear polynomial by truncating

shown by the

vertical

dashed

line.

The

linear

approximation contains the constant term a, which allows for the rigid body
displacement mode. In other words, during this mode, the element remains

and does not experience any strain or stress, that is, a 2 y = 0.


The requirement of constant state of strain (e y) for the one-dimensional
column deformation is fulfilled by the linear model, Eq. (3-3a) because of the
existence of term a 2 y. This condition implies that as the mesh is refined, that
in each element
is, the elements become smaller and smaller, the strain e
y

rigid

approaches a constant value.


In the case of one-dimensional plane deformations in the column, the

condition of constant state includes only e y

may

the

first

derivative or gradient

and more general


problems such as beam and plate bending. In such cases, it will be necessary
to satisfy the constant strain state requirement for all such generalized strain
or gradients of the unknowns involved, e.g., see Chapters 7, 11-14.
In addition to monotonic convergence, we may be interested in the rate
of

v.

Additional constant strain states

exist in other

of convergence. This aspect is often tied in with the completeness of the


polynomial expansion used for the problem. For instance, for the onedimensional column problem, completeness of the approximation function
requires that a linear function, that

is,

a polynomial of order () equal to one

Chapter 3

is

One- Dimensional Stress Deformation

needed. Completeness of the polynomial expansion requires that

including and up to the


satisfied for the linear

from Eq.

first

order should be included. This

model, Eq. (3-3a) since

(3-3d). In the case of

cubic approximation function


is

45

all

terms up to n

beam bending, Chapter


is

6,

to

and includes the order

It

may happen

not include
satisfy the

all

we

all

it

terms

automatically

are chosen

shall see that a

Then

required to satisfy completeness.

necessary to choose a polynomial expansion such that

up

is

includes

all

it

terms

3.

that an approximation function of a certain order (n)

may

terms from the polynomial expansion, Eq. (3-3d), and

still

requirements of rigid body motion and constant states of

As an example of a complete approximation

satisfying rigid

strain.

body motion and

constant state of strain in a two-dimensional problem, but not complete

of the polynomial expansion, see Chapter 12, Eq. (12-9).


For two-dimensional problems, the requirement of completeness of polynomial expansion can be explained through the polynomial expansion represented by using Pascal's triangle, see Chaps. 11-13.
Here we have given rather an elementary explanation of the requirements
for approximation functions. The subject is wide in scope and the reader
interested in advanced analysis of finite element method will encounter the
subject quite often. For example, the completeness requirement can be further
explained by using the so-called "patch test" developed by Irons [2]; it is
discussed in refs. [3] and [4]. Moreover, the approximation model should
satisfy the requirements of isotropy or geometric invariance [5]. These topics
are considered beyond the scope of this text.
in the sense

Step

3.

Define Strain-Displacement and Stress-Strain Relations

For stress-deformation problems, the actions or causes (Chapter 2) are


and the effects or responses become strains, deformations, and stresses.
The basic parameter is the strain or rate of change of deformation. The link
connecting the action and response is the stress-strain or constitutive law of
the material. It is necessary to define relations between strains and displaceforces,

ments and stresses and strains for the derivation of element equations in
Step 4. Hence, in this step we consider these two relations. We note at this
stage that although we use familiar laws from strength of materials and
elasticity for the stress-deformation

problem,

the relations relevant to specific topics.

flow problem (Chapter

4),

the relation

in later chapters

we

shall use

For instance, in the case of the fluid


between gradient and fluid head and

Darcy's law will be used.

Returning to the axial deformation of the column element, the


displacement relation, assuming small strains, can be expressed as
ey

= ^>
ay

strain-

(3-12a)

One- Dimensional Stress Deformation

46

Chapter 3

where e y = axial strain. Since we have chosen to use the local coordinate L
and since our aim is to find dv/dy in the global system, we can use the chain
rule of differentiation as

Now, from Eq.

(3-2)

dv

,~
101 v
(3 " 12b)

= d_(l^M)=\
dy\

dy

dL

dL dv
-ayTL'

we have
d_L

and from Eq.

dv
= Ty =

>

(3-12c)

1/2

(3-8)

= ^[(1 -

L)v t

+ i(l +

L)v 2 ]

1]N.

[-1

(3-12d)

Substitution of Eqs. (3-12b)-(3-12d) into Eq. (3-12a) leads to

= -}-[-

e,

1]H

(3-13a)

or

{,}
(1

where

[B]

(1//)

x [ 1

1)

{q}
2) (2

1)

is

a one-dimensional problem, the strain vector

contains only one term, and the matrix [B]

{e y}

(3-13b)

can be called the strain-displacement transforma-

1]

tion matrix. Because this

[B]
(1

is

only a row vector.

retain this terminology for multidimensional problems

We

shall

however, with multi-

dimensional problems these matrices will have higher orders.

The student can

easily see that Eq. (3-1 2d) indicates constant value

strain within the element; this

the displacement.

is

because

we have chosen

of

linear variation for

We can then call this element a constant-strain-line element.

STRESS-STRAIN RELATION
For

simplicity,

we assume

elastic (Fig. 3-4).

that the material of the

column element

is

linearly

This assumption permits use of the well-known Hooke's

law,

ay

Ey y

(3- 14a)

or in matrix notation,
{,}
(1

where

[C]

is

1)

[C]
(1

{,}

1)(1

(3-14b)

1)

the stress-strain matrix. Here, for the one-dimensional case,

matrices in Eq. (3- 14b) consist of simply one scalar term.

One- Dimensional Stress Deformation

Chapter 3

47

i\

A^
= U cl
u

III

Figure 3-4 Linear elastic constitutive or stress-strain (Hooke's) law.

Substitution of Eq. (3-13) into Eq. (3-14b)

now

allows us to express {a y}

in terms of {q} as

{y }
Step

4.

(3-15)

Derive Element Equations

number of procedures

Among

= [C][B]{qJ.

are available for deriving element equations.

and residual methods. Principles based on


and complementary energies and hybrid and mixed methods are
used within the framework of variational methods. As described in Chapter
2 and in Appendix 1, a number of schemes such as Galerkin, collocation, and
least squares fall under the category of residual methods. We shall use some
of these methods in this chapter and subsequently in other chapters.
these are the variational

potential

PRINCIPLE OF
In simple words,

MINIMUM POTENTIAL ENERGY


if

a loaded elastic body

is

in equilibrium

under given geo-

metric constraints or boundary conditions, the potential energy of the

deformed body assumes a stationary value. In the case of linear elastic bodies
is a minimum; since most problems we consider
involve this specialization, for convenience we shall use the term minimum.
Figure 3-5 shows a simple axial member represented by a linear spring
in equilibrium, the value

with spring constant k(F/L). Under a load P, the spring experiences a displacement equal to v.
The potential or the potential energy n p of the spring is composed of two
load (see Chapter 2)
parts, strain energy U and potential
p of the external

n,

The

strain energy

W
u+ W

B.

(3-16)

can be interpreted as the area under the stress-strain


when we minimize II,, we differentiate or

curve (Fig. 3-4). Mathematically,

One- Dimensional Stress Deformation

48

Chapter 3

IP

w^%
Figure 3-5 Idealized linear spring.

take variations of Tl p with respect to the displacement v. While doing this we


assume that the force remains constant, and we can relate variation of work

W by the load and the potential of the load as

done

SW

6Wr

(3-17)

where 8 denotes arbitrary change, variation, or perturbation. For our purpose

we can consider

it

to imply a series of partial differentiations.

sign in Eq. (3-17) occurs because the potential

into
is

work by

these loads,

The negative

of external loads

is

lost

W. Then the principle of minimum potential energy

expressed as

sn B = su + 8wD =

su-sw =

(3-18)

o.

There are two ways that we can determine the minimum of II,: manual
and mathematical. Both involve essentially an examination of the function
represented by lip until we find a minimum point. For simple understanding,
we first consider the manual procedure and write the potential energy for
the spring (Fig. 3-5), assuming undeformed state of the spring as the datum
for potentials, as

-Pv
Ikv - Pv,
2

(3-19a)

where kv = force in the spring and %(kv)v denotes strain energy as the area
under the load-displacement curve (Fig. 3-5). Since the load in the spring goes
from
to kv, we have to use average strain energy. The term Pv denotes
the potential of load P; since we have assumed P to be constant, this term
does not include . We further assume that P = 10 units and k = 10 units
per unit deformation. Then

U p = \\0v 2

Now we

search for the

a positive v

is

5v 2

minimum by examining

various values of deformations


table

lOv

assumed

v.

The

(3- 19b)

10v.

values of the potential Il p for

results are

shown

in the following

to act in the direction of the applied load

One- Dimensional Stress Deformation

Chapter 3

n,

-2.000
-1.000

+40.000

49

+ 15.000

0.000

0.000

0.125

-1.1719
-2.1875
-3.7500
-5.0000

0.250
0.500
1.000

2.000

0.0000

3.000

15.0000

4.000

40.0000

5.000

75.0000

etc.

Figure 3-6 shows a plot of lip versus


value at v
will

1.

deform by

On

Hence, under
1

known

is

can be seen that lip has a

when

minimum

in equilibrium,

we can perform the procedure of "going to and


minimum by using mathematics. It is

11, to find its

that a function assumes a

derivative

v. It

10 units, the spring,

unit.

the other hand,

on the function

P=

zero.

Applying

minimum

\kvbv

Pdv

well

value at a point(s) where

this principle to II, in

SU,

fro"

Eq.

(3- 19a),

we

its

obtain

=0

or
(kv

- P)Sv = 0.

(3-20a)

Figure 3-6 Variation of potential energy.

40

Is**
30
20

\
\ 10

\
-3

-1

^^" '2

Minimum

point

-10
-20
-30

One- Dimensional Stress Deformation

50

Since Sv

is

arbitrary, the term in parentheses

Chapter 3

must vanish. Therefore,

kv-P =
or

=P,

kv

(3-20b)

which is the equation of equilibrium for the spring. STl p = in Eq. (3-20) is
analogous to equating d\\ p \dv = 0, which will result in the same equilibrium
equation (3-20b). Substitution of the numerical values gives
\0v

Therefore, v

We
method

1.0 unit, the

10.

same answer

as before.

note here that in most problems solved by using the

Hp

is

a function of a large

ments. Consequently,

it

is

number of parameters

most economical and

matical methods because the manual procedure

finite

element

or nodal displace-

direct to use the


is

mathe-

cumbersome and

often

impossible.

The mathematical procedure involved minimization of lip. For simplicity,


we may view the process as simply taking derivatives of Tl p In general, how.

ever, the minimization will involve calculus of variations. In

treatment in this book

we

Now we return to the column element (Fig.

n> =

most of the

shall use the simple differentiation concept.

3-7)

and write FL

as

[5]

c e dv
iff ^ ' >

iff

fvdv

if

T'vds

S Pv

Figure 3-7 Generic column element with loads.

jp

U\

%
\

(3-2 la)

One- Dimensional Stress Deformation

Chapter 3

where \o y e y

strain energy per unit

Ty =

(weight) per unit volume,

area,

= part

M number

maximum

value of

vt

level,

M=

Y = body

Pa =

applied

displacement corresponding to

the joint forces

P a is now appropriate.
Pu can be treated as

Pih

Since a

applied to the total structure,

is

force

2.

contribution of the joint force

we

volume,

of points at which joint forces are applied; here the

comment concerning

joint force

surface loading or traction per unit surface

of the surface on which surface loading acts,

nodal (joint) forces at local

and

volume,

51

Pig

applied at point

a local

in the global sense. Later

more convenient to add contributions of concentrated


when we consider potential energy of the entire body;
becomes clearer when the total structure is considered.

shall see that

it is

joint or nodal forces


their relevance

The terms

Eq. (3-21 a) are essentially similar to those in Eq.

in

now we

except that

assign

volume

(3- 19a),

to the element instead of treating

it

as a

spring.

For the present, we assume that the cross-sectional area A of the element
is

constant; then Eq. (3-2 la) reduces to

- A\

Tyvdy -

n,

Ty

is

the (surface) loading per unit length along the centerline of the

Here

idealized line

o y e y dy

column

Yvdy-

Pav

(3-21b)

(Fig. 3-7).

Equation (3-2 lb) can

now

be expressed

in

terms of the local coordinate

system by using the transformation of Eq. (3-2) as

dy

(3-22)

-L-dL.

Therefore,
1

11,

Next we

ay y

dL-^C

YvdL-l-j TyvdL-tPVi'

(lc)

J"'

and a y from Eqs.

substitute for v, y9

(3-8), (3-13),

and

(3-14),

respectively, in Eq. (3-21c) to obtain, in matrix notation,

n* = T

j\
_1

(1

1)(1

-yj "'

^WL "t

[C]

{'

1)(1

[N]
(1

2) (2

(1

1)

dL-^

T,

{q}
1)(1

[N]

1)

x 2)(2 x

Pu
(1

f dL

{q}
1)(1

1)

(3-21d)

v,

l)(l

1)

One- Dimensional Stress Deformation

52

Chapter 3

or

=^

{qf

[W

(1

Al

2) (2

-yj

[B]

x 1)(1 x

1)(1

[N]
(1

[C]

{q}

2) (2

[N]

1)(1

2) (2

1)

M
r

2) (2

1)(1

(3-2 le)

(1

1)

dL

fy dL-

{q}

dL

{q}

(1

1)

1)(1

1)

where Y and Ty are assumed to be uniform.


Equation (3-21) represents a quadratic function expressed in terms of
v and v 2 In matrix notation, transposing in Eq. (3-2 Id) is necessary to make
.

the matrix multiplication in {e } r [C]{ >,} consistent so as to yield the scalar


3;

a y e y = Eel m Eq. (3-2 lc). The need for transposing will become
when we expand the terms in Eq. (3-2 lc). The last term denotes sum-

(energy) term
clear

mation,

P uv

+P

M=

2 iV 2 , if

2.

EXPANSION OF TERMS

We now
them

consider the

first

three terms in Eq. (3-2 le) one by one

and expand

as follows:

First term

"-*!> <r;j4<-.
AE
4/

{v\

2v v 2
l

..{:)-

v\)dL.

(3-23a)

Second term:

AIY

W
Here

is

assumed

fy

for uniform

[y(l

to be

L)v,

+ i-(l +

L)v 2 ~\^dL.

(3-23b)

uniform gravity load per unit volume. Similarly,

the third term yields

pi

Finally, the

'

sum of

Eq. (3-2 le) gives

J',

[t (1 "

L)Vl

(1

+ L >*\ dL

(3 " 23c >

Eqs. (3-23a), (3-23b), and (3-23c) and the last term in

One- Dimensional Stress Deformation

Chapter 3

l',[i
Notice that

2) to find its

invoke the principle of

minimum

+ L>

|</L

C P,^.

if J

minimum

potential energy (see Chapter


x

\\

i-(l

value by differentiating E^, with respect to v and v 2 as

Tyl

gig

L) Vl

a quadratic or second-order function in Vj and v 2

II p is

Now we

(1

53

(1

<" 2 ^

"
+

L)dL

"

2v ^ dL

Pl/

~ ^f\\

(3 " 25a)

'

(1

L)dL

^2

Although we used

(1

L)flfL

P2/

(3 " 25b)

and partial differentiations to find


advanced applications to multidimensional problems,
it is more concise and convenient to use calculus of variations; then both Eq.
(3-25a) and Eq. (3-25b) are written together as

the

minimum of lip,

differential calculus

for

SU P =
where 6

is

the variation notation.

We

(3-26)

0,

shall briefly discuss variational princi-

ples at various stages, but the majority of derivations in this

book

will

be

obtained by using differential calculus.

INTEGRATION
For the one-dimensional problem, the integrations in Eq. (3-25) are simple;
in fact, the first term in these equations is independent of the coordinate L

hence integrations involve only constant terms. Thus,

aeV
41

(2,

2v 1 )dL

= ^(2t>, -

-1

= <f(v, A

4\

2v 2 )L

4)
AIY
2

One- Dimensional Stress Deformation

54

Chapter 3

Integrations of other terms in Eq. (3-25) lead to

^0>, ^fi-Vi

v2)

+v

-Mf-TA- P u =

)-^-^- P

2l

0,

(3-27a)

= 0.

(3-27b)

In matrix notation,

AE

-1

or

= {Q}.

M{q}
Here
it is

[k]

stiffness

(3-28b)

matrix of the element, and with the linear approximation

identical to the matrix of stiffness influence coefficients in matrix struc-

tural analysis;

{Q}

it is composed of body
and the joint loads. It is

element nodal load vector, and

force (due to gravity), surface traction or loading,

interesting that the finite element derivations with the linear


result in

approximation

AIY

lumping of the applied loads (equally) at the two nodes:

= weight of the element, where Al = volume of the element, and fy = total


l

surface load.

We note that if higher-order approximations were used, the load


come out

vector would not necessarily

as

lumped loading; we

shall consider

such cases at later stages.

Equation (3-28) provides a general expression that can be used repeatedly


to find stiffness relations for

In the foregoing,

we

all

elements in the assemblage.

derived element equations by differentiating the

expanded function for Tl p This is possible because there are only two variables, v and v 2 Advanced problems, however, involve a large number of
variables, and we commonly write the results directly in terms of matrix
equations. For instance, differentiation of II P in Eq. (3-21e),
.

dn

_Q

*n*-o.
dv 2
leads to

*L

[BY
(2

[B]

[C]
1)(1

1)(1

dL{q}

Al
2

(2

2)

Y dL

[Nf
_1

1)(1

1)

dL

f,

(2x

1)(1

1)

{P//}
(2

/=

1)

1,2,

or
[k]{q]

= {Q},

(3-28b)

Chapter 3

One- Dimensional Stress Deformation

55

where
ai f

h=%

[B]

E[B]dL

and
(Q}

The terms

in

y J'

[NTF^L

+ -J-J"

[Nff/L -

{P,}.

Eq. (3-28) are the same as in Eq. (3-25), except that here they
The transpose in [N] r arises because after

are arranged in matrix notation.

differentiations [Eq. (3-25)] the force terms yield

order of 2 in the final results. After the process

?.

load vector that has an

on n^, the
a set of (linear) simultaneous equations [Eq. (3-28b)].
Often, the stiffness matrix [k] is called an operator, which means that
when operated on (nodal) displacements, the results are the (nodal) forces.

end

result

o. differentiation

is

COMMENT
The primary

task in finite element formulation can be considered to

involve derivation of element equations.


involve assembly, patching

The subsequent steps essentially


up or combination of elements, and use of linear

algebra for solution of the resulting simultaneous equations.


Step

5.

Assemble Element Equations

to

Obtain Global Equations

POTENTIAL ENERGY APPROACH


Although, for simple understanding, we considered the equilibrium of a
single element in the foregoing step, it is necessary to emphasize that it is the
equilibrium of the entire structure in which we are interested. Consequently,
we look at the total potential energy of the assemblage and find its stationary
(minimum) value. The procedure of assembling element equations can be
interpreted through minimization of total potential energy.

As an example, we

consider the column (Fig. 3-1) divided into three

elements with four nodes (Fig.

3-8).

Here we have numbered nodes starting from the top of the column and
have measured the global coordinate y as positive downward. This is just
for convenience; for instance, since the loads act downward, they are positive,
and displacements are positive. It is, however, possible to measure y as
positive upward or downward from any convenient point and to number
nodes in other consistent fashions.

By using Eq. (3-24) we can write the potential energies H p for each element in the assemblage and add them together to obtain total potential
energy IP, as

One- Dimensional Stress Deformation

56

Global

Local

Nodes

Displacement

Nodes

Properties

Displacement

Chapter 3

" y

1
J

v2

An to
,

v?

v4

Figure 3-8 Discretization of column and numbering.

^fj"'

W - 2v v
3

+ vl)dL

^f

[y

i>.

yO

+ L ^] rfL

y^ J'

[yd -

L) Vl

-1(1

+ L)v^dL

y^

[yd " L> + yd + L)4 ]rfi

fir

J'

%^ J'

/',

- P!,, -

[t

~~

L)v
'

+ T (I + LKl rfL

[yd " iK + yd +

(PL

+ P*u)v - (P|, +
2

),]<.

P\,)v 3
(3-29)

One- Dimensional Stress Deformation

Chapter 3

57

Here we assume that A,E, /, Y, and fy are different for different elements and
an element number, e denotes an element, the super= total number of elements. Note
script on P denotes an element, and
that the local joint or point loads at common nodes are added together and
2
yield the global joint or point loads at these points. For instance, PJ, + P U
gives global point load at global node point 2.
For global equilibrium we minimize 11^ with respect to all four nodal
displacement unknowns, v u v 2 v 3 and v 4 Thus,
their subscript denotes

dU

dv f

^f
-

<2

^ -

2v ^ dL

^r f y
1

({

L)dL

[yO -

- Ph =

L)\dL

(3-30)

0,

and so on. In combined variational notation we write

5n p

(3-31a)

which denotes
cttPp

dn<p
dv 2
(3-3 lb)

dU p
[

dv 3
dll'p

dv 4
After the required integrations and arranging the four equations in matrix
notation,

we

AiEi

\-AiEi
ll

obtain

-AiEi
h

A\E
h

v\

A2E2

-A2E2

h
-A2E2
h

V2

A2E2

A 3 E3

-A E

-A E
3

A 3 E3
h

h
A\\\Y\

_j_

Ty\ll

Axhfx

fylh

P\,

A2I2Y2

fylh
2

^3/3^3
fylh
A2I2Y2
fylh
-^
+ -2T + -l~ + -2~
A Y

+ 2 ^p
,

3 l3

~2

7^3/3

2/

Pii+Ph
(3-32a)

Ph + Pl

+
One- Dimensional Stress Deformation

58

Chapter 3

or

[K]{r]

(3-32b)

{R},

where [K] = assemblage stiffness matrix, {r} = [v v 2 v 3 v 4 ] = assemblage


nodal displacement vector, and {R} = assemblage nodal load vector.
r

METHOD

DIRECT STIFFNESS

The foregoing approach can be explained and understood alternatively


through the direct stiffness method [6]. We note, however, that the basic idea
of assembly evolves essentially as a result of the minimization of total potenenergy.

tial

close look at Eq. (3-32a)

added together;

this

shows that the

common node

corresponding to the

coefficients

indicated by enclosing

is

common node

stiffness

for elements

them

in

dashed

or influence

and 2 are

lines.

Simi-

added together. This interpretation


can lead to the familiar concept of obtaining the assemblage matrix by adding

larly,

loads at the

are also

individual element matrices of contributing elements through the direct

stiff-

ness approach.

Let us express the matrix equations for the three elements by labeling the
terms with subscripts as

*
>

Global
Local
1

A E
X

"

-r

Local Global
\v\

>vA

i_ [vl

>v 2 \

_-l

>
>

Global
Local
2

2
1

"

1
A 2 E2
h _-l

-r

>

"

1
A E
h _-l
3

r 1/
(3-33a)

Local Global
\v\

i_ [vl

>

Local

>v 2

>v

_A
Global

-r

\
2 l2

Y2

^}

(3-33b)

Local Global

h\
3

i_ [v 2

>v
vj
3)

>

_A

3 l3

Y n\
3

,
,

Ty3 l

K;:l-

(3-33c)

Chapter 3

One- Dimensional Stress Deformation

Here the superscript indicates the element; for instance,


node 2 for element 1, and so on (Fig. 3-8).

59

v\

is

the displacement

at

To assemble
order of

these three relations,

we note

that the total global degrees

and hence the assemblage matrix and load vector are of the
Consider Eq. (3-34) in which we assign

of freedom are
4.

4,

A,E

-A E
X

AxE

A 2 E2

-A2E2

v\

-A2E2

A2E2

fAthfi
2

A3E3
h

A3E3

v2

A3E3

A3E3

=v

v\

v\

v4

+ 7^, +p}/

AxhY\

Tylh

A2I2Y2

Tylil

+ p\l +

A212Y2
2

Tylh
2

==

(3-34)

<

blocks of 4

x 4

A 3 hY 3

Ty3h

A3I3Y3

+n

+ p32i

for the assemblage stiffness matrix

blage load vector.

fy 3 l 3

and 4 X

for the assem-

Now we insert the coefficients of the matrices for the three

elements [Eq. (3-33)] into the proper locations in Eq. (3-34). For instance,
the coefficient in Eq. (3-33b) for element 2 corresponding to the local indices

added to the global location

(2, 2) is

(3, 3)

and so on. Similarly the nodal

loads corresponding to the local index 2 are added to the global location 3

and so on.

We

notice that Eq. (3-34)

direct stiffness

approach

the displacements at the


that

is,

is

is

the

same

as Eq. (3-32).

We

also see that the

based essentially on the physical requirement that

common

nodes between elements are continuous;

there exists interelement compatibility of displacements at the

common

nodes.

BOUNDARY CONDITIONS
we

now we have

concentrated only on the properties of the column. Next


consider the physical conditions that support the column in space, because

Until

the foregoing stiffness properties are called into action only


is

when

the

column

supported. This leads us to the concept of boundary conditions or con-

straints.

As

the

name

displacement or

implies, a
its

boundary condition denotes a prescribed value of


on a part of the boundary of the structure or

gradient(s)

One- Dimensional Stress Deformation

60

body.

boundary condition

we

space. After

how

us

tells

body (column)

the

introduce these conditions,

Chapter 3

we have

is

supported in

a structure that

is

ready

to withstand applied forces; in lieu of these conditions, Steps 1-6 resulting


in

tell us only about the capability of the column to withstand


and not how they are withstood. In other words, without boundary

Eq. (3-32)

forces

conditions, the stiffness matrix [K]


ishes,

and there can be an

infinite

singular, that

is

number of

is,

its

determinant van-

possible solutions. Hence, the

equations in (3-32) cannot be solved until [K]

is

modified to

reflect

the

boundary conditions.
TYPES OF BOUNDARY CONDITIONS
As discussed in Chapter 2, we encounter three kinds of boundary conditions. They are prescribed (1) displacements (or other relevant unknowns),
(2) slopes or gradients of unknowns, or (3) both. They can also be called first
or Dirichlet, second or Neumann, and third or mixed boundary conditions,
respectively.

For

instance,

at the base

is

if

gradient or slope
fully fixed,

it

the

specified,
is

column

is

the

it is

specified,

constitutes the

supported such that the axial displacement

it

first
is

condition;

if

the base

the second condition;

is

fixed such that

and

if

the base

is

mixed condition.

The boundary conditions

in

terms of the given displacement are often

called geometric or forced, while those in terms of gradients are often called

natural: the latter can often relate to generalized forces prescribed

on the

boundary.

HOMOGENEOUS OR ZERO-VALUED BOUNDARY CONDITION


As an
that

is,

at

illustration, let us

nodal 4

assume that the displacement at the column base,


one can specify a nonzero displace-

(r 4 ) (Fig. 3-8), is zero;

ment also if the base experiences a given settlement.


The boundary condition can be simulated in the finite element equations
by properly modifying the assemblage equations [Eq. (3-32)]. To understand
this modification, we shall consider one of the available procedures by writing Eq. (3-32) in symbolic form as
1

R.

\Vt

v2
>

^32

We

delete the fourth

^34

K43

^44 _

row and

(3-35)

<

v3

"-33

JV

iJ

the fourth co 1 umn corresponding to v

which leads to the modified equations


"

23

33_

>r
<

as

'-':

^'3,

Jtl
>

= R

>

<

,*3,

(3-36a)

One- Dimensional Stress Deformation

Chapter 3

61

or

= {R}.

[K]{r]

(3-36b)

The overbar denotes modified matrices.


The assemblage

matrix [Eq. (3-32)]

stiffness

symmetric and banded or

is

Symmetry implies that k u = kJt and handedness implies


nonzero values of ktJ occur only on the main diagonal and a few off-

sparsely populated.
that

diagonals, whereas other coefficients are zero. These properties, which occur
in

many

make

engineering problems,

economical. This

solution of the equations easier and

achieved by storing in the computer only the nonzero

is

we

elements within the banded zone. In Appendix 2

discuss the solution of

systems of such simultaneous equations.

NONZERO BOUNDARY CONDITIONS


nonzero value, then the procedure is
which can include the homogeneous condition as a
special case. For instance, assume that v 4 = 5. Here Eq. (3-35) is modified
by setting the coefficient k Ait equal to 1, all other coefficients in row 4 as 0,
and R A = S. Therefore
If the specified displacement has

somewhat

different,

ku

fc,2

k 21

k 21

"

k 23

R>

v2
<

k 32

k 33

These equations can

now be

v3

34
_

= *

(3-37a)

v4

3
c5
.

solved for v u v 2 v 3 and v 4 Often,


,

when we take advantage of

when

dealing

symmetry
of a matrix for reducing storage requirements, it becomes advisable and
necessary to restore the symmetry of the matrix in Eq. (3-37a) that is broken
by the modification. It can be done by substracting from R l9 R 2 and R 3 the
quantities k l4 X 3, k 24r x 3, and k 34r x (5, respectively. Thus, the equations
with large matrices (Appendix 2)

the

reduce to

k 21

0"

*I1

k 23

R,

v2

>

kzi

k 33

Here kt4 x 8

(i

1, 2,

3) is

(3-37b)

v3
Ij

-0

<

k 34r x 6

v*.

transposed to the right-hand

side.

For instance,

for the third row,

0xt),+

k 32 v 2

k,*

x S

=R

or

+ k 32 v +
2

k 33 v 3

=R 3

k 34 S.

(3-37c)

The foregoing boundary conditions expressed in terms of the unknown


displacement are the forced or geometric constraints. The natural boundary

One- Dimensional Stress Deformation

62

Chapter 3

conditions usually do not need the special consideration required for the

geometric boundary condition. The natural boundary condition,


as a zero value for a derivative (slope) of the
cally in

the

an integrated sense

sum of

the

in the finite

unknown,

is

if specified

satisfied

automati-

element formulation. In other words,

computed values of the

derivative at the

boundary vanishes

unknowns and

different categories

approximately.
Different problems involve different

we

of boundary conditions, which


Step

Solve for Primary

6.

Equation (3-32)
set for the

is

shall discuss.

Unknowns Nodal Displacements


:

a set of linear algebraic simultaneous equations.

column problem

is

linear

because the coefficients

tJ ,

The

which are

composed of material properties (E) and geometric properties (/, A), are
constants and do not depend on the magnitude or conditions of loading and
deformations. For instance, we have assumed linear Hooke's law and small
strains and deformations in the formulation of the foregoing equations. We
note here that if we assume material behavior to be nonlinear and large
strains and deformations, then Eq. (3-32) will be nonlinear. Although we shall
briefly discuss nonlinear

topic

is

behavior at later stages, detailed discussion of

beyond the scope of this

The unknowns
assemblage vector

this

text.

in Eq. (3-37) are the nodal displacements given by the


r
[v x v 2 v 3 v 4 ] in which v 4 is already known. The

{r}

equations can be solved by using direct, iterative, or other methods; see

Gaussian elimination and a number of its modifications are one


sets of direct methods used for solution of the finite element
equations. The direct procedure involves two steps: elimination and back

Appendix
of the

2.

common

substitution

To

Example

We

[5, 7].

illustrate the solution

procedure,

we now

consider two examples.

3-1

assume the following data for the column problem

(Fig. 3-8):

Cross-sectional area,

E = 1000 kg/cm 2
= 10 cm;
A 1.0 cm 2

Surface traction,

fy =

Body

Y=

Modulus of elasticity,
Element length,

weight,

Boundary condition,

v4

kg/cm

.0

0.5

kg/cm 3

0.0.

We also assume that the column has uniform cross-section and material properties.
Substitution of these values into Eq. (3-28) leads to the following element and

assemblage equations

One- Dimensional Stress Deformation

Chapter 3

1000 x

nw

10

x 10 x

lJl v2

63

a*it\)
*H\) +

or

-100"

100

3c:KM;i

100^

100

and so on for the other two elements. Assembly of the three elements [Eq.

(3-32)]

yields
"

100

100

100

200

100

100

200

Vi

[2.5]

v2

5.0

100

100_

100

100

100

200

100

100

200

5.0

,Vt.

.2.5]

5|

10

^3

Introduction of the boundary condition v A

>

-100

'

10
5

leads to the modified equations

"

In the expanded form,

100v 2

-100^ +

200v 2

100v 2

lOOvj

For solving these equations,

first

+0
+

=
=

100v 3
200^3

7.5,

(3-38a)

15.0,

(3-38b)

15.0.

(3-38c)

we

follow the elimination procedure. Equation

7.5

(3-38a) gives

100^!
Substitution for

100^

in

100v 2

Eq. (3-38b) yields

-7.5

100v 2

200^2

100^3

15

or

100v 2

Now, Eq.

22.5

100v 3

(3-38c) gives

-22.5

I00v 3

200v 3

15

or
37.5
v3

100

cm.

Back substitution leads to


v2

100

cm

and
67.5
V\

100

cm.

Example 3-2
Instead of uniform

and

fyy

equal to 10 kg at the top, that

example we apply only a concentrated load


node 1. We can look upon this load as a global

in this
is,

at

One- Dimensional Stress Deformation

64

joint load

P lg

Chapter 3

In this case, the final equations are

lOOvj

-100^!

100v 2

+
-

200v 2
100v 2

=10,

-f

- 100^3 =
+ 200v =
3

(3-39)

0,
0.

Note that the equations on the left-hand side are the same as before only the rightis changed. This gives us an idea that for linear problems, where
Kij do not change, for a given structure, the elimination process is required only
once since we can store the elimination steps in the computer. For different loadings
only the back substitution needs to be performed. Thus, we obtain
;

hand-side load vector

Step

7.

vi

v3

=^cm

cm

-3$q

v4

(specified)

Solve for Secondary Unknowns; Strains and Stresses

For the displacement formulation based on potential energy, nodal disWe call them primary because they
are the main unknowns involved in the formulation and in Eq. (3-32). The
secondary unknowns are those derived from the primary unknowns evaluated
in Step 6. For stress analysis problems, these are strains, stresses, moments,
shear forces, etc. These designations will change depending on the type of
formulation procedure used (see Chapter 1 1).
placements are the primary unknowns.

STRAINS AND STRESSES


By

substituting the

computed displacements from Step


computed as follows

strains in the elements are

Example
For element

3-1 (continued)

ey (D

iQ

[-l

-7.5
1000"

and

Similarly, the strains for elements 2

^=

3 are

22.5
1000

and

37
tyK

(67.5)

l]^o(

1000

60

6 into Eq. (3-13),

One- Dimensional Stress Deformation

Chapter 3

Axial stresses in the three elements can

The negative

strains

OyV)

(3)

100Q

now

x 1000

X 1000

mi x

100

be derived by using Eq. (3-14):


-

-7.5 kg/cm 2

-22.5,

-37.5.

sign denotes compressive stresses.

Example 3-2

The

a y {\)

-7.5

65

and

(continued)

stresses are

-10
y(l)

1000'

-10
ey {2)

1000'

y (3)

1000'

-10
and
<T y

(D

<7,(2)

OyO)
Step

= ~10.0 kg/cm 2
= - 10.0,
= - 10.0.

Interpretation and Display of Results

8.

Figure 3-9 shows plots of displacements and stresses for the two previous
examples. For Example 3-1, the distribution of displacements within each

element

is

linear,

need not be

whereas for the entire column, the displacement distribution


The computed stresses are constant within each element,

linear.

whereas the actual distribution over the entire column can be linear. Existence
of the constant stress states is dictated by the assumed displacement model,
which is linear.
In

Example

the exact value

3-2, the computed displacements and stresses are the same


from closed form solutions, because for the exact solution,

oy

_4A =

^=

as

-10 kg/cm 2

and
Vl

_ PQl) _

~ AE ~

10 x 30
1000 x 1

30
100

An important observation can be made at this stage. The accuracy of the


assumed approximation will depend on the type of loading, geometry, and
material properties. In Example 3-1, the loading varies from zero at the top
to the highest value at the base for the gravity load Y and is uniformly

One- Dimensional Stress Deformation

66

= 0.675

v,

cm

Chapter 3

a = -7.500 kg/cm 2

22.50

Computed

= 0.600

v2

Actual

v3

37.50

= 0.375

/
/
/

v4 =

w,

0.000

Displacements

Stresses

(a)

0.30

<T\

-10.0 kg/cm 2

cm

t-

/
I

Computed and exact

v2

1
1
1
i

v 4 = 0.00

Disf )lacements

Stresses

(b)

Figure 3-9 Distribution of computed displacements and stresses,


(a) Results for

Example

3-1. (b) Results for

distributed for the surface loading

Example

Ty The computed
.

of the actual, and the accuracy of the

3-2.

stresses are

an average

element computations

finite

is

consid-

ered satisfactory.

For Example

3-2, the

numerical solution

is

identical with the exact solu-

A number of other factors


can influence the accuracy of numerical predictions and can require addi-

tion,

because there

is

only one concentrated load.

tional considerations for

improvements to account for the

factors.

One- Dimensional Stress Deformation

Chapter 3

67

The accuracy of the predictions can be improved by using two methods:


mesh and/or (2) higher-order approximation models. The decision
as to which approach should be used depends on the characteristics of the
problem, and trade-offs exist with respect to accuracy, reliability, and com(1) finer

puter cost.

becomes essential to use higher-order models. If the column


geometry and loading, it may be useful to consider
higher-order models. For instance, we can use a quadratic approximation
Very often

it

(Fig. 3-8) has irregular

for v as

= \L{L = [Ntfq},

X)v x

+ \UL +

\)v 2

(1

-L

(3-40)

where
[N]

1)

{qf

[L(L

L(L+1) (1-L 2 )]

and
[,

The element here has

three nodes (Fig. 3-10), with the third

middle of the element.

It is

In

fact,

we can choose

interpolation functions

node

in the

not necessary to have the third node in the middle.

the node anywhere within the element; then the

TV,

will

be different for such choices.


L =

l = +1

Figure 3-10 Element with quadratic or second-order approximation

model.

The

steps of the finite element formulations can be repeated to derive the

required equations for the higher-order element; this

an exercise; see Prob.

is left

to the reader as

3-8.

FORMULATION BY GALERKINS METHOD


In the case of the variational approach,

we applied

the procedure to a single

generic element and used the results recursively for deriving properties of all
elements. Then the individual element results were combined to obtain the
also emphasized
assemblage equations for the entire discretized body.

We

that in reality the variational procedure


it

was only

for convenience that

to the entire

domain, and

to analyze element

by element.

was applied

we chose

assemblage equations [Eq. (3-32)] were


derived by considering the total potential energy of the discretized body.

To

illustrate this concept, the

first

One- Dimensional Stress Deformation

68

As
dure

explained in Chapter 2 and in Appendix

is

1,

Chapter 3

Galerkin's residual proce-

also applied so as to minimize the residual over the entire domain.

Another important aspect of this procedure is that the approximation function and the functions q> [Eq. (2-12)] are defined over the total domain.
While applying Galerkin's method for finite element analysis, we use the
interpolation functions N as the weighting functions [Eq. (2-14)], which are
defined over the entire domain. In view of the foregoing characteristics of
Galerkin's method and because for the variational procedure we have defined
Ni relevant simply for an element, it is necessary to clarify a number of
aspects and explain the relevance of TV,- over the total domain.
{

Explanation and Relevance of Interpolation Functions

Figure 3-1 1(a) shows an approximation v to the exact solution v* defined


for the total

domain of a one-dimensional body. The approximate or trial


domain can be expressed in

function v for the displacement over the entire


the sense of Eq. (2-12) as

=V

(3-41)

JVjwJ,

k=\ j=\

where the superscript k denotes an element,


is the total number of elements, and m is the number of interpolation functions per element.
For the column with three elements (Fig. 3-8), Eq. (3-41) can be expressed
as follows

Sum for

Sum

elements, k

=2
;=i

1,2,3:

N)v)

-
;=i

Njvj

oxer interpolation functions, j

N)v).

(3-42a)

j=\

1, 2,

m;

= N\v\ + N\v\ + N\v\ +


+ Nlvl
+ N\v\ + N\vl + N\v\ +
+ N mvm
+ N]v] + N\v\ + N\v\ +
+ N m vl
linear approximation [Eq. (3-6)] m = 2; therefore

For the

N\v\
4-

-f

N\v\

N\v\

N\v\

(3-42b)

N\vl

N\v\.

(3-42c)

Here N\ denotes interpolation function for node 2 of element 1, and so on,


where the superscript denotes an element. Figure 3-1 1(b) shows the plots of
the linear interpolation functions over the three elements. Since v\

=v u

One- Dimensional Stress Deformation

Chapter 3

-K

N?

2 2 N k vk

v =

N?

69

N^

N?

N3

Vt

Vt

;<

/^

(b)

NJT 1

+ Ni

subscript => local node

subscript => element

i-1

(0
Figure 3-11 Approximation and interpolation functions in Galerkin's method, (a) Approximation for v over total domain D.
(b) Interpolation functions over

node

v\=v\ =
v

v 2 vl
,

D.

Interpolation function for

(c)

i.

= v\

v 3 and v\
,

=v

we have

= N\v + (N + N\)v + (Nl + N\)v +


= A^- = #i*i + #2*2 + N v + N v
l

N\v A
(3-43)

A,

N denotes interpolation functions around node


general N => N\ + N\. For the linear case [Fig. 3-1 1(c)],
where

1;

for instance, in

One- Dimensional Stress Deformation

70

y
yt

N (y)

- yt-x =
NV,

(3-44a)

-y = N'U

y i+i

-yt

bi+i

<y<y^

y t -\

-yt-\

Chapter 3

S t -u

<y<y
_
=

5,-1

i+ v.

-\

/i-i

Si,

where s

the

1,

and

/,

denotes the length of

The definition of N in Eq. (3-44a) is valid only for points within


domain D. For each end node one of the values will be relevant. Thus
interpolation function around node i has nonzero values only in the two

element
the

<s<

the local coordinates

is

(3-44b)

-y

?*+!

i.

adjoining elements

and

and zero values

in all other elements.

= cross-sectional area

<*Z&

"Y

X*#
Figure 3-12 Equilibrium of column segment.

With the foregoing explanation, we are ready


for formulating finite element equations for the

to use Galerkin's

column problem.

derive the governing differential equation for the column, Fig. 3-12.
sidering the equilibrium of forces for an elemental

method

We

first

By con-

volume of the column,

we have

Acy

- Aay - A^dy + f(y)dy =

or

dGy

~dy'~

f(y).

(3-45a)

oidajdy from Eq. (3-14a) after insertion of e y from Eq. (3-12a)


dropping the subscript to E for convenience)

Substitution
gives (after

AE,d

v*

dy

f(y).

(3-45b)

One- Dimensional Stress Deformation

Chapter 3

Here v*

is

the exact solution,

and we have considered /(j) as surface

f(y) per unit length per unit surface area.

Pa

body force Y(y) and joint load


form as

traction

possible to consider f(y) as


the latter can be defined in a function

PuW = Ptiyj)
where Puiyj)

71

x S(y

It is

- yj

(3-46a)

),

= magnitude of joint load at pointy and S(y yj)

is

the Dirac

delta function,

&(y-yj)
The

'

yj '

(3

(3-6))

is

R(v)

EA^ -f.

(3-47)

For convenience we wrote / instead off(y). Minimization of

46b )

residual for the differential equation (3-45b) (and the linear approxi-

mation function

to

"

R with

respect

[Eq. (3-6)] leads to

VSEA w- f N dy=o
'

(3 48)

'

where h is the total length of the column equal to y 4 the coordinate of the
last node (Fig. 3-8).
Integration by parts for the first term gives
,

' BA&NA = EA%l,N [ - EA


t

and hence Eq.

(3-48)

[ % f *.

(3-49a)

is

<%->=l>^-*4>
We

inserted

the

/,,

(3-49b)

direction cosine of angle between the axis of the

axis, as unity, since cos

9y

= cos =

column and

1.

Now if v from Eq. (3-43) is substituted into Eq. (3-49b), we obtain

/=
The index notation

l,2,3,4,y=l,

implies
set

=
=2
=3
=4
1

and
and
and
and

=
varyy =
vary j =
vary 7 =
vary j

1, 2, 3, 4,
1, 2, 3,

4,

2, 3, 4,

1, 2, 3,

4,

2, 3, 4.

(3-50)

One- Dimensional Stress Deformation

72

which leads

to

~dN,

dN,

dN

dy

dy

dy

dN2 dN

dN,

EA
Jyi

Chapter 3

2
"

dy

dy

dy

dN,

dN,

dNz

dy

dy

dN

dy

dy

dN2

dN,

dy

dy

dN,

dN,

dy

dN,

dN,

dy

dy

'

dN,'

dy

dN2 dN, dN2


'

dy

'

dy

dy

dN,

dN,

dN,

dy

dy

dy

dN,

dN,

dN,

dN,

dy

dy

dy

dy

dy

"

dy

dN4 dN2 dN4

dN,

.dy

dy
'

dy

dN,'

dy

<*>N,
dy

fN,

fN2
\dy

<Jy\

i\

dy.

+ EA\

^N
dy

(3-5 la)

fN,
fN,

or

ffi
It is

important to note that the terms

total length

and that the

f
in

(3-51b)

fR,

Eq. (3-51) are integrated over the

boundary term (dv/dy)Ni

limits in the

fj

are defined

at the ends.

Terms such

as

dN,

dN,

dy

dy

dN

!'

dy

+
vanish because
to

N =
3

in region

y 4 Thus only those terms


.

dy

dy

Jy s

y3

dy

to

dN,

dN,

dy

dy

i"

dy

y 2 and

N =

in regions

dN^

dN2

dy

dy

dy

dN,

EA

dy

r
Jy

dN2 dN2
dy

dN
m

dy

y 3 and
two ele-

to

side reduces to

dy

dN^

dN

dy

dy

dN,

dN

dy

dy

dy

dN2

y2

that have nonzero values in one or

ments contribute to the integration. Hence, the left-hand

dN,

dy

dN

3
'

dN

dy

dN

'

dy
or
[K]{rj

'

dy

dN4

dy

dy

dy

dN<

dN4

dNt

dy

dy

dN,

dy _

(3-52a)

&
*> 5>|

*>

<5>j

^
>j

j!

I> >!
5

73

One- Dimensional Stress Deformation

74

Note

Chapter 3

that the terms enclosed in the dashed lines represent [B] r [B] for the

three elements. This indicates that the contributions of the element stiffness
properties are added such that compatibility at

common

nodes

is

assured.

Integrations for the relevant limits will lead to the assemblage stiffness matrix
[K].

For

instance,

Assuming elements of equal length

=
=

si

*2

Now

ds

= (\/l)dy

dN\

dy

ds

ds

_L

'

dy

we have

(y

-ydlh

(y 3

- y)li-

= (l/l)dy.

and ds 2

dN\

/,

and

(3-53b)

By noting

that

dN = dN

dy

ds

>-

dy

ds 2

we have

= 7T

#22

and

f(lXl)<fc+ f'c-iX-i)^
Jo

(3-53c)

Jo

finally
1

EA

[K]

-1
2

-1

-1

(3-54a)

-1
which is the same as [K] in Eq. (3-32) for constant element length and
form A and E.
The first term on the right-hand side of Eq. (3-5 la) denotes

uni-

>

ffNXdy
ryi

y
1

'fN{dy

{R;

fN\dy

n
\
Jy
2

fNldy

n
\

(3-54b)

fN\dy\

\"fN\dy\
which

is

the

same

as the contribution to the assemblage load vector {R}

by

T, in Eq. (3-32).

The second term on

the right-hand side in Eq. (3-5 la) denotes essentially

the end boundary terms, because

points

and

4.

For

instance,

have nonzero values only

at the

end

Chapter 3

One- Dimensional Stress Deformation

dv
Ndy
since

N-

N\

at

dy

nodes

\,x

and

dy

Hence

4.

15

it

reduces to

(-
dv
dy

AEi

(3-54c)

m>
dv

at

nodes

and

4,

we have

(-<))
R

(3-54d)

K)

dy/*

Equation (3-54d) represents natural boundary conditions defined through


prescribed values of (normal) derivative of the displacement. The term
AE(dv/dy)

= AEe y = Ao y

denotes (internal) force, and hence

{R.}

(3-54e)

where F, and FA are forces at the boundary nodes.


Comments. If an external point load is applied

at node 1, F, equals the


no external load there, F = 0. When the
column base has a prescribed boundary condition, say v4 = 0, then the term

prescribed external load. If there

F4

will

not appear in the

(3-51)] are

is

final results

when

the assemblage equations [Eq.

modified for the given boundary condition. In other words, the

natural boundary condition at a point will not have influence on the final
results if that point has a prescribed geometric

boundary condition.

important to understand the foregoing idea that Galerkin's residual


procedure is applied to the entire domain and that its application yields
It is

boundary conditions which should be properly


use of Galerkin's method

we may

interpreted. In the subsequent

by element, with a clear understanding of the foregoing implications.


When Galerkin's procedure is applied to an element, we can look upon
consider

it

to be applied element

the interpolation functions for the element to have nonzero values within

One- Dimensional Stress Deformation

76

Chapter 3

themselves and zero elsewhere. For instance, in Fig. 3-11, N\ and N\ have
nonzero values within element 2 but zero values at other locations in the
domain.

The foregoing finite element equations from the Galerkin method are
same as those from the variational approach. This happens

essentially the

usually in the case of problems mathematically classified as self-adjoint

a rather elementary

manner

of the symmetric operator matrix


linear or the

problem

is

[8].

In

self-adjointness can be attributed to the existence


[K]. If the

governing equations are non-

non-self-adjoint, the

two

results

can be

different.

An example of a non-self-adjoint problem is given in Chapter 8.

COMPUTER IMPLEMENTATION
Although we could perform hand calculations for the one-dimensional column problem, almost all problems solved by using the finite element method
involve large matrices, and recourse has to be made to electronic
computation.

To understand computer

codes in a progressive manner,

duce a rather simple code (program)

DFT/C-1DFE

in

we plan

Chapter

6.

to intro-

This code

can handle the problems of axial deformation, one-dimensional flow, and


one-dimensional temperature distribution or consolidation.

The reader should study


problems

in

Examples

3-1

the code in Chapter 6 and use it for solving the


and 3-2 and some of the subsequent problems.

OTHER PROCEDURES FOR FORMULATION


In this section,

we

shall introduce alternative

methods of formulation by

using variational principles such as the principle of stationary complementary energy

and the mixed principle [5,9-11]. Although hybrid principles

are found to be successful for a


in this

number of situations, they

will

not be covered

elementary text except for one example for Torsion in Chapter

1 1

Comment. For an undergraduate course, the instructor may wish


postpone coverage of the following material for a

later stage or for

to

a graduate

course.

COMPLEMENTARY ENERGY APPROACH


IT C is defined as the sum of the complementary
and the complementary potential of the external loads

The complementary energy


strain energy

U = U + Wpc
C

(3-55a)

Chapter 3

One- Dimensional Stress Deformation

Graphical interpretation of Uc

shown

is

in Fig. 3-4.

11

According to the principle

of stationary (minimum for an elastic body in equilibrium) complementary


energy,

we have

8U = 8U + 5Wpc =
C

or

= 5U - SW = 0,

<5n c

where

is

the complementary

because the potential


strains,

is

lost into

(3-55b)

work of external

loads.

sign occurs

and displacements,

n =
<

W MW dv ~ If PFWdS,
T

* JIf
V

where [D]

is

S2

is

the part of the

For the column problem, Eq.

[C]"

1
,

{u} is the vector

boundary on which

of prescribed

{u} are prescribed.

(3-55c) can specialize to

A p o A Cydy - p T vdy.

n =

express

them

(3-55d)

In the case of the complementary energy approach,

unknowns and

(3-55c)

st

the strain-stress matrix

displacements, and

as

The minus

work. In terms of the components of stresses,

in

terms of nodal

we assume the

stresses, or joint

stresses

or nodal

Hence, the approach is also called the stress or equilibrium method.


For matrix structural analysis, this procedure can specialize to the well-known
force method. Advanced study and applications of this approach are beyond
the scope of this text however, we shall illustrate the method by using the
rather elementary problem of the axial column.
We express stresses (<i) in the column element (Fig. 3.7) as
forces.

to
where [NJ

is

[N.]{Q n },

the matrix of interpolation functions and {Q}

nodal or joint forces related to a


the element [5,6,9, 12,

The

(3-56a)

statically

is

the vector of

determinate support system for

13].

surface tractions {T} can be expressed as


{T}

where [NJ

is

= [NJ{Q

(3-56b)

n },

a relevant matrix of interpolation functions.

Substitution of Eqs. (3-56a) and (3-56b) into II C [Eq. (3-55d)] gives

U =
c

T
i J]J {Q n } [KY[D]\N a ]{Q n }dV
V

J|

{QJTNJW&

(3-57)

St

Differentiation of II C with respect to {Q}

[f]{QJ

(q)

and equating to zero leads to


(3

58a )

One- Dimensional Stress Deformation

78

Chapter 3

where
[f ]

JJJ

[N.r[D][NJ<fo

(3-58b)

and
{q}

[f] is

r
JJtNJ {u}.

called the element flexibility matrix

and

(3-58c)

{q} is the vector

of prescribed

displacements.

The element flexibility equations can be assembled into global equations


and then solved for the unknown forces (stresses). This procedure can be
difficult and cumbersome. It is often convenient to transform the element
matrix into the element

flexibility

stiffness matrix.

Example 3-3
Consider the column problem (Fig.

and

stress

approaches are shown

3-7).

The

force systems used in the displacement

in Fig. 3-13. In the case of the latter, the flexibility

is formed by eliminating the degree of freedom corresponding to the rigidbody motion; it is done by constraining one of the ends while the other is loaded,
which yields a stable element.

matrix

Fo = F

'///////A v?

=0

777777777} v\

=0

fi=,-F

(a)

(b)

Figure 3-13 Force systems, (a) Displacement approach, (b) Stress

approach.

For

F acting at node

1,

v2

= 0, and for F acting at node 2, Vt =

0.

Under

these

circumstances the axial stress can be expressed as


C7 y

Then use of Eq.

= ^[lW

[S a ]{Q n }.

(3-58b) gives the flexibility matrix as

(3-59)

One- Dimensional Stress Deformation

Chapter 3

79

P.J_ dy

[f]

A E A

J_

(3-60)

AE
The equation

relating the reactions

is

given by
(3-61)

From

matrix structural analysis, the

can be transformed to the

flexibility relation

stiffness relation as [9, 12, 13]

[k]

[f]-

[G][fr

AE

which

is

same matrix

the

[fFtGF

[owr
-AEl

-AE

AE

m.
(3-62)

Eq. (3-28).

[k] as in

Comment. The complementary energy approach can

also be used in con-

junction with the concept of stress functions. This approach can be easier to

implement and understand


ever,

it is

we have

somewhat

in

comparison to the foregoing procedure. Howformulate with one-dimensional problems;

difficult to

illustrated its application for the

two-dimensional torsion (Chapter

11).

MIXED APPROACH
In the mixed approach, both the element displacements and stresses are

assumed to be unknowns. The formulation results in coupled sets of equawhich nodal displacements and stresses appear as unknowns.

tions in

Variational

Method

In the mixed method, the variational functional, often


Hellinger-Reissner principle,

nR =

JJJ

({f{e}

is

expressed as

known

as the

[5, 10]

- *WTD]M - mx})dv

JJ

iuYmdSi

Si

Here (w

{(u

JJ

- u)f{T}<tf 2

{P) r {q,}.

u) denotes the difference between the actual

placements along the boundary

(3-63)

St

S2

and {PJ and

{q,}

and prescribed

dis-

are the vectors of applied

nodal or joint loads and displacements, respectively.

One- Dimensional Stress Deformation

80

Chapter 3

The stationary value of IT* yields a set of equilibrium and stress-displacement equations. In these equations both the displacements and the stresses
simultaneously appear as primary unknowns.
Residual Methods

These methods can be used to formulate finite element equations based


on governing differential equations. For example, Galerkin's method can
be applied to the equilibrium and stress-displacement (stress-strain) equations, which results in coupled equations in displacements and stresses.
Example 3-4

We illustrate

column problem
assumed to be linear:

the mixed approach for the

the vertical displacement

is

= #1 - L)v +

=
Here [NJ

The

is

the

same

axial stress

where

[a

polation functions.

Variational

As in Eq.

+ L)v 2

(3-8d)

(3-8e)

[NJ{q}.

as [N] in Eq. (3-8d).

can also be assumed linear as

ay

r
{<T}

i(l

(Fig. 3-7).

a2

=
=

i(l

- L)a, +

i(l

L)o 2

[NJ{aJ,

the vector of nodal stresses

is

Note

(3-64)

that here

and [NJ

we have assumed [NJ

is

the matrix of inter-

[NJ

[N].

Method

Substitution of

v,

a y and
,

-*[

ey into the special form of Eq. (3-63) leads to

{<fn}

T [NY~[N){<f}dL

(TV,

+ P2 v 2

(3-65)

).

Here for convenience we have not included the body force, surface traction, and the
term related to difference in displacements. These terms can, however, be included
without

difficulty.

Now we differentiate 11* with

respect to

a u a2

({tf})

and v u v 2

({q})

and equate

the results to zero:

|^ = => y

[W\B)dL

{q}

[NFrNtfLto}

0,

(3-66a)

One- Dimensional Stress Deformation

Chapter 3

0=>y['

[BRN]d!L{<Fj

{0}

81

(3-66b)

[P,,},

or

+ [k TU ]{q] = 0,
IKuYiQn) + {0} = {Q},
[k TT ]{<T}

(3-66c)

or
{0}|

(3-67a)

LfCrTTo] "Jl{q}i

[QV

where

3E

AT
6E

A[

A[

6E

3J

A]_

[Nf[N]JL

[K

[k T J

[N] r [B]^L

(3-67b)

-/I

^4

(3-67c)

2J

and

in

(3-67d)

13

For further illustration, consider the following properties of the column divided
two elements (Fig. 3-14):

A =
Pi
/

=
=

E=
Figure 3-14

cm 2

10 kg,
10 cm,

1000 kg/cm 2

Column problem

for

mixed procedure.

a<

f
10

1)

v?

cm
(D
o%\
r

al

10

2,

(D
W
\i

()=* element

,vl

cm

V7b77>

2)
2)
o 2 ,v 2
{

v3 =

One- Dimensional Stress Deformation

82

Chapter 3

Substitution of these properties into Eqs. (3-67) gives the element equations as

Element

-10

-10

3000

6000

-10

-10

6000
____

3000
____

1"

<*2
>

<

or

--

(3-68)

10

Vl

Ox

Element 2

v2

The element equations can be assembled now. For convenience, in the following
we have rearranged the nodal unknowns as [c v a 2 v 2 a 3 v 3 ]; hence the assem!

bled equations are

10

10

3000

6000

_J_
2

10
- 6000

a,

__1_

10

10

6000

3000

2
1

(3-69)

v2

Here the boundary condition v 3

10

Vx

20
3000

10

6000

Ox

<7

[V 3 )

0/

has been introduced as described previously.

Solution of Eq. (3-69) leads to

= 10 kg/cm 2
a 2 = -10,
<7
= -10,
Ox

vx

cm,

V 2'-

cm,

v3

10

(prescribed).

These results are similar to those in Example 3-2. We may note that this is rather
an elementary problem used simply to illustrate the procedure. Also, the results
would have been different if we had used other orders of approximation and
different loading

and geometry.

Comment. The matrix in Eq. (3-69) contains zeros on the main diagonal.
For the standard Gaussian elimination, this can cause computational difficulties since the

Then

it

is

zeros can appear as "pivots" in the denominator (Appendix

necessary to use the idea of partial and complete pivoting

2).
[7],

which involves exchanges of rows and columns during the elimination procedure. Mathematically, the matrix [K] in Eq. (3-69)

is

positive semidefinite in

constrast to the positive-definite character of the stiffness matrix in the dis-

placement approach.

Galerkin's

With

Method

we used

the displacement formulation,

[Eq. (3-45b)]. It

was derived on the

the governing equation

basis of the equilibrium equation [Eq.

and the stress-strain or stress-displacement equation [Eq. (3-14a)].


use both the equilibrium and the stress-strain displacement equation
for formulating the mixed approach with Galerkin's method. The residuals
according to the two equations are
(3-45a)]

We can

* = * -

fy

(3-70a)

and

**

-% + *'

Assuming the approximation models


(3-64),

we

for v

(3 - ?0b)

and a y

as in Eqs. (3-8)

and

have, according to Galerkin's method,

^
4'( dy

T^N dy =

(3-71a)

and

g + )#,</, =

<).

(3-71 b)

After proper integration by parts, Eqs. (3-7 la) and (3-7 lb) will lead to
the

same

results as in Eq. (3-67), obtained

by using the variational procedure.

COMMENT
we have presented formulaby variational and residual methods. In the variational methods, the
attention has been given to the displacement, complementary, and mixed
In this chapter and in Chapters 7 and 11

tions

procedures, while Galerkin's procedure has been used mainly in the residual

methods.

number of

useful hybrid procedures [11] are available in the

variational methods; Gallagher

As

procedures.

[9]

has presented examples of some of these

described in Appendix

1,

other residual methods are also

possible.

In this elementary text,

it

will

be

difficult to

cover the total range of the

methods of formulation. Hence, we have chosen only a few of them, particularly those which can be illustrated easily with simple problems.

BOUNDS
discussed rather qualitatively that different approaches yield
different bounds to the exact solution. In the case of the variational procedures, if the physical and mathematical requirements are fulfilled, the poten-

In Chapter

we

83

One- Dimensional Stress Deformation

84

tial

Chapter 3

(displacement) and complementary (stress) energy approaches yield,

respectively, lower

and upper bounds to the exact solution for displacement.

The approximate

(algebraic) value of potential energy with a given


approximation function is higher than the exact or minimum value (Fig.
3-6). That is, the approximate stiffness is higher than its exact value. Consequently, the approximate displacements u, are lower than the exact displacements. This is depicted in Fig. 3-1 5(a), which shows the convergence behavior

mesh is refined.
The physical explanation is that the displacement approach yields a stiffer
or stronger structure than what it really is, because the assumed displacement

as the

functions, although continuous within elements, provide an approximate


(global) distribution of displacements. Indirectly, this introduces additional

supports or constraints in the structure and makes

it stiffer.

In the case of the complementary energy approach, the approximate value

of complementary energy
flexibility

is

higher than the exact or

minimum

value.

The

value from a numerical approximation has higher value than

the corresponding exact value

that

is,

the stiffness has lower value than the

Consequently, the approximate displacements are higher

exact stiffness.

than the exact displacements, Fig. 3-1 5(b).

The

approach yields a discontinuous distribution of displacement.


may be construed to introduce gaps and overlaps in the
structure. This makes the structure weaker or less stiff than what it really is.
stress

Physically, this

Figure 3-15 Symbolic representation of bounds in


analysis, (a)
stress

Bounds

in

finite

element

displacement approach, (b) Bounds in

approach.

[K]

Number

Number

of nodes

of nodes

(a)

u => displacements

[f]

Number

Number

of nodes

(b)

of nodes

One- Dimensional Stress Deformation

Chapter 3

85

Thus the two approaches bound the exact solution from below and from
[Figs. 3-1 5(a) and (b)]. This property is useful, and often we can bound

above

the exact solution by using both approaches.

Comment. Although the mixed and hybrid approaches can

yield satis-

factory and often improved solutions, they do not possess bounds.

ADVANTAGES OF THE FINITE ELEMENT METHOD


At this stage, it may be
ment method.

useful to

some of the advantages of the

list

finite ele-

Since the properties of each element are evaluated separately, an obvious

advantage
element.
is

no

is

that

we can

incorporate different material properties for each

Thus almost any degree of nonhomogeneity can be

medium hence

restriction as to the shape of the

geometries cause no

difficulty.

Any

included. There

arbitrary

and

irregular

type of boundary conditions can also be

accommodated

easily; this can be an important factor.


have not yet discussed other factors such as nonlinear behavior and
dynamic effects. Subsequently, and during study beyond this book, the reader
will find that the method can easily handle factors such as nonlinearities,

We

and time dependence.


method that makes it so appealing and
powerful for solution of problems in a wide range of disciplines in engineering
and mathematical physics. This generality allows application of formulations and codes developed for one class of problems to other classes of
problems with no or minor modifications. For instance, one-dimensional
stress deformation (Chapter 3), flow (Chapter 4), and time-dependent flow of
heat or fluids (Chapter 5) can be solved by using the same code (Chapter 6)
with minor modifications. Similarly, formulation and code for such field
problems as torsion, heat flow, fluid flow, and electrical and magnetic potentials (Chapters 11 and 12) have almost identical bases of formulations and
aribitrary loading conditions,
Finally,

it is

the generality of the

codes.

PROBLEMS
3-1.

L and interpolation functions N^ and 2 if the local


coordinates were measured from the quarter point. See Fig. 3-16.

Derive local coordinate

Figure 3-16


-*~

1-

r*7/4-H

One- Dimensional Stress Deformation

86

(a) For L = (v - y )/(l/4),L(l) =


-v
N
L),
2 = i(l + L). (b) For L = (y
i(3
1, and Ni = |(1 - L), N2 = ^(1 + 3L).

Solution:

3-2.

-1, L(2)

Derive element equations

if

Chapter 3

the value of the elastic

L(l)

3 )/(3//4),

modulus

N =

and
t
-$, L(2)

3,

E and the area A

vary linearly as

E=

[{{\

-L)

JO +L)}<

and

A=[\{\ -L)
where

A and A 2 and

j and

^(1

+^)]{^}'

are the values at nodes

and

2, respectively.

Answer:
[k]

where the scalar X

_-l

"!]

given by

is

~0

6/

Mi

A,

E2

-i
3-3.

Consider the column


/

of each element

in Fig. 3-17

J" 'A,

10^0
0^01
1

0_

E,
{

A2

E2

with linearly varying E, A, and

Ty

Length

10 cm. Derive element equations, assemble them, intro-

duce boundary condition v A = 0, and solve for displacements,


of displacements and stresses.

strains,

stresses. Plot variation

Figure 3-17
2
E, = 1000 kg/cm
2
_Aj =0.5 cm
T y1 = 0.5 kg/cm

E 3 = 1500
Ao = 0.5

2000

and

Chapter 3

One- Dimensional Stress Deformation

87

Partial solution:

Equations for element 2:

soor

-1
Assemblage equations:

0~

-7
15

-8

-8

16

(V

0.20

0.55

<

-8

-8
i'i

3-4.

0.29732 cm, v 2

8_

0.26875 cm, v 3

0.65

*>3

.V 4

0.40,

0.17500 cm, and v 4

0.

fy to vary linearly from zero at


and compute displacement and stresses.

In Example 3-1, consider surface loading


the top to 2

kg/cm 2

at the base,

Hint: Consider

fy =
Then the load vector

-L)

H(l

|(1

+L)]{J")-

is

af + fy
-*K."+ *3
/

3-5.

For Example
and stresses.

3-1,

3-6.

Use the code


and 3-2.

in

3-7.

For Example

3-1,

stresses as the

number of nodes

Consider

yi

assume that the base

settles

by

0.1

cm. Find displacements

Chapter 6 and obtain computer solutions for Examples

4, 8, 16,

3-1

study convergence of solutions for displacements and


are increased using the code in Chapter

and 32 nodes; plot

results,

6.

and comment on improvement

in accuracy, if any.

Partial solution:

See Fig. 3-18. Note that in view of the linear model and plane
is not significantly influenced by the

axial problem, the displacement solution

refinement of mesh.
3-8.

Sketch the variations of

TV,-

in

Eq. (3-40) for the element in Fig. 3-10, and


A and E to be constant.

derive the element stiffness matrix assuming


Solution:

-16

14

-16

16

-16

32.

14
[k]

3-9.

Derive element

= AE

stiffness for

"ST

quadratic displacement (Prob. 3-8)

varies linearly as
'A,

A=[#l -L)

i(l

+^){ //}

if

the area

One- Dimensional Stress Deformation

88

3-10.

By

Chapter 3

using the data of Example 3-1, solve for the displacements and stresses in

Prob. 3-8.
3-11.

Derive interpolation functions

3-12.

TV,-, i

model when the intermediate node

is

1, 2, 3,

for a quadratic approximation

situated at one-third the distance

or right-hand node.

the

left-

(a)

Derive load vector {Q} due to surface loading


T,

and quadratic v

in

i(l

- L)f +
yi

i(l

fy

given as

+ L)fyt

Eq. (3-40).

Figure 3-18 Results with

mesh refinement and convergence,

(a)

Distribution of displacement v for eight elements, (b) Distribution

of stress ay for eight elements,


(d)

Convergence of

(c)

Convergence of displacements,

stress at midsection.

from

One- Dimensional Stress Deformation

Chapter 3

89

At top

^^

u.y

0.6

At midsection^
0.5

0.4

0.3

0.2

0.1

Number

12

16

12

16

of elements

(0

40.0

CM

30.0

20.0

10.0

1
8

Number

of elements

(d)

Figure 3-18 (Continued)

90

One- Dimensional Stress Deformation

Chapter 3

Solution:

iL{ {Q}

J "'

i(l

+ L)T]dL

the load vector due to gravity given by

Y=
Assume

L)T

+ 2Ty J

127V,

Compute

l)ki(l

(1-*)J

(b)

Dl

= -j\r UUL +
>

-L)F, +i(l +L)f2

i(l

linear variation for v.

Answer:

= Ali2Y + Y2]
l

[Q]
(c)

Find the load vector


TV

if

= iL(L -

!)?

^L(L

1)7V 2

(1

- L*)f

Answer:

Tv + 2TV

4TV

3-13.

27V,

+ 2fw + 16fJ

Idealize as closely as possible the irregularly shaped

solve for displacements

and

stresses

column

in Fig. 3-19;

by using DFT/C-1DFE.

L
E

=2000 kg/cm 2

Ty

= 2.0 kg/cm
= 10 kg

.0

kg/cm 3

y//////////////y

Figure 3-19

3-14.

Consider the coordinate system in Fig. 3-2(c) and derive the quadratic

inter-

polation function model [Eq. (3-40)] from the following generalized coordi-

nate model for v

a2x

a3 v2

[<J>]{a}.

One- Dimensional Stress Deformation

Chapter 3

91

Solution:

{a}

[A]-{qJ,

y\y*

yjyz

"

(y 2

[A]" 1

- y\)(yi - y\)
-O3 + y 2
- y\)(ys - yi)

(y 2

(y 2

(y 2

[N]

(y*

(y 3

y\)(yz

- y\)

(y 2

'

- y 2 )(y - y\)
~(y 2 + yi)
- y 2 )(y - yi)
3

-1

Sy 2 -

y\y 2

- yi)(y - y 2
yz + y\
- y\){y - y 2

- y\){yz - y 2 )

(^3

- ^2) (^3 - ^i)J

[][A]-

-y 2 )(y - y
-(y - y\)(y - ys)
(y - yi)(y - y 2
y\){yz y\) (y 2 - yi){y - y 2 (y - y 2 )(y - Ji)J]
Because y 3 - y = 7/2, y - y 2 = -7/2, y2 - y =
and (y - y )Kl/2) =
(y
{y 2

3)

L,

we

finally

7,

have

= [iL(L-l) iUL +

[N]

1)

-*].

REFERENCES
[1]

Zienkiewicz, O.
F.

C,

C,

Irons, B. M., Ergatoudis,

Dimensional Analysis,"
I,

[2]

[3]

and

J.,

Ahmad, S., and Scott,


Two- and Three-

"Isoparametric and Associated Element Families for


Bell, K., eds.),

in Finite

Element Methods in Stress Analysis (Holland,

Tech. Univ. of Norway, Trondhein, 1969.

Bazeley, G. P., Cheung, Y. K., Irons, B. ML, and Zienkiewicz, O. C,


"Triangular Elements in Plate Bending-Conforming and Nonconforming Solutions," in Proc. Conf. on Matrix Methods of Struct. Mechanics, Wright Patterson Air Force Base, Dayton, Ohio, 1965.

Strang, G. and Fix, G.J. An Analysis of the Finite Element Method, PrenticeEnglewood Cliffs, N.J., 1973.
,

Hall,
[4]

Bathe, K.
sis,

[5]

J.,

and Wilson, E. L., Numerical Methods


Englewood Cliffs, N.J., 1976.

Desai, C.

S.,

and Abel,

Nostrand Reinhold,
[6]

[7]

[9]

Element Analy-

J. F.,

New

Introduction to the Finite Element Method,

Van

York, 1972.

Turner, M. J., Clough, R. W., Martin, H. C, and Topp, L. C, "Stiffness


and Deflection Analysis of Complex Structures," /. Aeronaut. Sci., Vol. 23,
No. 9, Sept. 1956.
Fox,
Press,

[8]

in Finite

Prentice-Hall,

L.,

An

New

Introduction to Numerical Linear Algebra, Oxford University

York, 1965.

Prenter, P. M., Splines and Variational Methods, Wiley,

New

Gallagher, R. H., Finite Element Analysis Fundamentals,


Englewood Cliffs, N.J., 1975.

York, 1975.
Prentice-Hall,

One- Dimensional Stress Deformation

92

[10]

Washizu, K., Variational Methods

in Elasticity

and

Chapter 3

Plasticity,

Pergamon

Press, Elmsford, N.Y., 1968.


[11]

Pian, T. H. H., and Tong, P., "Finite Element Methods in

Mechanics,"

in

Advances

in

Continuum

Applied Mechanics, Vol. 12, Academic Press,

New

York, 1972.
[12]

Przemieniecki,

New
[13]

J. S.,

Theory of Matrix Structural Analysis, McGraw-Hill,

York, 1968.

Rowan, W. H., Hoadley, P. G., and Hackett, R. M.,


Computer Methods of Structural Analysis, Prentice-Hall, Englewood Cliffs,

Beaufait, F. W.,
N.J., 1960.

ONE-DIMENSIONAL FLOW

Flow of heat or

fluid

through solids

is

a problem that

is

frequently

encountered in engineering. In general, such a flow occurs in three spatial

we can assume the flow to occur


one dimension. Examples of these problems are heat flow
through uniform bars, fluid flow through pipes of uniform cross section,
vertical flow in a medium of large extent, and flow in and out of vertical
banks and cuts and toward vertical retaining structures.
As in Chapter 3, the flow domain (Fig. 4-1) can be idealized as a onedimensions. For some problems, however,
essentially in

dimensional line with the material properties attached to the

line.

The

governing differential equation for the one-dimensional steady-state flow can

be expressed as

/(*)

dx-

#x),

(4-1)

where k x is the material property, coefficient of permeability for flow through


porous media {LIT) or thermal conductivity for heat flow; <p* is the fluid head
(L) or potential for flow through porous media or temperature for heat flow
and/(x) = q(x) is the forcing function fluid flux (1/r), or heat flux, respectively, for the

two problems.

important and useful to note that Eq. (4-1) is similar to Eq. (3-45b)
that governs one-dimensional deformation in a column. This indicates that
It is

the

phenomenon of deformations

follow similar natural behavior.

in a

We

column and the one-dimensional flow

can even consider an interpretation that


93

Chapter 4

One- Dimensional Flow

94

(a)

(b)

(c)

Figure 4-1 Idealization for flow in pipe, (a) Flow through pipe,
(b)

One-dimensional idealization,

(c)

Discretization in three ele-

ments.

flow or rate of change of deformation due the (axial) load and the flow or rate
of change of potential or temperature due to applied fluid head or temperature are analogous.

Equation (4-1) governs steady or time-independent flow of heat and fluid


through one-dimensional media; in fact there are a number of other phenomena such as one-dimensional moisture migration that are also governed
by similar equations. The finite element formulation for Eq. (4-1) can therefore be used for
5,

we

all

these problems with only

minor modifications. In Chapter


by using the problems of heat

shall derive equations for transient flow

flow and consolidation, while in this chapter

we

use steady fluid flow through

media as the example problem. We may note, however, that the finite
element formulation for all these problems will essentially be the same except
for different relevance for material properties and meanings of the unknowns
such as temperature and fluid head.
For the fluid flow problem, then, (p is the fluid head or potential =
ply + z, where p = pressure, y = density of fluid, z = elevation, and
k x = coefficient of permeability in the x direction. The forcing function f(x)
can take the form of applied fluid flux intensity q(x). Now, we follow the steps
rigid

(Chapter 2) required in the

finite

element formulation.

Step

As

Chapter

in

Step

We

Choose Element Configuration

1.

we

3,

use the one-dimensional line element [Fig. 3-2(c)].

Choose Approximation Function

2.

choose a linear approximation model for

(p

fltj

cc 2

fluid potential as

(4-2a)

or
<p

where

=N

N = ^(1 L), N =

^-(1

+N

l(Pl

2 <p 2

+ L),

and

element nodal fluid heads. Equation (4-2)


here

q>

(4-2b)

[N]{<p],

is

{^}

[<p

p2

is

the vector of

similar to Eq. (3-7) except that

replaces v.

The requirements for choosing the approximation function for <p are similar
to those in Chapter 3. The linear approximation yields continuous variation of
<p within the element. For simple plane and linear flow, the interelement
compatibility

required only for the nodal values of (p. In other words, since

is

the highest order of derivation in the energy function, given subsequently in

Eq.

only up to order
The completeness requirement is
allows for rigid body motion and

(4-6), is 1, the interelement compatibility is required

equal to

= 0, that

is,

for nodal heads.

satisfied since the linear function,

Eq.

for constant state of gradient

Eq. (4-3a).

Step

Define Gradient-Potential Relation

3.

and Constitutive

The

gx

(4-2),

relation

Law

between the gradient of

(p

and

analogous to the definition of strain e y in Eq.

(p

can be defined in a manner

(3- 12a);

hence

.-
where gx

The

is

the gradient of

(a)

with respect to x.

(p

relevant constitutive law describes the flow behavior through porous

media. The simplest constitutive law that

we can

use

is

Darcy's law (Fig.

4-2),

given by
vx

where v x

= velocity

= -k x d = -k xgx

in the

direction

(4-3b)

and gx

= hydraulic

gradient.

The

negative sign occurs because the velocity decreases with gradient (Fig. 4-2).

The value of

the gradient can be obtained by differentiating

q>

in

Eq. (4-2)

with respect to x as

3?

= fx

? ({

" L)<Pl + ^ (1 +

L)<Pl]

'

(4_4a)

95

One- Dimensional Flow

96

9x

Chapter 4

3y?

3^

Figure 4-2 Relation between velocity and gradient, Darcy's law.

Use of the chain

rule of differentiation leads to

= 4-1-1

dx
I

As

before [Eq. (3-13)] [B]

i]

*
\ fl

[B]{<p}

is

the gradient-potential transformation matrix.

(4-4b)

Substitution of Eq. (4-4b) into Eq. (4-3) gives

-[RPKfJ,
where [R]
case

it is

= matrix

of coefficients of permeability

simply the scalar k x

It is

(4-5)

for the one-dimensional

interesting to note the similarity

between

Eqs. (4-4b) and (3-13) and between Eqs. (4-5) and (3-15); the gradient and
velocity terms here are, respectively, analogous to the strain
in the stress-deformation

and

stress

terms

problem.

Derive Element Equations

Step

4.

The

finite

element equations can be derived by using either a variational


For problems with certain mathematical

principle or a residual procedure.

properties such as self-adjointness, valid variational principles are available.

For certain other problems,

it

may

not be possible to establish a mathemat-

ically consistent variational principle.

Residual methods such as Galerkin's

procedure can be used to derive

element equations for such problems

finite

methods can be used as general procedures.


For the problem governed by Eq. (4-1) we shall first use a variational

in fact, the residual

principle. Later in this chapter,

method.

we

shall illustrate the use

of Galerkin's

One- Dimensional Flow

Chapter 4

The

variational principle corresponding to Eq. (4-1) can be written as

r^k

Q=A
where

97

is

(^fdx - pfydx,

the cross-sectional area, assumed to be uniform,

(4-6)

and q

is

the

prescribed fluid flux; here units of q are (L 2 /T).

In Eq. (4-6)

we used

the symbol

to denote a

system through which the flow occurs; this term

is

measure of energy in the


analogous to the energy

measure IT^ for the stress-deformation problem. It is only to distinguish this


measure for different problems that we have used different symbols in fact,
we can use any convenient symbol. Note the analogy between the strain
energy density term \o y e y in Eq. (3-21 a) and the term ^v x gx in Eq. (4-6) and
between the potential of the external load fyv in Eq. (3-2 la) and q<p in Eq.
;

(4-6).

Substitution of d<p/dx and

H=A

into Eq. (4-6) leads to

q>

[" #<p n } T [WkM{<p n }dx

Differentiation of

with respect to
f

<5n

= o=

<p

I""

and

^=

<p t

m]{<p n }dx.

(4-7)

yields

0,

(4-8)

1^ = 0,
or

f" [BYk x [B]dx{<p n }

P q[Wdx,

(4-9a)

or

M{<P}

where

[k]

{Q}.

(4-9b)

= element property (permeability) matrix,


[k]

=A

("
>

[WkJWx

XI

i-'J'

and {Q}

{WkAWL,

(4-9c)

= element nodal forcing parameter vector


{Q}= J{"qlWdx
Xl

h;

q[N] T dL.

(4-9d)

Evaluation of

[k]

and (Q

Substitution for [B] and [N] in Eqs. (4-9c) and (4-9d) and integrations
lead to

Ak.

[BY[B]dL

[k]

Ak

r[-

w-D

qi

{Q}

2 J_,

Notice that Eq. (4-10)

is

(4-10)

dL

li(l+L)J~

ql

(4-11)

[i

similar to Eq. (3-28) except for the fact that

has replaced E. The lumping of half the applied flux at each node

is

kx

the result

of the linear approximation model, and for higher-order models, the forcing
vector can be different.

The element equations can thus be expressed


Ak.
I

Step

As

as

ql/2

(4-12)

-1

ql/2

<Pi

Assemble

5.

explained in Chapter

3,

we can

differentiate the total of

elements and obtain the assemblage equations. The procedure


similar to the direct stiffness

ment

that the fluid heads at a

is

for all

essentially

approach and is based on the physical requirecommon node between two elements are equal.

is analogous to the interelement compatibility for displacements in


Chapter 3. Then for a domain discretized in three elements (Fig. 4-1) with
uniform area of cross section, we have

This

Ak x

-1

-1

'

[<P\

-1

-1

-1

\<P1

-1
1

>=

ql

(4- 13a)

2'

93

1
L

or

[K]{r}={R],

where

[K]

(4-13b)

assemblage property (permeability) matrix,

vector of nodal heads, and {R}

= assemblage

{r}

= assemblage

vector of nodal forcing

parameters.
Step

6.

Solve for Potentials

Solution of Eq. (4-13) after introduction of the prescribed boundary


conditions allows computations of nodal heads.
the following properties

98

To

illustrate this step,

assume

Example

4-1

=
=

A
kx
q
/

Boundary

cm 2

1.0

cm/sec,

cm 2 /sec,

0.0

10 cm.

conditions:

2.0 cm,
1.0

We

cm.

use Eq. (4-12) to find the element equations. For instance, for element

number

we have
1

icm:

10

and so on for the other elements. The assembled equations are


0"

-1

-1

-1

9i
cp 2

>

-1
The boundary conditions
set

1 1

k 44

=
=

<

<?3

<Pt

are introduced as follows


1

and

ku

=0,

2, 3, 4,

and

k Ai

=0,

1, 2, 3,

r =2,

R<

1.

Therefore

0~

<Pi

_2_

(Pi

10
<

>

TO

the

first

equation

>

<

<Pi

10

1_

From

we have

q>

2.

From
9>3

10

"^

the second equation,


_

10" =

"

10

<P*

or
-<Pi

but

(p

+2<Pi

- "?3

2; therefore
2<p 2

From
(p A

(p 3

2.

(i)

the third equation

-tp 2
but

- = 0,

2(p 2 - -<?4

Therefore,

-<Pi

2(Pz

1.

(ii)

99

One- Dimensional Flow

100

Equations

(i)

and

(ii)

Chapter 4

lead to

-7(p 2

4^3
3(p

Hence

and
2<p 2

-f =

2.

Hence

=f-

<Pi

Step

7.

Secondary Quantities

The secondary quantities for the fluid flow problem are the velocities and
The knowledge of <p from Step 6 now permits evaluation of

quantity of flow.
these quantities.
Velocities.

Element

=
Similarly for elements 2

and

0.0333 cm/sec.

"- =

10

-' "III

\3 cm/sec.
0.0333

^,

= -^[-1
=

Quantity of flow

I?

is

0.0333 cm/sec.

Qf = v x A =
where

i]{*

the velocity

0.0333

= 0.0333 cc/sec,

normal to the cross-sectional area A, and the subscript


1 and so on.

in v xl denotes element

Step

8.

Interpret and Plot Results

The computed values of cp and

v x are plotted in Fig. 4-3.

We can

see that

the finite element computations give exact solutions both for fluid heads and
velocities.

This

is

because the cross section and material properties are

One- Dimensional Flow

Chapter 4

101

cm

1 0.0333 cm/sec

(b)

Figure 4-3 Plots of

finite

element computations,

(a)

Fluid heads,

(b) Velocity.

uniform, and constitutive (Darcy's) law

is

linear.

nonuniform cross section and nonlinearity can


for (p and v.

FORMULATION BY GALERKIN'S METHOD


The residual corresponding to Eq. (4-1)
Rif)

k x %

As

indicated in Chapter

3,

yield nonlinear distributions

is

- q.

(4-14)

The details of derivation are almost identical to those for the column problem
in Chapter 3. The final results for the three-element system (Fig. 4-1) will be
>

-1

-1

dx).
\<Pl

-1

<Pl

|(

Lid
2

2
>

-1

1
\

[K]{r]

= {R r +
}

{RJ

(4-1 5a)

<

= {R}.

(*%1

(4-15b)

The boundary terms {R } include flux, k x (d<p/dx), or Neumann-type boundary


condition at end nodes 1 and 4. They are prescribed by given values of the
(normal) derivative d(p/dx. If the boundary is impervious, d<p/dx = 0. For
the node with geometric or Dirichlet boundary conditions, that is, prescribed
cp, the term is removed while introducing the condition in the assemblage
equations.

PROBLEMS
4-1.

Solve Example 4-1 by considering a nonzero value of q.

4-2.

Use code DFT/C-1DFE and

4-3.

Use code DFT/C-1DFE and compute the


30

cm

verify the results in

Example

4-1.

steady-state temperature in a bar

long with the following properties,

= 1.0,
= 0,

kx
<7

/=
and temperatures

10 cm,

at the ends,

T(0)

7(30 cm)

=
=

100 degrees,
degrees.

BIBLIOGRAPHY
Desai, C.

S.,

"An Approximate Solution


ASCE, Vol. 99, No. IR1,

Drain Div.,
Desai, C.

S.,

ments

wicz, O.

Desai, C.

"Finite Element

in Fluids,

S.,

C,

Vol.

eds.),

"Seepage

in

for

Flow

in

(Gallagher, R. H., Oden,

Wiley,

Engineering (Desai, C.

Methods

for Unconfined Seepage," /. Irrigation

1973.

New

Porous Media,"
J.

York, 1975, Chap.

T., Taylor,

in Finite Ele-

C, and

Zienkie-

8.

Porous Media," in Numerical Methods in Geotechnical


S., and Christian, J. T., eds.), McGraw-Hill, New York,

1977, Chap. 14.

Pinder, G.

F.,

and Gray, W. G.,

Finite

face Hydrology, Academic Press,

Remson,

I.,

Hornberger, G. M., and Molz,

face Hydrology, Wiley-Interscience,

102

Element Simulation
York, 1977.

in

Surface and Subsur-

New

New

F. J., Numerical Methods


York, 1971.

in

Subsur-

5
ONE-DIMENSIONAL
TIME-DEPENDENT FLOW
Introduction to Uncoupled

and Coupled Problems

In this chapter

we

treat

the case
is

when

in which temperature or fluid pressures


These effects can occur in two ways. For

problems

act in addition to external loading.

the magnitudes of temperature or fluid pressures are

relatively easier to include

them

in the finite

known,

it

element formulations because

Some examples
known temperature distribution in a structure and known fluid
pressure in a porous body. The general case occurs when temperature is
unknown just like the displacement. Then we need to consider interaction or
the effects can be superimposed o r considered as uncoupled.

of these are

coupling between deformation and thermal

effects.

UNCOUPLED CASE
As an

illustration, let us

perature Tis to cause a

assume that the

known

,.

where

a'

f*

The

stress-strain relation

ay

is

known change

column

in tem-

[Fig. 5- 1(a)],

= *'T,

(5-1)

= coefficient of thermal expansion.

e y the effective elastic strain e yn

of a

effect

strain e yo in the

If the total strain

is

denoted by

given by

= *, -

(5-2)

can then be written as

= Ee = E(e, yn

e yo ).

(5-3)

103

One- Dimensional Time- Dependent Flow

104

Chapter 5

EA

(a)

1000 kg

10cm
Area = 1.0 cm 2

a'
v,

= coefficient of thermal expansion


= 0.0000065 cm/cm-0F

=0

10cm
E =

10

2x 10 6 kg/cm 2

cm

(b)

Figure 5-1 Thermal effects in a bar idealized as one-dimensional,


(a)

and

Bar and one-dimensional


properties.

idealization, (b) Finite element

mesh

One- Dimensional Time- Dependent Flow

Chapter 5

For deriving

Tl p

= A-

element equations,

finite

energy function [Eq.


f"

a,dy

r Yvdy - P T vdy - Puv

related to the external forces are the

same

as before.

term related to the strain energy can be written by using Eq.

U'

(5- 4)

Jyi

Jyi

Note that the terms

modify the potential

to

(3-21)], as

Jyi

first

we need

105

The

(5-3) as

= \" E(e y -(J(e y -eJdy


Jyi

A C y2

=4

=U +U
1

can drop the

last

Ee y yo dy

A C

Jyi

Jyi

We

yz

Ee)-A\

y2

Ectfy.
Jy

(5-5a)

term of Eq.

contribute to element equations

(5-5a), since,

when

Up

is

being a constant,

differentiated.

the problem in Chapter 3 [Eq. (3-21)], the only

it

will

not

Thus, compared to

new term

that enters

is

the

second term in Eq. (5-5a), which after substitution for e y from Eq. (3-13) can

be expanded as

U2 =

Ap[Vl

v2

]^-^Ee

yo

dy

A{qf

J"

T
[B] [C]{eyo }dy

Jyi

-\L)\
aU-Vi

(v 2 L)

= AEe (-v + v
U2 is the part of the strain energy relevant to e
yo

Here
same as
loads.

(5-5b)

2 ).

U in Eq. (3-21) and corresponds to the


Now differentiation of 11^ with respect to v

yo

the other part

strain
x

is

the

due to the external

and v 2 gives

Mi =

AEey X-l),

(5-6a)

^=

AEeJY),

(5-6b)

or in matrix notation,

=AOy
II

!*

One- Dimensional Time- Dependent Flow

106

Chapter 5

In general terms,

{Q
Finally, if

=A

f" [BY[C]{e yo ] dy.

(5-7b)

j yi

we add {Q

in Eq. (5-7) to Eq. (3-28), the modified element

equations are
[k]{q}

{Q}

{Q

(5-8)

},

where all terms except {Q } have the same meanings as before. We call {Q }
an additional, extra, correction, "initial," or residual load vector.
Thus the known strains due to temperature changes become an additional
load which is superimposed on the external load {Q}. It appears on the
right-hand side of Eq. (5-8) and does not contribute to the left-hand side,
which contains the unknown displacements.
Initial Stress

As can be
can act as

seen from Eq. (5-7), the

"initial" stress or pressure.

known temperature or fluid pressure


instance, we may know from the

For

solutions of the Laplace equation (Chapter 12) the distribution of seepage

dam; such

pressures below a

pressures can be converted as additional load

vectors by using Eq. (5-7).

Residual Stresses

The concept of "initial"


term "residual,"

is

strain or stress,

which can be

called

by a

common

important and useful for incorporation of residual

conditions caused by temperature, fluid pressure, creep, lack of

phenomena. This concept has

also been used widely in a

fit,

and other

number of techniques

for nonlinear analysis.

Example

5-1

We now illustrate the concept


tion.

The

details are

shown

tions to include the effect of a

solve for displacements

and

of

load vector by using the foregoing formula-

initial

in Fig. 5-1 (b). It

is

required to derive the element equa-

in temperature of 1 00F and then to


combined (uncoupled) superimposed thermal

change (increase)

stresses for

and external load = 1000 kg at the top of the bar.


According to Eq. (5-1), the initial strain is

effects

= aT =

yo

Hence, the resulting

0.0000065 x 100

= 0.00065 cm/cm.

stress [Eq. (5-3)] is

Eyo

2 x 10 6 x 0.00065

1300 kg/cm 2

One- Dimensional Time- Dependent Flow

Chapter 5

and the

initial

load [Eq. (5-7)]

is

{Q
The assembled

initial

x 1300

{~ J}-

load vector {R} for the three elements

-1
1

{Rol

13(
1

is

-1300
1

<

1300,

By

107

following the procedure in Chapter 3 and by adding {R

vector,

to the external load

we obtain
1

2 x 10 x

-1
2

-1

-1

0"

\v 2
>

<

10

-1

-1

-1300

\v\

<

^3

.-1000,

1_ [v*
.

1300.

-1300

300

Note that here (Fig. 5-1) we have measured y positive in the upward direction.
The boundary condition v = is now introduced, and the solution of the resultx

ing equations yield


v\

v2

=0.0015,

v3
V4

The

total strains

due to the

0.000

cm

(specified),

= 0.0030,
= 0.0045.
effects

of both the external loads and thermal effects

are found as

[B]{q}=y[-1

yi
y2
yt

=
=
=

C)

+0.00015, cm/cm,
+0.00015,
+0.00015.

These strains are equal to the sum of the strain due to the external load, ey
-0.00050, and that due to the thermal effect, yo = 0.00065.

COMMENT
This example

concept of

initial

is

included mainly as an elementary illustration of the

load vector.

It

does not necessarily represent a practical

One- Dimensional Time- Dependent Flow

108

Chapter 5

Moreover, unless the ends are constrained, the temperature


not influence the stresses in the bar.

situation.

may

change

TIME-DEPENDENT PROBLEMS
As

illustrations we shall consider two problems governed essentially by the


same equation. They are time-dependent temperature effects [1] and timedependent deformations with expulsion of fluid from the pores (of a homogeneous medium); in geomechanics, the latter is called consolidation [2].
Both problems can be represented essentially by the following second-

order partial differential equation

called a parabolic equation:

[1],

dT*

d 2 T*

where T* = temperature (or fluid pressure) and a = thermal diffusivity =


k/pc, where k = thermal conductivity (Btu/hr-ft-0F or W/m K), p = density
(lbm/ft 3 or g/cm 3 ), c = specific heat (Btu/lbm-F or J/kg.K), x = space
coordinate, and t = time (Fig. 5-2). It is again interesting and useful to note
that Eq. (5-9a) is similar to Eqs. (3-45b) and (4-1) that govern deformations
in a column and flow in a medium, respectively. The difference is that now we
have the extra term on the right-hand side to account for the time dependence
of the phenomenon.
It is often more convenient and necessary, particularly for problems
involving nonhomogeneities and layering, to express Eq. (5-9a) as

/9 2

k
for the thermal

problem

and

[1]

At*

T*

-6^ =

ec

(5 " 9b >

-W

as

*L*g = m,*!
dx 2

yw
for the consolidation

problem

[2, 3, 4];

(5-9c)
v

dt

'

here/?*

= (excess)

pore water pres-

the coefficient of permeability (ft/sec or cm/sec), y w is the unit


coefficient of volume comweight of water (lb/ft 3 or g/cm 3 ), and
v
sure,

is

m =

pressibility (ft 2 /lb or

cm

/g).

The boundary conditions

associated with Eq. (5-9) can be expressed as

initial conditions,
(i)

and boundary

(iii)

We

is

= f *(*),

< x < h, < 0,


t

(5-10)

conditions,
(ii)

where h

T*(x, 0)

T*(0,t)
T*(h,t)

= f*(t),
= f (h),
h

t>0,

(5-1 la)

/>0,

(5-1 lb)

the total length of the domain.

note that for the time-dependent problem extra boundary conditions

occur as

initial

conditions [Eq. (5-10)] to define the

initial

or starting condi-

One- Dimensional Time- Dependent Flow

Chapter 5

Element

109

(a)

t-At

+ At

(b)

Figure 5-2 Solution in time domain, (a)

Time

variation of tem-

perature in a typical element, (b) Finite difference approximation


for

first

derivative.

tions of the body. In the case of static or time-independent problems in

Chapters 3 and 4 such

We

initial

conditions are not required.

shall first follow the steps for the

thermal problem and then

illustrate

the case of consolidation.


Step

1.

Discretize and Choose Element Configuration

The one-dimensional medium


Step

2.

is

divided into line elements, as in Fig. 3-2.

Choose Approximation Model

We choose a linear model to express temperature within an element as


r = i(i-L)r (0 + i(i-f-x)r2 (0
1

= [N(x)]{T },
n

(5-12)

One- Dimensional Time- Dependent Flow

10

where

{T}

= [T T

Here

the vector of unknown temperatures at nodes

2 ],

respectively.

Chapter 5

and

T2

are time dependent,

and [N]

and 2,

dependent on the

is

spatial coordinate x. As stated in Chapter 4, this function satisfies the requirements for approximation function for T for the one-dimensional flow.

Step

Define Gradient-Temperature and Constitutive Relation

3.

The following
heat flow and

is

rate equation can be

q'

where

q'

= rate

or

Step

(5-13)

W/m-K), and A

= area normal

to

= thermal
x

direction

).

4.

First

= -kA 6

of heat flow in the x direction (Btu/hr or W), k

conductivity (Btu/hr-ft-F or
(ft

assumed to describe the mechanism of

used as the constitutive relation:

Derive Element Equations

we

consider the transient heat flow problem. Use the following

variational functional at a given time level for obtaining the element equations

[5, 6]

['iHy + % T i dx - \j Tdx

(5 - M)
-

where q = applied heat or fluid flux and CI denotes the variational functional.
We have here used the symbol Q to distinguish from the n p used for potential
energy in the stress-deformation problems.

By

using Eq. (5-12),

we have

g = ^[i(i-i-)r
=

+i(i+L)r2

(5-15a)

[B]{T}

and

J= t

= [#! -

L)T,

+ ^(1 + L)T2

dT,
dt

[N]

\dT2

"fcl

dt

[N]{f} ;

(5-15b)

One- Dimensional Time- Dependent Flow

Chapter 5

111

where {T} r = \t f 2 ] and the overdot denotes derivative with respect to time.
Note that [N] is function of space coordinates and hence constant for the time
derivative. Substitutions from Eqs. (5-15a) and (5-15b) into Eq. (5-14) yield
x

A_l_
2 2

FY[BY<x[B]{T n }dL

+ ^^{t} T lWmTn}dL
(5-16)

Expansion of Eq. (5-16) leads to (assuming a to be constant)

(5-17a)

-1

-1

</L

V-4 2.

#1#2
</L

#i# 2

--f J"

(JV.7,

+N T )dL.
2

(5- 17b)

Further expansion of the matrix terms leads to

2T T2
x

(N\t,T,

(N T
X

We now

T\)dL

+ N,N

+N

t,T2

+ N N t T + N\t T )dL
x

T2 )dL.

(5-18)

take differentiation (variation) of

important difference between

this variation

with respect to 7\ and

and the variation

in

T2 An
.

Chapters 3

and 4 may be noted. Here, the functional involves time derivatives, f and
f 2 During the variations with respect to T and T2 we make an assumption
that t t and t 2 remain constant. This assumption makes the variational
x

One- Dimensional Time- Dependent Flow

112

approach mathematically
dQ.

Ace

Wt-TxlF'j:
+ 4[j
dQ

_
~ 2l

At*.

dT,

differentiation yields

2T >~ 2TJ dL

+ NyNjjdL - St

(Nit,

dL

= 0,

(5-19a)

J"

(-2T

The

less rigorous.

Chapter 5

+ 2T2 )dL

+ 4[^(N N2 T +Nlt
1

)dL

= 0.

,dL

-#J>

(5-19b)

Therefore,

dQ
dT

Aa

f,

2/ ,

- T )dL

(7\

+
dQ
dT2

^J

^a

+ N N t - lf.(l) =

(tfff ,

ti>

r,

+r

2 )</L

^f*

(5- 19c)

0,

2)

QTtNjt +
l

ATif 2 )

(5-19d)
Integration of the terms gives

^= A - T
g = ^(-r, +
-5j-(7\

2)

%(2t,

+1 2)

r2 ) + f(t, + 2f 2 ) -

Rearranging

^a
/

in

"
1

|-

= 0,

f=

(5-20a)

o.

(5-20b)

matrix notation leads to

-r

_-l

i_

"2

1"

kr-6- .1

2_

f^il

^/

ft

-* r

>l

7-

_W

_?'
2

(5-2 la)
1

or

LKKTJ

[kjffj

= {Q(r)}.

(5-2 lb)

Layered Media
In the case of layered media, the formulation based on Eq. (5-9b) or
(5-9c) will lead to the following element equations

One- Dimensional Time- Dependent Flow

Chapter 5

-i

r2

a,
1 2

a-fo

113

(5-21c)

where A, = ^4/:// and A 2 = Apcl/6 or ^j = Ak/yJ and A 2 = AmJ/6, corresponding to Eq. (5-9b) for the thermal problem and Eq. (5-9c) for the
consolidation problem. In the case of the latter, temperature is replaced by
pore water pressure. Although the same symbol q is used for prescribed flux
in Eqs. (5-21a) and (5-21c), their units will be different and should be defined
carefully. Here [k a ] = element thermal diffusivity matrix, [kj = element
matrix related to time dependence, and {Q(0} = element nodal vector of
forcing (flux) parameters, which can be time dependent. The matrices in
Eq. (5-21) can be expressed as
[K]

A P[Bfa[B]<fr,

(5-22a)

J XI

[k,]

{Q(t)}

=A
=

<j

\" [W[N]dx,

(5-22b)

I" [Wdx.

(5-22c)

J Xl

Solution in

Time

Additional derivations are required in the time-dependent

problems

because of the appearance of the second term on the left-hand side of Eq.
(5-21).

Up

using the

we have discretized the domain only in the spatial


need to obtain solutions in time. This can be done by

to Eq. (5-21)

direction, x.

Now we

finite difference [7]

respect to time.

The

first

type of discretization for the derivatives with

derivative dT/dt can be written approximately as

[Fig. 5-2(b)]

where At = time increment. Equation (5-23) essentially gives the slope of the
chord joining points A and B as an approximation to the continuous derivative dT/dt. By using Eq. (5-23), we can now write approximations to the time
derivative at the two nodes of an element [Fig. 5-2(b)]

^r.(' + ^)-r.(O
dT\

The time

We

_T +
2 (t

At)

-T

integration scheme [Eq. (5-23)]

2 (t)

is

(5.24a)

(524b)

the Euler-type procedure

[7].

use this simple scheme only for the sake of introduction. In fact, a

number of improved and mathematically superior schemes such as the


Crank-Nicholson procedure are commonly used in finite element applications.

One- Dimensional Time- Dependent Flow

114

Chapter 5

Substitution of Eq. (5-24) into Eq. (5-21) leads to

or

(w + iw)

;i;:3**^w

<->

or

>#:s-*
where

= ft.] +

[k]

{0}

{Q(r

^ft]
A/)}

+ [k,]{

At any time

/, the terms on the right-hand side in Eq. (5-26) are usually


known, {Q(/ + At)} from the specified forcing functions and the second term
from the known values of T at the previous time level. Because the initial
conditions [Eq. (5-10)] are given, we know the initial values of Tat any point

at time

= 0.

Hence, we can solve Eq. (5-26) for Tat

the previous time level are

The

At, since the

Tat

known.

solution process for the time-dependent problem thus involves two

and then (2) propagation or marching in time


The foregoing problem belongs to the class of problems
problems, since we start from an initially known state.

steps: (1) discretization in space

at various time levels.

called initial value

Step

4.

We now

Derivation by Galerkin's

Method

consider formulation of

finite

element equations [Eq. (5-21)] by

using Galerkin's residual method. Equation (5-9a) can be written as

tF-t?-*

(5 - 27a >

or

a
(

fe-|) r * = -

(5 - 27b)

or

LT*

= 0,

(5-27c)

where L is the differential operator. Denoting the approximate solution by


we have the residual R,
*<*>

= a 0-w-

T,

(5 " 28)

One- Dimensional Time- Dependent Flow

Chapter 5

On

the basis of the explanation in Chapter

Assuming

for the Galerkin formulation.

3,

linear

we

115

consider a generic element

approximation for an element

we have

as in Eq. (5-12),

r = i(i-L)r (0
I

+ i(i+L)r (0
2

= t NT

(5-29)

t.

Here the interpolation functions

peratures

are functions of x,

TV,

We

are the functions of time.

and the nodal tem-

note here that for a given time

t,

Eq. (5-29) gives variation of the temperature along the length of the element.

At

time dependence

this stage, the

only, 7^(0,

Now

T2 (t)

is

included in the temperatures at the nodes

[Fig. 5-2(a)].

according to Galerkin's method,


(5-30a)

Integration by parts of the

3 dx
a3
** &<

first

term leads to

NA

a 3-

dx

'L

-f

-3dt

N dx =

(5-30b)

or
x

\dTdN

dT M
N dx

dT

= *<Z N
AT

"dx-lti^

+L

T=

into Eq. (5-30c) leads to

I
Substitution of

X2
,

NT
t

<

(5-30c)

<

l,2,y=l,2.
(5-3 la)

The index notation has an implication


In matrix notation, we have
dNj
dx
J Xl

dN1 dN2

dl^^dNi

dx

dx

dx

dN2 dN2

dx

dx

3.

dT,

<Mj.'

dx

similar to that explained in Chapter

Ty

dx

dx

dt

N,N2

>dx

dT2

N
{

dt

\..dT

dx
(5-3 lb)

dT
dx-

The two terms on the left-hand


[k,]

as in Eq. (5-21).

side yield the

same element equations [kj and

One- Dimensional Time- Dependent Flow

116

Because

xu

N =
x

at

x 2 and

at x,

and A^

Chapter 5

and

at x,

7V2

at

the left-hand side yields

<dT

0i

i-(,i

(5-32)

which denotes (joint) fluid fluxes at nodes 1 and 2, respectively. As explained


in Chapter 3, when we assemble the element equations, the terms in Eq.
(5-32) can vanish, except those at the ends, and they can constitute conditions specified at the ends.

Step

5.

As an

Assembly for Global Equations

one-dimensional bar of metal (Fig. 5-la)

illustration, consider a

through which heat flow occurs. The bar


(Fig. 5-3) with four

divided into three elements

is

nodes; the elements are assumed to be of uniform length

L = -1

L = +1

Figure 5-3 Discretization of bar.

and cross

Use of Eq. (5-21) allows generation of element equations


By observing the fact that temperatures at adjacent
continuous, we can add the element equations to obtain global

section.

for the three elements.

nodes are

equations as follows

"1-1

Ace
I

~2

TV
T2

0"

2-1
0-1 2-1 T
0-1
J

-1

>

6At

1_

"2

R*

*2

>

<

6At

_0

t+At

t;

T2
T

2_

0"

_0

t+At

0~

2_

r1
T

(5-33a)

or in matrix notation,

^[K *]{r} f+Af + ^[K nw


a

+ At

= Wt+

+k%

Here [K*] and [K*] are the assemblage matrices and

{r}

r}

(5-33b)

and {R} are the

Chapter 5

One- Dimensional Time- Dependent Flow

assemblage nodal

unknown

(temperature) and

and the superscript

vectors, respectively,

117

known

forcing function

asterisk denotes assemblage matrices

Note that we can take out a// and l/6At as common if the
assumed to be homogeneous; then Eq. (5-33b) can be written as

in Eq, (5-33a).

material

is

Ar[K*]{r}, +A

where AT =
to be unity)

ccAt/l

+ |[K*]{r} r+Ar = ^{R}, +A + -g-[K*]{r}

(5-33c)

is

a nondimensional time increment (with area

Btu
pc

ft

lbm-F
Btu

hr-ft-Flbm

A assumed

ft^

hr*

Therefore,

AT = 5!

*L
'

hr

ft

= nondimensional.

Boundary Conditions

At

= 0,

we have

scribe values of

at

At

l9

at

<

We

=f

as uniform

choose to pre-

Hence Eq.(5-33)

is

7-[KJ]{r} 0+Af

The

the initial condition [Eq. (5-10)].

T T2i T3i and r4

initial

conditions are applied at

Now we

^-[K*]{r} 0+Af

{R} 0+/

6At

[KJ{r}

(5-34)

= 0.

can introduce end (geometric) boundary conditions


r,

=r,(Q,

r =r

4 (/z,

*)

= *i.

'>o,

=S

t>

4>

(5-35)
o.

A procedure

similar to the one described in Chapter 3 can be used. Here S


and (5 4 are the known values applied at nodes 1 and 4, respectively, and can
be assumed to remain constant with time.
For clarity of illustration let us assume that Ar = 1, T(x, 0) = T and
{R} = {0} then Eq. (5-34) can be expressed as
x

0"

"4

I
sym.

o-

T2

-i

T,

sym.

r4

-!
f

7-1

r,

4-

Af

+
3

"*"

T,
(5-36)

_0

_i

^-*

_j_ _f_o
"*"

To

li
3

\2

One- Dimensional Time- Dependent Flow

118

It is

important to note that the right-hand side essentially yields an equivalent

or residual load vector {R


f

Chapter 5

known temperatures

corresponding to the

at

= 0.
now be

Equation (5-36) can


[Eq. (5-35)].

modified to include boundary conditions

The modification can be performed by

setting

KAA
J = 2, 4,
2, 3,
J =

*n =1,

1,

Ku=0,
K4J = 0,

(5-37)

1,

R =5 U
t

=S

R*

A9

which leads to
0"

'

-i

I
I.

Note

that Eq. (5-38) lost

its

'Ti

s,

T2
T
Tt

(5-38)

r.'

A,

A/

symmetric nature. Often,

it is

economical and

convenient from the computational viewpoint to restore the symmetry of the


matrix.

It

can be achieved, as described

Chapter

in

3,

by modifying further

the other equations as

= 1,2,3,

Kn = 0,
Kj, = 0,
and the right-hand term

0"

|-|o

0-1

f
1

Step

6.

(5-39a)

as

"1

2, 3, 4,

Solve for Primary

[TV

8.

\T2

To

\t3

r.

k.

At

+
+

<5,

(5-39b)

i*

<5 4

Unknowns

Example 5-2

To

illustrate solution

of Eq. (5-39b),

let

us adopt quantitative values for

T 5u and
,

S 4 as follows:

f =
Si

S4

=
=

0.0,

< x < h, < 0,

10degrees|

20 degrees/

f>Q

(5-40)

One- Dimensional Time- Dependent Flow

Chapter 5

Then Eq.

(5-39b) for the

first

time increment

(t

0"

119

At) becomes

10

5_0

(5-41 a)

>
1

"I

oo
I

[T4t

2oJo

Therefore,

!^2

S'3

'

7Y

00
3

'

or

- 5T =
-5T + \6T =
\6T2
2

50,

100,

or by multiplication by 5 and 16, respectively, of the two equations,

8or2 - 25 T =
-8or2 + 2567 231 T =
3

250

1600

1850 =

=>TT =
:
3

we have

8.01

and

16T2

16T2

x 1850

50,

50,

231

40.04

T2 =

5.63.

SECOND TIME INCREMENT


For

the next time increment

-1

-1

0"

Tl
\

-i

_5

T2
T

At

fi

(/

[T4

At

on

Lo

2 At

Introduction of boundary conditions,

i
o

5.63

1 I

8.01

<5i

tJ

we have

[25.63]
]

2A/),

10
f

40.53

=h 57.67

>

20

10 and S 4

Af

(5-41 b)

48.0lJ

20, yields

10

01 (7V

-I
<

-1

T2
T

40.53
>

x 10

6
(

48.01

x 20

1_ IT*] 2Af

20
10
90.43/6
(5-41 c)

148.01/6

20

One- Dimensional Time- Dependent Flow

120

Chapter 5

Therefore
8

It

T -

|r

90.43

'

148.01

or

- 5r =
-5T2 + 16T =
16T2

90.43,

148.01,

or

80r2
80r2

25T3
256r3

=
=

452.15,

2368.16,

or

231T3 =2820.31,

T3 =

12.209,

and

- 5 x 12.209 =
16r2 = 151.4756,
T2 = 9.467,
16jT2

90.43,

and so on for other time increments.


Step

7.

Compute

the Derived or Secondary Quantities

The derived quantities can be rate of flow and quantity of flow. For
example, from Eq. (5-13) rate of heat flow q' for an element can be written as

-kA[B]{T}

= =^[-i
Since

we know T and T2
x

(5-42a)

!]{}

for each element, computation of q'

forward. For instance, for element 2 at time At,

T =
x

10 and

is

straight-

T2 =

5.63.

Therefore
g'(2)

= =^(-10
=

+5.63)

4.37^ Btu/unit

time.

(5-42b)

ONE-DIMENSIONAL CONSOLIDATION
Consolidation

the

is

[3, 8, 9]

phenomenon which describes time-dependent deformamedium such as soil under applied (external)

tion in a saturated porous

The material deforms with time while the liquid or water in the
phenomenon involves coupling
or interaction between deformation and pressure in the fluid.
Under a number of assumptions [2], it is possible to approximate the
phenomenon to occur only in one (vertical) direction; then the stressdeformation behavior of the skeleton of the medium and the behavior of fluid
can be treated separately. The stress-strain behavior is expressed through an
loading.

pores gradually squeezes or diffuses out. This

effective stress

concept given by

a=o'+p,
(5-43)
= total applied stress (Fig. 5-4), a' = effective stress carried by the
skeleton, and p = pore fluid (water) pressure. The stress-strain or

where a
soil

constitutive relation for the deformation behavior of the skeleton can be

expressed as (see Ref.

[2]

or other undergraduate texts on geomechanics)

Aa'

= -Ae,

(5-44)

av

where a v

= coefficient compressibility; e = void ratio,

proportional to axial

and A denotes change due to a load increment.


Under the above assumptions, the governing equation for one-dimensional consolidation is given by Eq. (5-9c); for homogeneous media, it can

or vertical strain

be expressed as

*&-
where

cv

<5

45 >

= coefficient of consolidation = k{\ + e )/y w a k = coefficient of


y w = unit weight of water, y = vertical coordinate, and
v,

permeability,
eQ

= initial void ratio.

The

constitutive law for the flow behavior

is

Darcy's

law:

v=-*- dP.
Equation (5-45)

is

(5-46)

of the same form as Eq. (5-9) for heat flow, and

all

steps of

the finite element formulations are essentially the same.

We

shall therefore detail

an

illustrative

problem for consolidation by

using the results derived previously.

Consider a consolidating mass (Fig. 5-4) divided into three elements


[Fig. 5-4(d)].

Assume

the following properties:

Length of element,

Coefficient of consolidation,

cv

Applied vertical load,

= 10 cm;
= cm /sec;
= g/cm
1

121

One- Dimensional Time- Dependent Flow

122

Chapter 5

TTTTT

ty

(a)

(b)

p(0,t) =

p(y, 0) = p

10

cm

10

cm

A
A

= a

2H
Saturated

30 cm

10

cm

^^^%m%
v&z
p(2H,t) =

^(2H,t)

(d)

3y

(c)

Figure 5-4 One-dimensional consolidation, (a) Consolidating mass,

One-dimensional idealization,

(b)

(c)

Boundary conditions,

(d)

Discretization.

By using Eq.
"
1

10

(5-2 la),

we

obtain, for each element (with

-r

[2

11

-1

i_

{;]+? _1

2_

QM

1),

(5-47)

or
[k.]{p}

The simple forward

( +

[k,]{p}

= {Q(0}.

difference Euler-type integration process will lead to

**L =m))- + ^ti

(5 - 48)

One- Dimensional Time- Dependent Flow

Chapter 5

123

Assembly of the three element equations gives


1

-1

-1

0"

Pi

-1

10

p3
t

10

Pi

Pa

0)

P(0,

t)

>(2H,

t)

= 1.0g/cm
= 0.0,
= 0.0,

f+Al

2
,

Pi
(5-49)

assumed homogeneous medium and A7" = 1.


initial and boundary conditions (Fig.
t

Pi
1

Assume

p(y

\P*\

R>

Pa

Pi

'2

Ri

JU.

12

lj .Pa. + Al

Ri

We

Pi

10

-1

01 Pi

>1

0<x<

5-4) as
h

,t<0,

t>0,

(5-50a)

t>0.

Here we assume that the top and bottom of the consolidating mass are
pervious, and hence the pore water pressure there
times; this constitutes the

first

possible to incorporate impervious boundary, that

type condition. For instance,

dp
(2H,
By

is

is,

the second or

we can have impervious


r)

essentially zero at all

or Dirichlet-type boundary condition.

0,

>

0.

base,

It is

Neumann-

and then
(5-50b)

it is not necessary to make any


boundary condition. The condition of no flow
across the base implied in Eq. (5-50b) is achieved simply by leaving the
boundary node as it is and treating it just like other nodes where p is assumed
as the unknown and obtained through the finite element computations.
The procedure for introduction of the boundary conditions [Eq. (5-50a)]
is essentially the same as for the temperature problem.
An important observation can be made at this stage. The problem of
consolidation is a coupled problem involving interaction between the
deformation of the skeleton of the medium and the pressure in the pores of
the medium. However, it is only because of one-dimensional idealization
and the accompanying assumptions [2, 8, 9] that we are able to consider separately the deformation [Eq. (5-43)] and pore pressures [Eq. (5-45)]. In reality
and in a mathematical sense, the effects of the two phenomena, deformation
and fluid pressure, are coupled.

In the

finite

element formulation herein,

special modification for this

COMPUTER CODE
The code DFT/C-1DFE mentioned

in Chapters 3 and 4 and described in


Chapter 6 can be used to solve the problems of transient heat flow and

consolidation by specifying the required option.

At

this stage, the reader

may

study the portion of the code relevant to these problems. In the following are
described

some

from application of this code.

results

Example 5-3
Figure 5-5 shows a one-dimensional idealization of a homogeneous
into 10 elements

and

Initial

The following

11 nodes.

conditions:

T(x, 0) or p(x, 0)

Boundary conditions:

divided

100 units,

0=0,
or p(2H,t) = 0;

7(0, /) or p(0,

T(h,t)

a or

medium

properties are given:

cv
/

=
=
=

unit,

unit,

0.1 unit.

Temperature Problem. In the case of heat flow, this problem represents timedependent cooling of a bar initially at a temperature of 100 whose ends are cooled
to 0 and kept at that temperature for all subsequent time levels.

pictorial distribution of

the bar

is

shown

computed temperatures

in Fig. 5-6(a). Distribution

at various time steps

along the bar at various times

is

along

shown

in Fig. 5-6(b).

Thus,
sional

it

is

possible to

compute

distribution of temperatures in a one-dimen-

medium by using the finite element

procedure. The temperatures and quantity

of heat flow can be found at any point in the bar and at any time

level,

giving the

under cooling or heating due to a given change in temperature.


It is possible to include different geometrical and material properties for each element; hence, nonhomogeneous media can be easily accommodated. It is also possible to include natural boundary conditions, that is, boundaries that are insulated
against heat. For complete insulation, it can be done just by leaving the node at that
boundary free, that is, by allowing the finite element procedure to compute temperature at the node. This is possible because the formulation procedure includes the
natural or gradient boundary condition automatically, in an integrated sense.
Consolidation. In this case, the problem represents transient deformations of a
porous saturated medium such as a soil foundation subjected to a load of 100 units.
Because the medium is saturated, at /
0, the fluid carries all the applied load
(pressure) and the initial conditions are p(y, 0) = 100 and a'(y, 0) = 0. As time
elapses, the pressure dissipates and the load is gradually transferred to the soil
skeleton and a' increases; at / = oo, o'iy, oo) = 100 units. If the top and bottom
entire thermal history

<

boundaries are pervious,

The time-dependent
temperature (Fig.
124

we can assume

that p(0, t)

=p(2H,

t)

0.0 at

all

times.

distribution of the p(y, t) are similar to the distribution of

5-6). Often, in

geomechanics,

it is

convenient to define degree of

One- Dimensional Time- Dependent Flow

Chapter 5

125

A"
A

Figure 5-5 Finite element mesh.

consolidation

U as

[2]

U=

\7

pdy

hM* e

Pody
where p

M=

is

the initial pressure,

(7T/2)(2/

Tv =

(5-51)

nondimensional time factor

cv t/H 2

and

1).

Figure 5-7(a) shows the distribution of U versus time factor Tv Figure 5-7(b)
shows values of excess pore water pressure along the depth at various values of U
and Tv
The foregoing results indicate that the finite element procedure can be used for
consolidation or settlement computations of foundations idealized as one-dimensional and subjected to (vertical) structural loads. The procedure can yield history
of settlements and pore water pressures with time and the final settlements under
.

a given load increment.

Once

the pore pressure at any time under a given load increment

Eq. (5-43) can be used to compute change in

From

a\

since a, the applied load,

the knowledge of the change in a' and the material property a m

possible to find the change in strain void ratio (Ae)

from Eq.

(5-44).

The

is
is

found,

known.

it is

then

settlement

can be found by multiplying the strain by the total length of the medium
Av(t)

where Av(t)

is

Ae{t)2H,

the vertical settlement at time

t.

(5-52)

(d

126

'ajruejadwaj.

X)

* 2

B
E R
u ')
jo

c a
C= x
09

8!

43
<_

C
B
^jjj

-,

X
V
>

',)

c
X u

.2

g
c
uu
sO

o
ed

>

!
u
o
u
s3 c
cti

i-

127

'ajniejaduj9j_

09

K o
F

'CI

y ed
>

0.10

0.08

0.06

j!

0.04

/
0.02

/
0.20

0.40

0.80

0.60

1.0

Degree of consolidation, U
(a)

Excess pore water pressure, p/p

0.25

0.50

0.75

1.0

(b)

Figure 5-7 Finite element results for consolidation problem, (a)

Time

factor vs. degree of consolidation, (b) Pore water pressure vs. depth.

128

One- Dimensional Time- Dependent Flow

Chapter 5

It is

129

also possible to include layered systems with horizontal stratification.

An

example of such a system follows.

Example

5-4.

Media

Consolidation of Layered Clay

Figure 5-8(a) shows a layered system of soil with four layers

shown

[10].

The material prop-

any convenient units can be assigned to the dimensions and to the properties. Program DFT/C-1DFE was used to solve the problem.
Figure 5-8(b) shows distribution of excess pore water pressures at various time
levels during consolidation. The problem of temperature distribution in layered
media can be solved almost identically.
It can be seen that the distribution of pore water pressure at the interfaces
between the layers is not continuous. If the magnitudes of material properties between two layers differ widely, the discontinuity can cause difficulties. Then it may
be necessary to derive criteria for restricting spatial and timewise meshes [8, 9, 11]

erties are

in the figure [10]

Figure

5-8

Consolidation

layered system

[10]. (b)

in

system,

layered

(a)

Details

of

Dissipation of pore water pressure with

time.

Pervious

Layer

o
CN

cv =

10

m 2 /day, m v

k = 0.1

= 0.01

m 2 /t

m/day

'

>

b
CM

Layer 2

=5. m = 0.001, k = 0.005

Layer 3

=20,

o
CN

'

c v = 40,

E
Layer 4

00
CM

rrv,

k = 0.04

Pervious

(a)

= 0.01. k = 0.20

0.001

One- Dimensional Time- Dependent Flow

130

Chapter 5

Excess pore pressure, p/p


0.4

0.6

0.8

1.0

(b)

Figure 5-8 Contd.

or to use alternative formulation in terms of velocities or stream functions as

unknowns.

As

explained before, once the pore water pressures are known,

it is

easy to find

the deformations at any time level.

PROBLEMS
5-1.

Derive an expression for the "initial load vector" for a line element with linear
approximation for/? and due to given fluid pressure p For p = 10 kg/cm 2
compute {Qo}; solve the problem in Fig. 5- 1(b) with an external load of
.

One- Dimensional Time- Dependent Flow

Chapter 5

1000 kg at the top by including the


[Qo}
5-2.

effect

of p

Hint:

yi

= -A

{BYPo dy.

Formulate the finite element equations for the temperature (or consolidation)
problem with linear variation of areas and material properties as

A =i(l -L)A +^(1 + L)A 2


X

a=i(l -L)a, +i(l +


5-3.

Solve Example

5-2 for

L)a 2

two time steps with the properties as before except the

conditions, which vary as

initial

7*(0, 0)

T(h/2, 0)

T(h, 0)
5-4.

131

=
=
=

10,

20,
10.

Derive equations for the three elements in Fig. 5-3 with the following properties

Element

k, p, c;

Element

2:

2k, p, c;

Element

3k, p,

c.

Assemble the equations and solve for the conditions that

<x<

= 0.0,
t) = 10,
t) = 20,

T(x, 0)
7X0,
T(h,

Assume

that

ment and
5-5.

A u &i and A 2

ct 2

t>
t>

h,

< 0,

0,
0.

are tne values at the

two end nodes of an

ele-

that they vary linearly.

Consider gradually refined mesh for both space and time, and by using DFT/
C-1DFE, study the behavior of the numerical solution for one-dimensional
heat flow (or consolidation); assume the boundary conditions in Example
5-3.

Use

2, 4, 8, 16,

1.0, 2.0, 10.0.

and 32 elements and At

Make various combinations

0.05, 0.1, 0.2, 0.25,

and

0.5

of these values and obtain numer-

Plot error vs. AT = QcAt/l 2 Define error as the difference


between the numerical solution (p) and the exact solution T*(p*)\
ical solutions.

error

T*

or

p*

p,

where
P*

=
J*

cos

nn){sm^)e-^^^

whereto is the uniform initial pressure, for the consolidation problem. Hint:
Note that with increasing value of At, say beyond 2.0, the solution will become
less and less accurate [8, 9].
5-6.

Solve Example 5-2 with forcing fluxes given as q

5-7.

By using DFT/C-1DFE,

solve

Example

5-3

=0.1

unit/unit length.

by considering only half the

One- Dimensional Time- Dependent Flow

132

Chapter 5

depth with the boundary condition at midsection as {dpldy){H, t) = 0.0.


Hint Since the problem is symmetrical about the midsection, dp/dy = 0.
:

5-8.

By

using

DFT/C-1DFE,

dary, that

solve

Example

5-3 for

an impervious bottom boun-

is,

Ty^H,t)=0.
5-9.

Formulate the consolidation problem if the soil were deposited gradually with
1 unit of depth to 10 units of depth in 10 years [12]. Assume linear

time from

Assume required parameters.

variation of the deposition.

REFERENCES
[1]

Carslaw, H.

S.,

and Jaeger,

J.

C,

Conduction of Heat

in Solids,

Clarendon

Press, Oxford, 1959.


[2]

[3]

Terzaghi, K., and Peck, R.


New York, 1955.
Desai, C.

S.,

B., Soil

Mechanics

in

Engineering Practice, Wiley,

and Johnson, L. D., "Some Numerical Procedures

for Analysis

of One-Dimensional Consolidation," in Proc. Symp. on Appl. of Finite Element Methods in Geotech. Eng., C. S. Desai (ed.), Waterways Expt. Station,

Vicksburg, 1972.
[4]

Schiffman, R.

L., private

[5]

Courant,

and Hilbert, D., Methods of Mathematical Physics, WileyYork, 1965.

R.,

Interscience,
[6]

communication.

New

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

[7]

Crandall,

[8]

Desai, C.

S. H.,

to the Finite

Engineering Analysis, McGraw-Hill,

Element Method, Van

New

York, 1956.

and Johnson, L. D., "Evaluation of Two Finite Element FormuOne-Dimensional Consolidation," Comput. Struct., Vol. 2, 1972,

S.,

lations for

pp. 469-486.
[9]

Desai, C.

S.,

and Johnson,

for Consolidation," Int. J.


[10]

L. D., "Evaluation of

Num. Methods

Some Numerical Schemes

Eng., Vol.

7,

1973, pp. 243-254.

Schiffman, R. L., Stein, J. R., and Jones, R. A., "PROGRS-I, A Computer


Program To Calculate the Progress of Ground Settlement Reference
Manual," Report No. 71-4, Computing Center, University of Colorado,

Boulder, 1971.
[11]

Desai, C.

S.,

and Saxena,

K., "Consolidation Analysis of Layered Aniso-

S.

Num.

tropic Foundations," Int. J.

Anal. Methods Geomech., Vol.

1,

No.

1,

Jan.-Mar. 1977, pp. 5-23.


[12]

Koutsoftas, D. C, and Desai, C.


Finite Elements: Solution of
76-17, Dept. of Civil Engg.,

Some

VPI

&

S.,

"One-Dimensional Consolidation by

Practical Problems," Report No. VPI-E-

SU, Blacksburg, Va,

May

1976.

COMBINED COMPUTER CODE


FOR ONE-DIMENSIONAL
DEFORMATION, FLOW,
AND TEMPERATURE/
CONSOLIDATION
By now

the reader

element method.
3, 4,

and

must have

realized the general nature of the finite

A comparison of the derivations and the results in Chapters

5 indicates that there

is

significant similarity in the theory

and

formulation of the three categories of problems. The physical phenomena of

deformation [Eq.
(5-9)]

(3-45)], steady

flow [Eq.

forcing function, either a load or a flux,

medium
these

(4-1)],

and transient flow [Eq.

are governed by analogous differential equations.

in similar

is

The

effect

of a

transmitted or diffused through a

manners, thus indicating the close relationship between

phenomena.

All but very trivial problems that are solved by using the finite element
method require use of the computer. The amount of information to be
digested and processed is so great for most problems that it is not possible to
perform the calculations manually. Hence, a knowledge of and exposure to
the development and use of computer programs or codes become necessary.

To introduce this topic gradually, we present in this chapter details of a


code for the three one-dimensional problems covered in Chapters 3, 4, and 5.
It is shown here that the generality of the method permits use of essentially
the

same program

Flow,

The program is called Deformation


One-Dimensional Finite Element
with sample input/output and documentation is

for all three problems.

Temperature/Consolidation,

DFT/C-1DFE.

listing

provided at the end of the chapter. In the following, details of various stages
of the code are described.
133

PHILOSOPHY OF CODES

computer code should be concise, but

same time,

at the

should be

it

properly documented so that the user can understand and employ

spending an undue amount of time. This

is

it

without

particularly true of a code

To achieve this purpose, and to make


we have provided sufficient commentary

written for a beginner, as in this book.

the code relevant to academic aims,


in the

code and

in the following.

STAGES
The explanation of

number of

the code can be divided into a

include input, computer implementation, and output.

stages,

which

These stages are

described in the subsequent listing of the code.


Stage

Input Quantities

1.

This step deals with the input quantities such as the

common

title,

parameters, material properties, and nodal and element data.

The first card gives details of the number of the problem and
a number of problems can be executed in the same
run. If the problem number NPROB is set equal to zero (a blank card), the
program will automatically exit from the computer. The second card includes
information on number of nodes (NNP), number of materials (NMAT),
number of surface or boundary traction cards (NSLC), option for whether
body force is applied or not (NBODY), and option for choosing the category
Input Set

the

1.

of the problem

title

of problem stress deformation, flow, or temperature/consolidation


:

semibandwidth (IBAND), which in the case of the


for the code is 2; and number of selected time
desired

(NTIME). The

third card or cards (as

give the properties relevant to the category

linear

levels at

many

that

is,

as

(NOPT)

approximation used

which output

number of

is

materials)

deformation, steady flow,

or time-dependent flow; Table 6-1 shows the meaning of the terms for the

weight).

body weight DENS denotes material density (unit


For consolidation of nonhomogeneous media y wi = unit weight of

water

input in place of

categories. In the case of

is

RO.

Input Set 2. This set gives the data on the nodal points. Here

M denotes

number and KODE designates the boundary condition at the


denotes the y coordinate (or any one-dimensional coordinate) of the

the node point

node.

node

point.

^or
134

VLY

gives the specified value of the

an explanation of symbols, see pp. 138-139.

boundary condition

it is

Combined Computer Code for One- Dimensional

Chapter 6

TABLE

135

Material Properties for Various Problems

6-1

Terms

AMV

RO

DENS

Remarks

1.0

1.0

Eq. (3-14)

E*

1.0

1.0

Ji

1.0

1.0

1.0

1.0

yi

PROP
Stress deformation

Homogeneous
Nonhomogeneous
Steady flow

kora

Homogeneous
Nonhomogeneous

ki

or

a,-

Eq. (4-3b)

Time-dependent
Temperature

Homogeneous
Nonhomogeneous

1.0

1.0

Eq. (5-9a)

ki

Ci

Pi

yi

Eq. (5-9b)

Eq. (5-9a)

yi

Eq. (5-9c)

Consolidation

Homogeneous
Nonhomogeneous
*i

1,2,...,//

Cy

1.0

1.0

ywi

number of

tied in with the value of

vi

different materials.

KODE.

If

KODE = 0,

the node

is

"free,"

and we

VLY to be a concentrated load or forcing function; if KODE = 1,


VLY = specified value of displacement, fluid head, temperature, or

can input
then

pore water pressure. Note that


if

gradient or slope

zero,

is

should be provided for

Within input
nodal data. This

set 2,
is

all

if

we

there

input

is

a natural boundary condition,

KODE = 0.

i.e.,

In general, these data

nodes, except in the case of the following.

we have made

applicable only

generated nodes are "free," that

is,

if

provision for automatic generation of


the nodes are equidistant,

and

if

the

KODE = 0, then nodal data are provided

first and last node in the region where these conditions are
Nodal data should be provided whenever there is a change in the
length and properties of the element and in the boundary condition.
We also compute the length / (ALL) of the element and store it for

only for the


satisfied.

subsequent computations.
Input Set 3.

An

The

total

number of elements

NEL

is

set

equal to

NNP

1.

by the two end node (global) numbers and a number


corresponding to the type of material of that element. Hence, for each element, four quantities are input: element number M, node numbers IE(M, 1)
and IE(M, 2), and material type number IE(M, 3). The value of the latter
varies from 1 to NMAT.
element

As

is

identified

explained in the user's guide

matically the element properties

if

(p. 140), it is

possible to generate auto-

certain conditions are fulfilled.

Combined Computer Code for One- Dimensional

136

Chapter 6

Three different variations of (element) areas can be assigned. In the case


all elements, IAREA = 1 and only one value of AREA(l)

of uniform area for


is

input. If

IAREA =

first and the last element


and the computer assigns intermediate area by
variation. If all elements have different areas, IAREA = 3,

2,

areas are input for the

referred to their midsections,

assuming a linear
and areas for all elements (at their midsections) are input.
Input Set 4. Here NSLC denotes the number of surface traction loading
(fy) cards, and TY denotes the value of surface loading. For instance, if in
Fig. 6-2, Example 6-2, Ty is applied on two of the elements, NSLC = 2. The
computer converts this loading to lumped loads at the two surrounding nodes
according to Eq. (3-28). KEL denotes the element on which TY is applied.
If no TY is specified, set NSLC = 0, then no data are required for surface
tractions.

Input Set

5.

This set contains three subsets of data for the time-dependent

temperature or consolidation problems. The

increment (DT), total time


tions,

(TOTIM)

and option (INOPT)

first

subset contains the time

desired for the time-dependent solu-

specifying uniform,

for

linear,

or arbitrary

variation for the initial conditions.

In the second subset, input the specific values of time levels TIM(I) at

which output

The

is

desired; the

number of these values is equal to NTIME.


on initial conditions T(x, 0) [Eq. (5-10)].

third subset contains data

Here, according

INOPT, data

are input as uniform, linearly varying, or

arbitrary values of initial temperature at the nodes. If

INOPT =

1,

input

simply one value UINIT(l) of initial temperature of pore pressure. If INOPT

= 2,

input the values at the

first

and the

last

nodes, and the computer will

provide intermediate values by linear interpolation. If

INOPT =

3,

input

temperature or pore pressure for each node.

Note: The information in input set 5 is not required for stress-deformation


and steady flow problems, that is, when NOPT = 1 or 2.
Stage

2.

Initialize

Various quantities are


set;

NCT

initialized here.

TIME =

0.0

and

NCT =

are

denotes the number of time steps and can be printed out at each

time step. H(I)

is

set

equal to UINIT(I), the

initial

conditions. Matrices

] in Eqs. (5-22a) and


and (5-22c)] and R(I)
are the element and assemblage forcing

QK(I,J) and QP(I,J) are the element matrices [kj and [k

(5-22b), respectively. Vectors Q(I) [Eqs. (3-28), (4-11),

[Eqs. (3-32), (4-13),

and

(5-33b)]

vectors, respectively. Matrix A(I,J) stores the assemblage matrices.

AK(I,J)

is

used repeatedly for the time solution and

during the Gaussian elimination process;

its

value

is

Matrix

is

modified at each step

set

equal to A(I,J) before

Combined Computer Code for One- Dimensional

Chapter 6

starting a

new time

step.

For the

stress

137

and steady flow problems, A(I,J) and

AK(I,J) are computed once, and no such resetting

is

required. AP(I,J)

the

is

assemblage matrix corresponding to [K *] [Eq. (5-33b)].


f

Intermediate quantities required for computation of the time factor in the


consolidation problem are evaluated here.
Stage

3.

Compute Element Matrices

Element matrix

computed

for

all

gravity load are

[k]

and load vector {Q} due to surface traction or

flux are

elements one by one. The contributions at nodes due to the

computed according

to Eq. (3-28)

and are added to the

element load vector {Q}.


Stage

At

4.

Assemble

this stage,

we add

the contributions of the element matrix and load

vector to the appropriate locations in the global matrix [K] and load vector
is done by identifying the global nodes corresponding
and 2 for each element.

{R} [Eq. (3-32)]. This


the local nodes

Stage

5.

to

Concentrated Forces

The concentrated

forces

VLY(I) are added

to the corresponding locations

in vector {R}.

Stage

6.

Boundary Conditions

The boundary conditions

are introduced here to modify the assemblage

matrices and load vectors [Eqs. (3-32), (4-13), and (5-33)]. In view of the fact
that

we have taken advantage of bandedness and symmetry,


make this modification.

special logic

is

used to

Stage

7.

Time Integration

This is operational only for the time-dependent problem (NOPT = 3 and


and pertains to Eq. (5-33). Here the equivalent load vector [K,]{r} is
computed and added to the assemblage vector {R}.

4)

Stage

The

8.

final

Solve Equations

equations are solved by performing the Gauss-Doolittle elimina-

tion procedure.

For

NOPT =

or

2,

this is

performed only once. For

Combined Computer Code for One- Dimensional

138

NOPT = 3 and 4,
steps = TOTIM/DT.

it

is

Chapter 6

performed as many number of times as the time

The space of vector R(I)

used here to store the values of computed

is

displacements, fluid heads, temperatures, or pore water pressures.


Stage

This

Set {R}

9.

is

{H}

relevant only for

computed are stored

{R} f+Af

NOPT =

and

4.

Here the values of R(I) just

in H(I) as the value {R} f at the previous time stage.

Output Quantities

Stage 10.

Depending on the value of NOPT, computed

results are printed

out as

follows

NOPT =

=2
=3
=4

results for stress-deformation

results for flow

results for

problem,

problem,

time-dependent temperature problem,

results for consolidation

problem.

In the case of consolidation, additional quantities such as time factor and

degree of consolidation are also printed out.

Explanation of Major Symbols and Arrays

A =

AK =

Assemblage matrix which is set equal to AK; only once for stress deformation and steady flow but at every time step for transient flow.
Assemblage matrix which is computed only once. This matrix corresponds to
[K][Eqs. (3-32), (4-13) or (5-33)].

ALL =
AP =
AREAEL =

DENS =
DT =
GWT =
H=

Length of each element.


Assemblage matrix corresponds to [K] [Eq. (5-33)].
Area of elements. If IAREA = 1, input one value of uniform area,
first and last values, = 3 input all values.

=
=
=

2 input

Density (unit weight) for different materials.

Time increment

[Eqs. (5-23)

and

(5-33)].

Contribution of gravity or body force at an element node.

Vector that stores


Rt+At

IAREA =

computed

at

initial
/

values of temperature (or pore pressure) and also

A/ time

level for use as

for the subsequent time level.

Option for specifying areas of elements


for uniform area over all elements.
1
:

2: for linearly varying areas; input values for only the first
3: for arbitrary variation, that

is,

and

input area for each element.

last

elements.

Combined Computer Code for One- Dimensional

Chapter 6

IBAND =

139

Semibandwidth. Set 2 for the one-dimensional problem with linear approximation.

= Node 1 of element M.
= Node 2 of element M.
Material type of element M.
INOPT = Option for initial conditions
= uniform temperature or pressure at all nodes.
= 2 linearly varying temperature or pressure input values only for the first and

IE(M,1)
IE(M,2)
IE(M,3)

the last nodes.

= 3: arbitrary; input values for all nodes.


KEL = Element(s) on which surface tractions (TY) are applied.
KODE = Code for boundary conditions:
= specified displacement, potential, temperature, or pore
= 0: free node where concentrated force can be applied.
NBODY = for no body force, = if body force specified.
NCT = Step of time increment can be printed out if desired.
NMAT * Number of materials.
NNP = Number of nodes.
NOPT = stress deformation, = 2 steady flow, = 3 transient
1

pressure.

temperature,

consolidation.

NPROB = Problem number if = 0, program exits from the computer.


NSLC = Number of surface traction cards.
NTIME = Number of time levels at which output is desired.
PROP = Material property:
= E, elastic modulus for stress deformation.
= k, coefficient of permeability for flow.
= a, thermal diffusivity for temperature (for homogeneous medium).
= c coefficient of consolidation (for homogeneous medium). For
;

layered

media, see Table 6-1.

QK =
QP =

R=

Element matrix [k] in Eqs. (3-28) and (4-9) and [k a ] in Eq.


Element matrix [k ] in Eq. (5-21).
Assemblage forcing vector {R} in Eqs. (3-32), (4-13), and

(5-21).

(5-33).

Computed

nodal displacements, potential, temperature, or pressure are stored in

this

vector after Gaussian elimination.

RO =
TF =
TIM =
TIME =
TITLE =
TUINIT =

TY =
UAV =

Mass
Time

density of material.
factor.

Selected time levels at which output

is

desired.

Elapsed time.
Title

and description of problem.


initial values = (sum of initial pore pressures) x (number of nodes).

Total of

Applied surface traction load.


Average degree of consolidation

at a given time.

UINIT =

Values of

USUM =

Average dissipation of pore pressure at a given time.


Value of applied boundary condition. If KODE = 1,

VLY =

initial

temperature or pore pressure at time

placement, potential, temperature, or pressure;

Y=

it

0.

implies specified dis-

KODE =

0,

it

implies free

node where forcing parameters (concentrated load) can be specified.


y coordinate for stress deformation (of vertical column and strut) and for
consolidation.

if

<

x coordinate

for flow

and temperature.

c
Si

L.

Ql

*-

co

-o

'I

Si

=3

o
I

Q
-I

CJ
LUI

"5

(D
i<D

CO
in

.Q

140

oc

irf

15
C
E
*

CN
in

00

s-

a O
.2

u.

co
-

<

IS

is

<n

_J

-D

3
UJ

S
w

<

>

4->

Q.

3^

CD
09
ro

CO
CD

C*

1ro

03

>

4->

CO

CD

+_,

r 0)
n C

a>
i_
CD

to
oj

=
(0
>

D C
a.
~c >:

*-

i
2

03

fc

<*>

it

g>

4-

M-

o "-T D
_
il_
o o O
"+CN CO

II

fe<o
-

OH

II

II

CO

c >

to <
a
S
= s
O
"-

_2>

<

o5

IO
"-H

h-

<

oi

<u

Ou-

3
3

o
U_

LLI

141

'

o 2
C Q.

"8

CO

2!

il
z

o o
c
E
.c

o
o

c
o

c/>

.2

k"

o
M

'5

CN

co

15

ro

'-^

II

II

*-

!tr

!=

'-o

CO

O)

>

a>

CO

+_<

liJ

+-

CO

a
3
o

h-

<
nr

142

O
u_

^
1-

<r>

LU

c
O

co

'6

<

1
= O

C
o

-a
2?

"5

r-

C6

C
g tr

Combined Computer Code for One- Dimensional

Chapter 6

Note

1: (a) If the

143

nodal points are equidistant, data are required only

for the

and the

first

last

node. The data for intermediate

nodes are generated by the computer. If a boundary condition or force

is

applied at a node, data must be input for such

nodes,
(b)

KODE = 0:

Node

is

free,

and concentrated force can be

applied.

KODE =

Boundary condition

potential,

in terms of displacement,

temperature,

or pore pressure

is

specified.

KODE is set equal to zero for the nodes that are generated by
the computer.

Note

2: If the elements

RO, DENS, data

have same material properties,

PROP, AMV,

and the last element; data for


intermediate elements are generated by the computer, and their PROP, AMV,
RO, and DENS are set equal to those for the previous element. If there is a
change in material property, data must be separately provided. For instance,
for 1 1 elements, if there is a change in material property at element 6, then
data for elements 1 and 5 and 6 and 1 1 should be provided.

Example

See Fig. 6-1.

are required only for the

first

6-1

By using DFT/C-1DFE, compute displacements and axial

stresses. Plot

the results.

10 kg

E
u

(3) E= 1000 kg/cm


/ = 10 cm (for each
= 1.0 cm 2

E~

A =0

element)

Vl

.it

\/h/////,

Figure 6-1

Example 6-2
See Fig. 6-2.
stresses.

By using DFT/C-1DFE, compute and

plot

displacements and

"

Combined Computer Code for One- Dimensional

144

(3)
^S

Chapter 6

E= 1000 kg/cm 2
/=10cm
_A= 1.0 cm

= 1.0 kg/cm
3
y = 0.5 kg/cm

mW0
Figure 6-2

Example 6-3

The pipe in Fig. 6-3 is subjected to potentials


and 4, respectively. Use DFT/C-1DFE, and
tials and velocities.

(p

10cm +- 10cm
,

'

= 100.0

cm 2

find

2 and 1 at the two end points 1


and plot distributions of poten-

+-

10cm

k x = 1.0 cm/sec,

q = 0.0

Figure 6-3

Example 6-4

The bar in Fig. 6-4 is initially


The boundary conditions are
7(0,

The coefficient of diffusivity CC


and adopt A/ = 1.0 unit.

at a

/)

=
1

temperature of 100F; that

T(10,
.

/)

0,

is

100F.

is, 1 1

nodes,

> 0.

Divide the bar into 10 elements, that

10 units

T(x, 0)

Figure 6-4

Use DFT/C-1DFE, and compute the

distribution of temperature at various

time levels as the bar cools.

PROBLEMS
See the problems in Chapters 3-5.

LISTING

AND SAMPLES OF INPUT/OUTPUT:

DFT/C-1DFE
C

C
C
C

Q
C

C
C

20

MAIN
FINITE ELEMENT C3DE FOh ONE-DIMENSIONAL DEFORMATION. ^LOW,
MA I
TEMPERATURE AND CONSOLIDATION
MAIN
OFT/C-iDFC
PROGRAM NAME
MAIN
DEVELOPED BY C.S.OESAI
**t^*^**Wt**A ********** ******** *************--****** *t*****W****yAIN
DIMENSION A(41,3), AK(4l,3), AP(41,3>, H(*l), >(4l), UK(2,2), 0P(2main
1,2), 0(2), LP(2), TIM|20I Y(4l), IE(40,3), Vt.Y(4l), &ROPIIO), ARFMAIN
2AEL12J), DENSC101, K0DE(41), ALLIED, TY(2DJt KELC201, ^IICI, AMVMAIN
MAIN
3(10), UINIT(41), TITLE(ld)
****** **W** **? * *********** **V** :**.**** *****: ***.******** ft*******y^J\J
MAIN
**
STAGE 1***- INPUT QUANTITIES
***A ******
***********************************> *^
FPR EXPLANATION OF VARIOUS STAGES
CHAPTEF 6
MAIN
SEE
***
INPUT SET I
MAIN
REAO 920, NPRUB. TITLE
MAIN
IF INPRQ3.LE.0I GO TO 910
MAIN
WRITE (6,930) NPROB, TITLE
AIN
WC ITE (6,940)
MAIN
WRITE (6,950)
MAIN
*** PROBLEM PARAMETERS
MAIN
REAO (5,960) NNP , NMA T, NS LC , NBODY ,NCPT , I 3 ANO, NT ME
MAIN
WRITE (6,9/0) NNP.NMATtNSLCtNBODY.NOPT, IbANOtNTIME
MAIN
MAIN
*** MATERIAL PROPERTIES
WRITE (6,990)
MAIN
DO 3
1=1, NM AT
MAIN
READ (5,980) PROPC I ) , AMV (I , PC ( 1 ) , DENS
MAIN
CONTINUE
MAIN
WRITE (6,1000)
I, PROP (I ),AMV( I),PC( I) ,UENS(I) ,I=1,NMAT)
MAIN
**
INPUT SET 2
MAIN
*** NODAL POINT OATA
MAIN
WR ITF (6,1010)
MAIN
N=l
MAIN
READ (5,1020) M, KOOE M , Y( M , VLY( M
MAIN
IF (M-N) 50,80,60
MAIN
CONTINUE
MAIN
WRITE (6,1030) M
MAIN
GO TO 910
MAIN
AUTOMATIC GENERATION OF NCDAL POINT OATA * *
MAIN
DF=M+1-N
MAIN
RY=(Y(M)-Y(N-1) )/DF
MAIN
KGDE(N) =
MMN
Y(N)=Y(N-l)+RY
MAIN
VLY(N)=0.0
MAIN
IF (N.EQ.l) GO TO 100
MAIN
****
**
COMPUTE
MAIN
ELEMENT LENGTH
ALL(N-l)=Y(N)-Y(N-l)
^AIN
MAIN
CONTINUE
WRITE (6,1040) N,K0DE(N) ,Y(N),VLY(N)
MAIN
N=N+1
MAIN
IF (M-N) 110,90,70
MAIN
IF (N.LE.NNP) GO TO 40
MAIN
******** a* **************** ********^*******w****************it*5**MA IN
***
MAIN
INPUT SET 3
*** ELEMENT OATA
MAIN
jp

Jf

X.

<

33

C
C

43
53

60
70

80
C

93
100

110
C
C
C
C

123
130
140

******<t'*******S'*****************W**********TA*t4!**?r: V t**yt**^**i*MJ,\j
WRITE (6,1050)
MAIN
NEL=NNP-l
MAIN
N=0
MAIN
REAO (5,963) M , I E M,
MAIN
, 1= 1 , 3
MAIN
N=N+1
MAIN
IF (M-N) 140,160,153
WRITE (6,1360) M
MAIN
GO TO 910
MAIN
(

145

35

40
50
50
7")

80
90
103
110
J20
133
]^Q
150

160
170
130
190
200
210
220
233
240
250
260
270
283
293
300
313
320
330
340
350
363
373
380
390
400
410
423
430

440
450
460
473
480
490
533
510
520
533
540
55)
560
57Q
583
590
600
613
62

630
640
650

1 E

190

N,2

210

N- 1 , 2

MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
AIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN
MAIN

IE(N,3)=IE(N-l,3)
(M-N) 173,170,130
(NEL-N) 180,180,120
CONTINUE
INPUT ELEMENT AREAS
READ (5,960) IAREA
GO TO (190,210,230), IAREA
READ (5,980) AREAEL(l)
IF
IF

DO 200

200

IE(N,i)=IE(N-l,l)+i
I

160
170
180

))

and Samples (cont.)

Listing
150

))

= 1, NEL

AREAEK I)=AREAEL(l

GO TO 240
READ (5,980) AR AEL( 1 ,AR?AE L NEL
AL=Y(NNP)-Y( I) -(Al_L( 1 +A LL NEL /2
SLUPE=(AREA5L( NEL -ARE AEL 1
/ AL
NEL1=NEL-1
(

DISTY=0.0
DO 220

=2, NEL

AREAEL

(I

AREAEL(

/2
+ S L OP E*D STY
(

.0

CONTINUE
GO TO 240

230
240
250
Q
C
c
Q

260
270
280
C
C
C
C

290

300
310

320

330
340
350

fc (

>

30

***
MAIN 940
INPUT SET 4
*** sukface traction cards
MAIN 950
******************** ** *************-*********** **** ******* * *****vAIN 960
MAIN 973
IF (NSLC.EU.O) GO TO 280
WRITE (6,1080)
MAIN 980
MAIN 990
DO 263 I=1,NSLC
READ (5,1090) KEL(I),TY(I)
MAIN1000
MAIN1010
DO 270 I=1,NSLC
WRITE (6,1100) I ,KEL I , TY I
MAIN1020
(

CONTINUE
MAIN1030
**********************************:******************************* na sjio40
***
INPUT SHT 5
MAIN1050
***
DATA FOR TIME DEPENDENT PRPRLFMS
MAIN1060
****** ** ************ ******** ******** *********** ******************* m,a mio 70
IF (N0PT.LT.3) GO TO 370
MAIN1080
WRITE (6,1110)
MAIN1090
READ (5,1120) DT,T3TIM, INOPT
MAIN1100
WRITE (6,1130) DT,TOTIM, INOPT
MAIN1110
WRITE (6,1140)
MAIN1120
1= 1 ,NT I ME
READ (5,980)
MAIN1130
T IM(
MAIN1140
DO 290 I=1,NTIME
WRITE (6,1150) I,TIM(I)
MAIN1150
MAIN1160
INOPT =1
UNIFORM INITIAL COND I TI CNS,=2 LINEAR, =3 ARBITRARY
MAIN1170
WRITE (6,1160)
GO TO (300,320,340), INOPT
MAIN1180
READ (5,980) UINIT(l)
MAIN1190
MAIN1200
DO 310 1=1, NNP
MAIN1210
UINIT( I)=UINIT(1)
GO TO 350
MAIN1220
MAIN1230
READ (5,980) UIN IT 1 ,UINIT NNP
MAIN1240
NNPl=NNP-l
MAIN1250
DO 330 I=2,NNPi
MAIN1260
AL=Y(NNP)-Y(l)
MA IN 1270
SL0PE=(UINIT(NNP)-UINIT( 1) )/AL
MAIN1280
UINITt I)=UINIT( I +SLOPE*Y I
MAIN1290
GO TO 350
I =1
MAIN1330
NNP
READ (5,980)
Ul NI T(
MAIN1310
CONTINUE
= 1, NNP
MAIN1320
DO 360
MAI.M1330
WRITE (6,1170) I,UINIT(I)
i

360

773
780
793
800
810
923
R30
MA I
840
MAIN 850
MAIN 860
MAIN 37 3
MAIN 883
MAIN 890
MAIN 9)3
MAIN <U3
MAIN 920

READ (5,980)
AREAEL
), 1=1 NEL
CONTINUE
DO 2 50 M=1,NEL
A E a:- L w
M, 2
WRITE (6,1070) M, E M, 1
I
E m, 3)
******************** ******** ************ ********* V ******** < **r****MAlN

50
76 3
7

\!

0ISTY=DISTY *(ALL( I-D +ALL


220

660
670
680
690
700
713
720
730
740

146

I )

Listing

and Samples (cont.)

370

*t*****r*****a ****************** **************-p******t***********MAlN13'V0


*** STAGE 2*** INITIALIZE
MAIN1350
**********************+************************************ MA IN1360
CONTINUE
MAIN1370
NCT =

MA IN 1380

380

TIME=0.0
00 380 I=l,NNP
IF (N0PT.LT.3) UINIT(I)=0.0
H(I)=UINIT(I)
R(I)=0.0
00 383 J= 1,1 BAND
A(I,J)=0.
AK(I,J)=0.
AP(I,J)=0.

MAIN1390
MAIN1400
MAIN1410
MAIN1420
MAIN1430
MAIN1440
MAIN1450
MAIN1460
MAIN1470

DO 520 M=l ? NEL

MAINU80

Q
C
C

II1=IE(M,1)
MAIN1490
II2=IE(M,2)
MAIN1500
ALEN=ABS(Y( I II )-Y( 112)
MAIN1510
MT=IE(M,3)
MAIN1520
DO 390 1=1,2
MAIN1530
Q(I)=0.0
MAIN1540
DO 390 J=i,2
MAIN1550
QK(I,J)=0.
MAIN1560
QP(I,J)=0.
HAIN1570
IF (N0PT.LT.4) GO TO 410
MAIN1530
FOR CONSOLIDATION (OF LAYERED MEDIA) FIND TIME FACTOR ON THF PAS I S MA I M1590
OF AVERAGE CV, THIS IS AN APPROXIMATION , ALTERNATIVELY THIS
MAIN1600
CAN BE DONE ON THE BASIS OF ONE OF THE LAYERS
MAIN1610
MAIN1620
ANEL=NEL
MAIN1630
CV=0.0
MAIN1640
DO 400 MM=1,NEL
MAIN1653
MTT=IE(MM,3)
MAIN1660
CV=CV+PROP (MTT)/(RC(MTT)*AMV(MTT))
MAIN1670
CVA=CV/ANEL
HH=( Y(NNP)-Yd) )/2.0
MAIN1680
MAIN1690
TFF=CVA/(HH*HH)
MAIN1700
CONTINUE
****************** *******-********<-*********v9**"e******>" ****#***VAIN1710
*** STAGE
3*** COMPUTE
ELEMENT MATRICES MAIN1720
****** ft*************************************************** ********MAIN1730
MAIN1740
IF (N0PT.LT.3) GO TO 450
IF (N0PT.EU.4) GO TO 420
MAIN1750
TEMPERATURE PROBLEM
MAIN1760
0Rl=(AREAEL(M)*PR3P(MT) )/ALEN
MA IN 17 70
MAIN 1780
DR2=(AMV(MT)*R0(MT)*ALEN)/DT
MAIN1790
GO TO 440
MAIN1800
CONSOLIDATION PROBLEM
MAIN1310
DRl=PROP(MT)/(RO{MT)*ALEN)
0R2=(AMV(MT)*ALEN)/DT
MAIN1820
MAIN1830
TUINIT=0.0
MAIN1840
DO 430 1=1, NNP
MAIN1850
TUINIT=TUINIT+UINIT( I)
MAIN1860
CONTINUE
MAIN1870
QK(1,1)=DR1
MAIN1830
QK(2,2)=QK(l,l )
MAIN1890
QM1,2)=-QK<1,1)
MAIN1900
QK(2,1)=QK(1,2)
QP(1,1)=0R2*(1. 0/3.0)
MAIN1910
MAIN1920
QP(l,2)=QPll,l)/2.
MAIN1930
QP(2,2)=QP(1,1)
MAIN1940
QP(2,1)=QP<1,2)
MAIN1950
GO TO 500
MAIN1960
CF=(AREAEL(M)*PROP(MT) l/ALLCMI
MAIN1970
QK(1,1)=CF
MAIN1980
QK(2,2)=CF
MAIN1990
QK(1,2)=-CF
MAIN2000
QK(2,1)=-CF
)

390
C
C
C

400

410
C

C
q

420

430
440

450

147

Listing
C
C
C

460

470
480
C

490
500

510
520
530

C
C

C
C
C
C

540
C
C

C
C

550
C
C

q
C

560

and Samples (cont.)

**************************************************** ************** ma J nj20 10


*** STAGE 4*** ASSEM8LE
MAIN2020
****************************************************************** ma I N2030
IF (NSLC.EQ.O) GO TO 480
MAIN2040
COMPUTE ELEMENT FORCE VECTOR Q DUE TC TRACTION OR FLUX
MAIN2050
00 460 IM=1,NSLC
MAIN2060
MK=KEL(IM)
MAIN2070
IF (MK.EQ.M) GO TO 470
MAIN2080
CONTINUE
MAIN2090
GO TO 480
MAIN2100
Q( 1)=Q(1)+(TY( IM)*ALL(MKI)/2.0
MAIN2U0
MAIN2120
Q(2)=Q(2)*U(1)
CONTINUE
MAIN2130
ADD FORCING VECTOR DUE TO BODY FORCE OR FLUX TO ELEMENT LOAD VECTOMA I N2140
IF (NBODY.NE.l) GO TO 490
MAIN2150
GWT=(AREAEL(M)*ALL<M)*DENS(MT) 1/2.0
MAIN2160
Q(l=Q(l)+GWT
MAIN2170
Q(2)=Q(2)+GWT
MAIN2180
CONTINUE
MAIN2190
CONTINUE
MAIN2200
LP(1)=M
MAIN2210
LP(2)=M+1
MAIN2220
00 510 LL=1,2
MAIN2230
I=LP(LL)
MAIN2240
R< I)=R(I )+Q(LL)
MAIN2250
DO 510 MM=1,2
MAIN2260
J=LP(MM)-I+1
MAIN2270
IF (J.LE.O) GO TO 510
MAIN2280
AKU ,J>=AK(I,J)+QK(LL,MM)
MAIN2290
IF (N0PT.LT.3) GO TO 510
MAIN2300
AP( I,J)=AP( I,J)*QP(LL,MM)
MAIN2310
MMN2320
CONTINUE
MAIN2330
CONTINUE
VAIN2340
NCT=NCT+1
IF (N0PT.GE.3) TIME=TIME+DT
MAIN2350
MAIN2360
IF (N0PT.EQ.4) TF=TFF*TIME
HAIN2370
DO 540 1=1, NNP
MAIN2380
DO 540 J=l,IBAND
**** IMPORTANT NOTE ***** IF NO (TIME DEPENDENT) FORCING
MAIN2390
ELEMENT PARAMETERS SUCH AS FLUX ARE APPLIED, THEN INITIALIZE VECTOMA IN2400
R AT THE START OF EACH TIME STEP**
MAIN2410
IF TIME DEPENDENT FORCING
MAIN2420
PARAMETERS ARE APPLIED THEN STAGES 3 AND 4 SHOULD 8E PERFORMED
MAIN2430
AT EACH TIME LEVEL AND THE ELEMENT CONTRIBUTIONS ADDED TO P
MAIN2440
CONCENTRATED TIME DEPENDENT FOPCES CAN HOWEVFP 3E APPLIED
MAIN2450
IF (N0PT.GE.3) R(I)=0.0
MAIN2460
A( I,J)=AK( I, J)
****************************************************************** MA I N2470
MAIN2480
*** STAGE
5***C0NCENTPATED FPRCFS
MAIN2490
ADD CONCENTRATED FORCES TO ASSEMBLAGE LOAD VECTOR R
****************************************************************** ma N2500
MAIN2510
DO 550 1=1, NNP
IF (KODE(I).NE.O) GO TO 550
MAIN2520
R(I )=R(I)+VLY{ I)
MAIN2530
MAIN2540
CONTINUE
****************************************************************** MA I N2 550
6***
*** STAGE
BOUNDARY CONDITIONS
MAIN2560
*************************************************** ******** *******m A I N2570
MAIN2590
ADD K(ALPHA) *K(T)
MAIN2600
DO 560 1=1, NNP
MAIN2610
DO 560 J=1,IBAND
MAIN2620
A( I,J)=A( I,J)*AP(I,J)
MAIN2630
CONTINUE
MAIN2640
DO 600 N=1,NNP
IF (KODE(N).EQ.O) GO TC 590
MAIN265D
MAIN2660
BOUND=VLY(N)
MAIN2670
DO 580 M=2,IBAN0
MAIN2680
K=N-M*i
MAIN2690
IF (K.LE.O) GO TO 570
I

148

Listing

570

580

590
600
C
C

R<K)=R(K)-A<K,M)*BOUND
A(K,M)=0.0
AP(K,M)=0.
K=N+M-i
IF (K.GT.NNP) GC TO 580
R(K)=R(K)-A(N,M)*B0UND
A(N,M)=0.0
AP(N,M)=0.
CONTINUE
A(N,1)=1.0
AP(N,i)=0.
R(N)=BOUN0
CONTINUE
CONTINUE

and Samples (cont.)


MMN2700
MAIN2710
MAIN2720
MAIN2730
MAIN2740
MAIN2750
MAIN2760
HAIN2770
MAIN2780
MAIN2790
MAIN2800
MAIN2810
MAIN2820
MAIN2830

*** STAGE 7*** TIME


MAIN2850
INTEGRATION
************ *********** *****r***x*********:,* T:**********r**-*AT N2860
IF (N0PT.LT.3) GO TO 650
MAIN287D
00 640 1=1, NNP
MAIN2880
MAIN2890
BB=0.
IF (I.EQ.i) GO TO 620
MAIN2900
MAIN2910
IP=I-1
1AIN2920
KK=1,IP
00 610
MAIN2930
JJ = I-l-KK
MAIN2940
COMPUTE RIGHT HANO SIDE 'EQUIVALENT LOAD* K(T)* R<T)
"*AIN2950
IF
JJ.GT.I8AND) GO TO 6 10
MAIN2960
BB=BB+AP(KK,JJ)*H(KK)
MAIN2970
CONTINUE
MAIN2980
CONTINUE
MAIN2990
J=1,IBAND
DO 630
MAIN3003
II = H-J-1
MAIN3C10
IF (II.GT.NNP) GO TO 630
MAIN3020
BB=BB+AP(I ,J)*ri( II
MAIN3030
CONTINUE
MAIN3040
R(I)=R<I)+BB
MAIN3050
CONTINUE
MAIN3060
CONTINUE
{

610
620

630

640
650
C
C
q

660
670

680

8*** SOLVE
EQUATIONS
MAIN3080
*** STAGE
MAIN3090
EQUATION SOLVER - GAUSS- DOOL ITTLE ELIMINATION PROCEDURE
******************************************************************** MA I N3100
MAIN3110
NRS=NNP-1
NR=NNP
HAIN3123
*MN3130
DO 670 N=l,NRS
MAIN3140
M=N-1
MAIN3150
MR=MIN0(IBAN0,NR-M)
MAIN3160
PIV0T=A(N,1)
MAIN3170
DO 670 L=2,MR
MAIN3180
C = A(N,L)/PWOT
=
MAIN3190
1
M+L
MAIN3200
J=0
MAIN3210
DO 660 K=L,MR
MAIN3220
J=J+1
A( I, J)=A(I , J)-C*A(M, K)
MAIN3230
A(N,L)=C
MAIN3240
DO 680 N=1,NRS
MAIN3250
M=N-1
MAIN3260
MR = MIN0( IBAND,NR-M)
MAIN3270
C=R(N)
MAIN3230
R(N)=C/A(N,1)
MAIN3290
DO 680 L=2,HR
MAIN3300
I=M+L
MAIN3310
R( I )=R( I)-A(N,L)*C
MAIN3320
R(NR)=R(NR)/A(NP,1)
MAIN333}
DO 690 1=1 MRS
AIN33^0
N=NR-I
MMN3350
M=N-1
MAIN3360
MR=MIN0( IBANUNR-M)
MAIN3370
,

149

Listing

and Samples (cont.)

00 690 K=2,MR

MAIN3380
MAIN3390
R(N)=R(N)-A(N,K)*R(L)
MAIN3400
******************** ** r ******* ******** ************************** **MAIN3413
*** STAGE 9*** SE T R(T)=H( )= RCT+OT)
MAIN 3420
************************************** ************************** **MAIN3430
00 700 1=1, NNP
MAIN3440
H(I|=R<II
MAIN3450
************************************** ************** ************ **MAIN3460
*** STAGE 10*** OUTPUT
QUANTITIES MAIN3470
************************************** ************************** **MAIN3430
IF (N0PT.GE.3) GO TO 710
MAIN3493
WRITE (6,1180)
MAIN3503
CONTINUE
MAIN3510
GO TO (720,750,780,830), NGPT
MAIN3520
*** OUTPUT FOR STRESS-DEFORMATION PROBLEM MAIN3530
WRITE (6,1190)
MAIN3540
MAIN3550
WRITE (6,1200)
MAIN3560
00 733 1=1, NNP
MAIN3570
WRITE (6,1250) I,R(I)
MAIN3530
COMPUTE
STRESSES
MAIN3590
WRITE (6,1210)
MAIN3600
DO 740 L=2,NNP
MAIN3613
MT=IE(L-1,3)
MAIN3620
STRFSS = (P(L)-R(L-1) *PROP( MT )/ ALL L- 1)
MAIN3630
L1=L-1
MAIN3640
WRITE (6,1230) LI, STRESS
MAIN3650
GO TO 890
MAIN3663
PRINT OUT RESULTS FOR FLOW PROBLEM//)
*** OUTPUT FOR FLOW
MAIN3670
MAIN3680
WRITE (6,1240)
WRITE (6,1270)
MAIN3690
MAIN3700
00 760 1=1, NNP
MAIN3710
WRITE (6,1250) I,R(I)
MAIN3720
COMPUE
VELOCITIES
MAIN3730
WRITE (6,1220)
MAIN3743
DO 770 L=2,NNP
MAIN3750
MT=IE(L-l,3)
MAIN3760
VEL0=(R(L)-R(L-1))*PR0P(MT)/ALL(L-1)
MAIN377D
VELO=-VELO
MAIN3780
Li=L-i
MAIN3790
WRITE (6,1230) Ll,VELO
MAIN3800
GO TO 890
*** OUTPUT FOR TRANSIENT TEMPFRATURE
MAIN3810
MAIN3820
IF DT IS CHOSEN TO BE LESS THAN 0.0001 , CHANGE TOLER
MAIN3830
TOLER=0.000i
IF (TIME.GT.OT) GO TO 790
MAIN3840
MAIN3850
WRITE (6,1180)
MAIN3860
WRITE (6,1280)
MAIN3873
CONTINUE
IF (TIME.GT.TOTIM) GO TO 900
MAIN3880
DO 800 I=1,NTIME
MAIN3890
DIF=ABS(TIME-TIM(I))
MAIN3900
MAIN3910
IF (DIF.LT.TOLER) GO TO 810
MAIN3920
CONTINUE
GO TO 530
MAIN3930
WRITE (6,1260) TIME
MAIN3940
WRITE (6,1290)
MAIN3950
MAIN3960
DO 820 1=1, NNP
MAIN3970
WRITE (6,1250) I,R(I)
MAIN3983
GO TO 530
*** OUTPUT FOR
CONSOLIOATIGN MAIN3990
MAIN4000
T0LER=0.000l
MAIN4010
IF (TIME.GT.OT) GO TO 840
MAIN4020
WRITE (6,1180)
WRITE (6,1300)
MAIN4030
MA 1^4040
CONTINUE
MAIN4050
IF (TIME.GT.TOTIM) GO TO 900
L = M-K

690
C
C
C

700
C

c
c

710
C

720

730
C

740
C
C

750
760
C

770
C
c

780

790

800

810

820
C

830

840

150

Listing

850
860

870

880

890
900
910

C
C

920
930
940
950
960
970

and Samples (cont.)


MAIN4060
MAIN4070
MAIN4080
MAIN4090
MAIN4100
MAIN4110
MAIN4120
MAIN4130

00 850 I=1,NTIME
DIF=ABS(TIME-TIM(I)I
IF (OIF.LT.TOLERI GO TO 860
CONTINUE
GO TO 530
CONTINUE
USUM=0.0

DO 870 1=1, NNP


UZ=R(I)/TUINIT
USUM=USUM*UZ
UAV=1.-USUM
WRITE (6,1310) TIME,TF,UAV
PRINT OUT NOOAL PORE PRESSURES
WRITE (6,1320)
00 880 1=1, NNP
WRITE (6,1330) I,R(I)
IF (UAV.GE.0.98) GO TO 900

MAI N4 140

MAIN4150
MAIN4160
MAIN4170
MAIN4180
MAIN4190
MAIN4200
MAIN4210
MAIN4220
MAIN4230
GO TO 530
MAIN4240
CONTINUE
MAIN4250
CONTINUE
MAIN4260
GO TO 20
MAIN4270
CONTINUE
MAIN4280
WRITE (6,1340)
MAIN4290
STOP
**#*****************#********#**** ********!**************.****maIN^300
MAIN4310
MAIN4320
FORMAT (I5,3X,18A4)
MAIN4330
FORMAT (/1H1,10X,8HPR08LEM=, 15, 3H.. ,18A4////)
FORMAT
MAIN4340
1QX,16HINPUT QUANTITIES////)
PARAMETERS//)
MAIN4350
FORMAT (10X,38HINPUT TABLE
1A .. PROBLEM
FORMAT (1615)
MAIN436D
FORMAT (5X.39HNUMBER OF NODE POINTS
. . .=, I 5/5X,39HNU"MA I N4370
. . .= , 15/ 5X ,39HNUMBEP
1BER OF MATERIALS
OF TRACTION CMAIN4380
..MAIN4390
...=,I5/5X,39HCPTI0N FOR BODY FORCE =0 OR 1
2ARDS
I5/5X, 39HSEM I-MAIN4400
3.=,I5/5X,39H0PTI0N FOR PROBLEM TYPE
. . .=
OF OUTPUT TIME MAIN4410
4BAND WIDTH
. . .=
I 5/ 5X , 39HNUMBER
15)
MAIN4420
5LEVELS
MAIN4430
FORMAT (8E10.3)
FORMAT (////10X,36HINPUT TABLE IB. . . MATE R I AL PROPER T I S//5X ,55H MMAIN4440
1AT
K
RO/OEN OF WAT
MATDENS//)
MAIN4450
C OR MV
FORMAT
5X,I5,2X,E10.3,2X,E10.3,2X,E13.3,2X,E10.3)
MAIN4460
FORMAT (////10X,33HINPUT TABLE 2
NODAL POINT DATA//5X,40H NODE MAIN4470
DISP/FORCE/)
MAIN4480
KODE
Y-COCRD
FORMAT (2I5,2E10.3)
MAIN4490
MAIN4500
FORMAT (10X,19HERR0R IN NODE CAR0=,I5)
MAIN4510
FORMAT (5X,I5,2X,I5,2X,E10.3,2X,E12.3)
FORMAT (////10X,30HINPUT TABLE 3 .. ELEMENT DATA//5X, 40HEL NO N0MAIN4520
AREA/)
MAIN4530
NOOE J MTYPE
IDE I
MAIN4540
FORMAT (5X,21HERR0R IN ELEMENT CARD, 15)
MAIN4550
FORMAT (5X,I5,2I8,2X,I6,2X,E10.3)
FORMAT (////10X,35HINPUT TABLE
SURFACE TRACTI0NS//5X,26HNUMBMAIN4560
TRACTION/
MAIN4570
1ER
ELEM
FORMAT (I5,E10.3)
MAIN4580
FORMAT (5X,I6,2X,I5,2X,E10.3)
MAIN4590
FORMAT (////10X,51HINPUT TABLE
5A
DATA FOR TIME DEPENDENT PR0BMAIN4600
MAIN4610
1LEMS//)
FORMAT (2E10.3,I5)
MAIN4620
FORMAT (5X,15HTIME I NCREMENT=, ElO. 3,2X ,20HT0TAL SOLUTION T IME= , E10MAIN4630
1.3,5X,7H0PTI0N=,I5)
MAIN4640
FORMAT
////10X.44HI NPUT TABLE 5B,
DATA FOR OUTPUT TIME L EVELS//5MAI N4650
IX,26HNUMBER
OUTPUT TIME/)
MAIN4660
FORMAT (5X,I6,10X,E12.3)
MAIN4670
FORMAT (////10X.37HINPUT TABLE 5C. . INITIALS CONDITIONS //5X, 25HNOMAIN4680
10E
TEMP/PRES/)
MAIN4690
FORMAT (5X,I4,10X,E10.3)
MAIN4700
FORMAT (//1H1,10X,18H0UTPUT QUANTITIES)
MAIN4710
FORMAT (////10X,46H0UTPUT TABLE
I
.. STRESS-DEFORMATION PROBLEM /MAIN4720
MAIN4730
1/)
(

980
990
1C00
1010

1020
1030
1040
1050

1060
1070
1080
1090
1100
1110

1120
1130
1140
1150
1160

1170
1180
1190

151

Listing
1200
1210
1220
1230
1240
1250
1260
1270
1280
1290
1300

1310
1320
1330
1340

and Samples (cont.)

FORMAT
/5X,27H NODE
DISP LACEMEN T /)
MAIN4740
FORMAT
/5X,24HELEM
S TRESS)
MAIN4750
FORMAT /5X,27HELEM
VELOCITY/)
MAIN4760
FORMAT 5X, I4,10X,E10.3)
MAIN4770
FORMAT
////10X,32HCUTPUT TABLE 1
FLOW PROBLEM/)
MAIN4780
FORMAT
5X,I5,10XtE12.3I
MAIN4790
FORMAT
//10X, 14HELAPSED TIME =,E1 0.3//)
MAIN4800
FORMAT /5X.24HN0DE
POTE NTIAL)
MAIN4810
FORMAT
////10X,38H0UTPUT TABLE 1 . . TEMPERATURE PROBLEM//)
MAIN4820
FORMAT
/5X28H NODE
TE MPERATURE///)
MAIN4830
////10X,52H0UTPUT TABLE 1 .. RESULTS FOR CONSOLIOATICN PRMAIN4840
FORMAT
MAIN4850
10BLEM/
FORMAT
5X,13HELAPSED TIME=tE10.3, 2X.12HTIME FACTOR E10.3,2X. 24HDMAIN4860
MAIN487D
1EGREE OF CONSOLIDATIONS 10. 3/)
FORMAT
/10X,28H NODE
POR E PRESSURE//)
MAIN4880
MAIN4890
FORMAT
10X, I5,10X,E13.4)
/////17H ** JOB
FORMAT
MAIN4900
END **
MAIN4910
END
(

<

(
(

PROBLEM

1.

EXAMPLE 6-1/1-D STRESS DEFOR/POINT LOAD


AT TOP/ CONSTANT AREA
INPUT QUANTITIES

INPUT TABLE

PROBLEM

1A

PARAMETERS

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
OPTION FOR BODY FORCE =0 OR 1
CPTION FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS

INPUT TABLE

MAT

0.100E 04

INPUT TABLE

NODE

KOOE
1

B.

. .

MATERI AL PROPERTIES
OR MV

RO/DEN OF WAT

0.100E 01

0.100E 01

..

NODAL POINT DATA

Y-CCORD
O.OOOE 00
O.IOOE 02
0.200E 02
0.3 ODE 02

DISP/FORC!

O.OOOE 00
O.OOOE 00
O.OOOE 00

O.IOOE 02

MATDENS
O.OOOE 00

Problem
INPUT TABLE
EL NO

NODE

NODE

1.

(cont.)

..

ELEMENT DATA

MTYPE

AREA

O.IOOE 01
O.IOOE 01
O.IOOE 01

1
1
1

OUTPUT

QUANTITIES

OUTPUT TABLE

DISPLACEMENT

NODE

-O.OOOE
-O.IOOE
-0.200E
-0.300E

2
3

2
3

2.

00
00
00
00

STRESS
-O.IOOE 02
-O.IOOE 02
-O.IOOE 02

EL EM

PROBLEM

STRE SS-OEFO* MATI ON PRG5LEM

..

EXAMPLE

6-2/1-D

STRESS DEFOR/SURFACE

AND BODY FORCES/CONSTANT AREA


INPUT QUANTITIES

INPUT TABLE

PROBLEM

1A

PARAMETER S

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
OPTION FOR BODY FORCE =0 OR 1
OPTION FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS
VELS

INPUT TABLE

NODE

KODE
1

.=

. . .

. .

.=

RO/DEN OF WAT

O.IOOE 01

O.IOOE 01

. .

OR MV

O.IOOE 04

. .

=
=
=
.

.MATERI AL PROPERTIES

INPUT TABLE IB.


MAT

. . .

..

MATDENS

-0.500E 00

NODAL POINT DATA

Y-COOPD
O.OOOE 00
O.IOOE 02
0.200E 02
0.300E 02

DISP/FORCE
O.OOOE
O.OOOE
O.OOOE
O.OOOE

00
00
00
00

153

Problem
INPUT TABLE
EL NO

NODE

NODE

(cont.)

2.

..

ELEMENT DATA

MTYPE

INPUT TABLE

NUMBER

ELEM

OUTPUT

TRACTION

-O.IOOE 01
-O.IOOE 01
-O.IOOE 01

QUANTITIES

NODE

STRESS-DEFORMATION PROBLEM

DISPLACEMENT
-O.JOOE
-0.375E
-0.600E
-0.675E

ELEM

00
00
00
00

STRESS

0.375E
0.225E
0.750E

2
3

3.

O.IOOE 01
O.IOOE 01
O.IOOE 01

SURFACE TRACTIONS

..

OUTPUT TABLE

PROBLEM

AREA

EXAMPLE

02
02
01

6-2/1-D

STRESS DEFOR/SURFACE

AND BODY FORCES/LINEAR AREA


INPUT QUANTITIES

INPUT TABLE

1A

PROBLEM

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
CPTION FOR BODY FCRCE =0 OR 1
CPTICN FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS
154

PARAMETERS

Problem
INPUT TABLE IB.
MAT

C OR MV

RC/OEN OF WAT

0.1O0E 04

0.100E 01

0.100E 01

Y-COORD
O.OOOE
0.100E
0.2 00E
0.300E

INPUT TABLE

NOOE

NO

2
3

2
3

NUDE

INPUT TABLE

NUMBER

2
3

OUTPUT

2
3
4

ELEM

DISP/FORCE
O.OOOE
O.OOOE
O.OOOE
O.OOOE

00
02
02
02

..

ELEMENT DATA

MTYPF
1

4 ..

00
00
00
00

AREA

2
3

0.150E 01
O.IOOE 01
0.500E 00

SURFACE TRACTIONS

-O.IOOE 01
-O.IOOE 01
-O.IOOE 01

QUANTITIES

OUTPUT TABLE

NODE

-0.500E 00

TRACTION

ELEM
1

MATDENS

NODAL POINT DATA

..

KODE
1

EL

.MATERIAL PROPERTIES

INPUT TABLE

NODE

(cont.)

3.

STRESS-D6FCRMATICN PROdLFM

DISPLACEMENT
-O.OOOE
-0.242E
-0.442E
-0.567E

30

00
00
00

STRESS

-0.242 6

2
3

0.200E

02
02

-0.125E 02

155

PROBLEM

EXAMPLE

4.

STRESS DEFOR/SURFACE

6-2/1-D

AND BODY FORCES/ARBITRARY AREA


INPUT QUANTITIES

INPUT TABLE

1A

PROBLEM

..

NUMBER OF NOOE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
CPTION FOR BODY FORCE =0 OR
CPTICN FOP PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS

INPUT TABLE

MAT

O.IOOE 04

INPUT TABLE

NODE

B. ..

OR MV

RO/DEN OF WAT

0.100E 01

0.100E 01

..

MATERIAL PROPERTIES
C

NOCAL POINT DATA

Y-COORD

KODE

DISP/FCRCE

O.OOOE 00
O.IOOE 02
J.20DE 02

O.OOOE
O.OOOE
O.OOOE
O.OOOE

D.3 00E 02

INPUT TABLE

NODE

EL NO

2
3

NODE

NUMBER

ELEM

..

ELEMENT DATA

-ITYPtf

2
3
4

INPUT TABLE

..

00
00
00
00

AREA
1
1

0.150E 01
O.IOOE 01
0.500E 00

SURFACE TRACTIONS

TRACTION

-O.IOOE 01
-O.IOOE 01

-0,1 ODE 01

156

PARAMETERS

MATDENS

-0.500E 00

Problem

QUANTITIES

OUTPUT

OUTPUT TABLE

STRESS-DEFORMATION PROBLEM

..

DISPLACEMENT

NODE

-O.OOOE
-0.242F
-0.442E
-0.567E

2
3

00
00
00
00

STRESS

EL EM

-0 .242E 02
-0 .200E 02
-0 L25E 02

2
3

PROBLEM

(cont.)

4.

5.

EXAMPLE
AREA

6-3/1-D

STEADY FLOW/CONSTANT

INPUT QUANTITIES

INPUT TABLE

PROBLEM

1A

PARAMETERS

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
OPTION FOR BODY FORCE =0 OR 1
OPTION FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS

INPUT TABLE IB. ..MATERIAL PROPERTIES


MAT

0.1J0E 01

INPUT TA3LE

NODE

KODE
1

OP M V

RO/OEN OF wAT

0.100E

0.100E 01

..

01

MATOENS
J.OD.)E 00

NODAL POINT OATA

Y-COQRD
O.OOOE 00
O.iOOE 02
0.200E 02
0.3 00E 02

OISP/FORCE
0.200E
O.OOOE
O.OOOE
O.IOOE

01

00
00
01

157

Problem
INPUT TABLE
EL NO

NODE

2
3

OUTPUT

(cont.)

..

ELEMENT DATA

MTYPE

NODE

5.

2
3

QUANTITIES

FLOW PROBLEM

NODE

POTENTIAL
0.200E
0.167E
0.133E
0.100E

01
01
01
01

VELOCITY

ELE*

0.333F-01
0. 3336-01
0.333E-01

2
3

6.

0.100E 03
0.100E 03
0.100E 03

OUTPUT TABLE

PROBLEM

AREA

EXAMPLE5-2/CHAPTER5/HANDCALCULATION
INPUT QUANTITIES

INPUT TABLE

PROBLEM

1A

PARAMETERS

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OP TRACTION CARDS
OPTION FOR BODY FORCE =0 OR I
OPTION FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LF VELS

INPUT TABLE
MAT

INPUT TABLE

KODE
1

158

. .

4
1

..=

...=
...=

3
2

. . .

MATERI A L PROPERTIES
OR ^V

RO/DEN OF WAT

0.100E 01

0.100E 01

3.10JE 01

NODE

B.

...=
...=
...=

..

NODAL POINT DATA

Y-COORD
O.OOOE 00
0.100E 01
0.200E 01
0.30DE 01

DISP/FCRCF
0.100E
O.OOOF
O.OOOF
0.200E

02
00
00
02

XATDENS
D.OOOE 00

Problem
INPUT TA3LE
_

VCOc

',;

\EOE

(cont.)

..

ELEMENT DATA

"TYPE

APE/

O.IOOE 31

2
3

:.

<*

O.IOOE 01

INPUT T^bLr
time

6.

INCREMENT^ O.IOOE

^:tal

31

INPUT TABLE
'-.'-E

5A

PI-

I-E DEPENDENT

SOLUTION TINE= 0.3OOE

53..

DATA

OUTPUT

Lt^i.

..

iooe ci

=C^

11

==:~LEMS
DPTION-

OUTPUT TIME LEVELS

TINE

O.IOOE 31
D.200E 01
0.300E 01

2
3

5c.

input ta3le

initials conditions

tef/?pes

s^:e

O.OOOE 00
0.0 00= 00
I.OOOE 33
C.C3GE JO

2
3

OUTPUT

QUANTITIES

OUTPUT TAoLc

EL^SE.^
IOC

3
H

..

TIM.E

TEMPERATURE P^C?LE W

3.13GE

TEMPERAT

31

:
.

9. IOOE 02
3.5b3E 31
3. 331' 31
2.213? 32

159

Problem

(cont.)

6.

ELAPSED TIME
NODE

J.

2006 01

TEMPERATURE

0.100E
0.968E
0.129E
0.200E

ELAPSED TIME

02
01
02
02

0.300E 01

TEMPERATURE

NODE

0.

IOOE

02

0.117E 02
D.150E 02
0.2O0E 02

PROBLEM

EXAMPLE 6-4/TRANSIENT HEAT FLOW/CON


STANT AREA AND INITIAL TEMPERATURE

7.

INPUT QUANTITIES

INPUT TABLE

PROBLEM

1A

PARAMETERS

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
CPTION FOR BODY FORCE =0 OR 1
OPTION FOP PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS

INPUT TABLE IB.


MAT

O.IOOE 01

INPUT TABLE

NODE

KODE

2
3

4
5

6
7
8
9

10
11

160

. .

3
2

12

MATERI AL PROPERTIES
OR MV

RO/DEN OF WAT

O.iOOE 01

O.IOOE 01

11

..

NODAL POINT DATA

Y-COORO
O.OOOE
O.IOOE
0.200E
0.300E
0.400E
0.500E
0.600E
0.700E
0.800E
0.900E
O.IOOE

00
01
01
01
01
01
01
01

01
01
02

DISP/FORCE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.COOF
O.OOOF
O.OOOE
O.OOCE
O.OOOE

00
00
00
0)
00
00
00
00
00
00
00

NATDENS

O.OOOF 00

Problem
INUT TABLE

NODE

NO

EL

NODE

7.

(cont/

..

ELEMENT TATA

MTYPE

AP5A

2
3

&

7
8

a
9

10

INPJT TA3LE

THE INCREMENT^

.1
]

10
11

5A

..

TOTAL

O.IOOE 3C

J.100E 01
O.IOOE 01

>1

0.100E 01
O.IOOE 01
0.100E 01
0. 100E

01

O.IOOE 3 1
O.IOOE 01
O.IOOE 01

DATA FC^ TIME DEPENDENT PROBLEMS

SOLUTION TI*E =

30 OE

03

OPTION:

INPUT TA5LE 53.. DATA FO^ OUTPUT TINE LEVELS

NUMBER

OUTPUT

TIME
.

j
3

100E On
50DE 00
ICOE

01

200E 01
300E 01
.400E CI
600E 01
OOE
30E

01
02

.150E

02
02

3
1

200E

0.300E 02

INITIALS CONDITIONS

INPUT TA?1
NCOS

TEMP/?" =S

3.10JE

0.1 DOE

13

O.IOOE

03
33

10

O.IOOE 03
O.IOOE 13
O.IOOE 03
O.ICDE 03
O.IOOE 13
O.IOOE B3

11

0.1 DOE

0. 133E

6
7
8

03

03

161

Problem

7.

OUTPUT

QUANTITIES

(cont.)

..

ELAPSED TIME

O.IOOE 00

TEMPERATURE

NODE

-0.030E
0.884E
0.101E
0.999F

00

O.IOOE
O.IOOE
0.999E
0.101E
0.8846
O.OOOF

33
03
02
03
02
00

02
03
02
0. 100E 03

2
3

4
5

6
7

8
9

10
11

ELAPSEO TIME
NODE

2
3

4
5

6
7

8
9

10
11

ELAPSEO TIME
NODE

2
3

4
5

0.500E 00

EMP=RATURE

-0.030E
0.633E
0.935E
0.99RE
O.IOOE

03

3.1005
3.133E
0.993E
0.935E
3.633E
0.0 DOE

03
03
02
02
02
33

03
OZ
02
32

3.103E 31

TEMPERATURE

-3. 33JE

J 3

0.49OE 32
0.822E j
J.963E 32
0.995E 02
P.

100E

0.

7
8
9

0.99 5E
3.960E
0.822E
0.*96E
O.OOOE

13
11

162

TEMPERATURE PROBLEM

OUTPUT TA3L5

33
02
32
32
J?.

:o

Problem

7.

EL-^SEO TIME

(cont.)

0.200E 31

TEMPERATURE

MODE

0.000E
3.3725
0.669E
0.3575
3.946E
3.974E
3.948E
0.857E
3.669E
0.3725
O.OOOE

ELADSEO TIME

6
7

8
9
]

11

33E

0.300E OL

3.576E
0.767E
0.877E
0.913E
0.877E
0.767E
0.576E
0.311E
O.OOOE

02
32
02
00

-O.OOOE 30
0.311E 02

ELAPSEO TIME

32
32

TEMPERATURE

N00 E

3J
02
02
02
02

32
02
32

02
02

02
32

02

00

O.OOOE 01

TEMPERATURE

-0.333E 00
J.271E 32
0.539E 32
3.691E 02
0.3025 32
0.

9
10

11

*0E

0.302E
0.6 9 IE
3.5395
3.271E
O.OOOE

02
32

02
02
32

33

163

Problem
ELAPSEO TIM E

7.

NODE

-0.000E
0.217E
0.411E
0.564F
0.662E
0.695C
0.662E
0.564E
0.411E
0.217E
0.000E

*
5

6
7

10
11

02

02
02
02

02
02
02

02
00

0.800E 01

-O.OOOE 00

0. 177E
0.336E
0.463E
0.544E
0.572E
0.544E
0.463E
'J. 336c
0.177E
O.OOOE

2
3

4
5

6
7

10
11

02
02
02
02
02
02
02
02
02
00

O.iOOE 02

NODE

TEMPE^ATUR

-O.OOOE 00

2
3

4
5

6
7

8
9

10
11

164

00
02

TEMPERATURE

NODE

ELAPSEO TIME

0.600E 01

TEMPERATURE

2
3

ELAPSED TIME

(cont),

3.145E
0.2 76E
0. 330E
J.446E
0.469E
0.446E
0.380E
0.276E
0.145F
O.OOOE

02
02

02
02
02
02
02
02
02

00

Problem

(cont.)

7.

ELAPSED TIME

3.150E 02

TEMPERATURE

NODE

O.OOOE 00
0.383E 01

2
3
4

0. 168

0.231E
0.272E
0.286E
0.272E
0.2315
0.163E
0.883E
O.OOOE

6
7
8
9

10
11

PROBLEM

oz
02
02
02
01

01
02
01
JO

EXAMPLE 5-4/CONSOLIDATION OF LAYERED


MEDIA

8.

INPUT QUANTITIES

NPUT TABLE

PROBLEM

1A

NUMBER OF NODE POINTS


NUMBER OF MATERIALS
NUMBER OF TRACTION CARDS
CPTION FOR BODY FORCE =0 OR
OPTION FOR PROBLEM TYPE
SEMI-BAND WIDTH
NUMBER OF OUTPUT TIME LEVELS

0.400E-01
0.200E 00
0.50QE-02
0.100E 00

2
3

INPUT TABLE

NODE

KODEE

3
4
5

6
7
8

10

21
4
1

4
2
8

MATERIAL PROPERTIES

INPUT TABLE V
MAT

PARAMETERS

OR MV

0.100E-02
0.100E-01
0.10GE-02
0.100E-0L

kO/DEN

OF

0.100E
0.100E
0.100F
0.100E

WAT

MAT DENS

01
01

O.OOOE on
O.OOOE 00

01
01

O.OOOE 00

0.000=

00

NODAL POINT DATA

Y-CCORD
O.OOOE
0.400E
0.800E
0.120E
0.160E

00
01
31

02
02

0.2 00E 02
0.2 40E 02

0.280E 02
0.320E 02
0.360E 02

DISP/FORCE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE

00
00
00
00
00
00
00
00
00
00

165

Problem

0.400E
0.440E
0.480E
0.520E
0.560E
0.600E
0.640E
0.680E
0.720E
0.760E
0.800E

11
12

13
14
15
16
17
18
19

20
21

INPUT TABLE

NODE

NO

NODE

2
3

3
4

7
8

7
8

9
10
11
12

10
11
12
13
14

AREA

2
2
2

11
12

13
14
15

15
16
17
18
19

13
14
15
16
17
18
19

20

20

20

21

<

TIME INCREMENT= 0.500E 00

4
5

6
7

3
3

16
17
18
19

01

4
4
4

..

O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
0.100P
O.IOOE
O.IOOE

01
01
01
01

01
01
01
01

01
01
01
01
01
01
01

DATA FOR TIKE DEPENDENT PROBLEMS

TOTAL SOLUTION TIME= 0.200E 03

INPUT TABLE 5B.

NUMBER

01
01

01
0. 100E 01

10

5A

00
00
00
00
00
00
00
00
00
00
00

O.IOOE
0.100E
O.IOOE
O.IOOE

5
6
7
8
9

INPUT TABLE

166

ELEMENT DATA
MTYPE

32
02
02
02
02
02
02
02

..

0.000E
0.000E
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE
O.OOOE

02
02
02

2
3

(cont.)

8.

OUTPUT

DATA

=0R

TIME

0.5 00E
0.650E
0.280E
0.600E
0.985E
O.llOE
0.150E
0.200E

00
01
02
02
02
03
03
03

OPTION=

OUTPUT TIME LEVEL!

Problem
[NPUT TABLE

5C.

8.

INITIALS CONDITIONS

TEMP/PRES

DE

2
3
4

0.100E 01
O.IOOE 01
0.100E 01
O.IOOE 01

0.1 JOE 01

O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE
O.IOOE

7
3

10
11

12
13
14
15

16
17
18
19

20
21

OUTPUT

NODE
1

2
3

4
5

6
7

8
9

10
11
12

13
14
15
16
17
18
19

20
21

ELAPSED TIME= 0.650E 01

01
01

01
01
01
01
01
01
01
01
01
01
01
01

01
01

QUANTITIES

OUTPUT TABLE
ELAPSED TIME= 0.500E 00

(cont.)

..

RESULTS FOR CONSOL OAT ON PROBLEM


I

TIME FACTOR= 0.695E-02

DEGR.FE

OF

CONSOLIDATION 0.1*4E 00

PORE PRESSURE

-O.OOOOE
0.5435E
0.8194E
0.9286E
0.9718E
0.9890E
0.99616
0.9996E
0.9999E
O.iOOOE
0.1000E
O.IOOOE
O.IOOOE
O.IOOOE
O.IOOOE
0.9999E
0.9996E
0.9968E
0.9720E
0.7549E
O.OOOOE

00
00
00
00
00
00
00
00
00
01
01
01

01
01
01
00
00

00
00
00
00

TIME FACTQR= 0.904E-01

DECREE OF CONSOLIDATION 0.335E 00


167

Problem
NODE
1

2
3

4
5

6
7
8
9

10
11
12

13
14
15
16
17
18
19

20
21

ELAPSED TIM= 0.280E 02

NODE

2
3

4
5

6
7
8

9
10
11

12
13
14
15
16
17
18
19

20
21

ELAPSEO TIME= 0.600E 02

NODE
1

168

8.

(cont.)

PORE PRESSURE

-O.00O0E 00
0.1533E GO
0.3 033E 00
00
00
00
00

0.4474E
0.5836E
0.7116E
0.8320E
0.9465E
0.9655E
0.9734E
0.9865E
0.9909E
0.9924F
0.9857E
0.9503E
3.8463E
0.8040^
0.6923E
0.5118?
0.273 OF
0.0000E

00
00
00
00
00
00
00
00
00
00
00
00
00

TIMF F<\CTOk
FACTORS= 0.3895 00

DECREE OF CONSOLIDATION

0. 539E

00

PORE PRESSURE

-J.OOOOF jO
0.1182E UO
0.2360F
0.3532E JJ
0.4693E 00
0.5840E no
3.6970E 31
0.8080E 00
0.8275E
0.8421E 00
0.8516E 00
0.8560E
0.8550E 00
0.6973E 00
0.5071E 00
0.2768E 00
0.2538E 00
0.2108E 00
0.1509E 00
0.7867E- -01
O.OOOOE 00
TIME FACTOR= 0.834E 00

PORE PRESSURE

-O.OOOOE 00
0.9214E-01

DEGREE OF CONSCL IDAT ION= 0.672E 00

Problem

8.

(cont.)

0.1840E 00
0.2752E 00
0.3656E 00
0.4548E 00
0.5425E 00
0.6284E 00
0.6434E 00
0.6542E 00
0.6607E 00
0.6629E 00
0.6608E 00
0.4828E 00
0.2902E 00
0.8632E-01
0.7420E-01
0.5837^-01
0.4087E-01
0.2094E-01
O.OOOOE 00

3
4
5

6
7
8

10
11
12
13

14
15
16
17
18
19
20
21

ELAPSFD TTME= 0.985E 02

NODE
1

2
3

4
5

6
7
e

9
10
11
12
13
14
lb
16
17

18
19
20
21

ELAPSFD TI^E= 0.110E 03

NODE
1

2
3
4
5

6
7

FflCTQK= 0.137E 01
TIME FflCTCK=

DFGRFE n F fONSFt. IDAT ICN= 0.765E 00

PORE PRESSURE

O.OOOOE 00
0.6731E- -01
CO
00

0.1344E
0.2011E
0.2670F
0.332 2F
0.3962E
0.4539E
0.4698F
0.4776E
0.4822F
0.433 7F
0.4820E
0.3446F
0.1979E

oc
>0

00
oc

00
00

00

JO
0.4575E- -01
0.37A>8E- -01
0.2ti91E- -01
0.1959E- -01
0.9 894E- -02

O.OCOOF
TIME FACTQR= 0.153^ 01

DEGREE OF CC*SOL IOAT

IC1N=

0. 78hE

PORE PRESSURE

-O.OOOOE 00
0.6125E-01
0.1223E 00
0.1629E 00
0.2430E 00
0.3022E 00
0.3605E 00
169

00

Problem

9
10
11
12
13
14
15
16
17
18
19

20
21

ELAPS C D TIMS= 0.150E

33

2
3

4
5

6
7
8

10
11
12

13
14
15
16
17
18
19

20
21

0.4175E
0.4274F
0.4345E
0.4308E
0.4401E
0.4385E
0.3130F
0.1793E

00
00
00
00

00
00
00

0.4060E- -01
0.3330E- -01
0.2547E- -01
3.1722F- -01
0.8684E- -02
0.0000E 00

2
3

-O.OOOOE 00
0.4439E-01
0.8803E-01
3.1317E 00
0.1749F 00
0.2175E 00
0.2595E 00
0.3005E 00
0.3077E 00
0.3128F 00
0.3158E 00
0.3167E 00
0.3156F 00
0.2250F UO
0.1284E 00
0.2847F-01
0.2326E-01
0.1772E-01
J.1195F-01
0.6016E-02
O.OOOOE 00

-O.OOOOE 00
3.2922E-01
0.5 83 5E-01
0.8729F-01

0.U59E
0.1442F
u,1720 c
0.1992E
J.2J39E
0.2073E
0.2093E
0.210GE

7
8

10
11
12

DEGREE OF

PORE PRESSURE

DEGREE OP CONSOLIDATION* 0.846E 00

PORE PRESSURE

TIME FACTOR= 0.278E 01

NODE

170

(cont.)

TIME F4CT0R= 0.2J9F 31

NUDE

ELAPSEO TIME= 0.200E 03

8.

00
00
00
00
00
00
00
JO

CNSCL IOAT ION= 0.698E 00

Problem 8 (cont.)
13

14
15
16
17
18
19

20
21

** JOB

0.2092E 00
0.1491E 00
0.3507E-01
0.1882E-01
0.1536E-01
0.1170E-01
0.7887E-02
0.3969E-02
O.OOOOE 00

END **

171

BEAM BENDING
AND BEAM-COLUMN

INTRODUCTION
Problems of beam bending and beam-column analysis using one-dimensional
shows a beamcolumn subjected to the transverse load p(x) and axial load P. We first treat
the case of bending only, without the load P. Under the usual assumptions of
beam bending theory [1], the governing differential equation can be written as
idealization are considered in this chapter. Figure 7- 1(a)

dx 1

(fW^)=

/J

(4

(7-la)

= transverse displacement; F(x) EI(x), flexural rigidity; and


= coordinate along the centroidal axis of the beam. If F(x) assumed

where w*

is

uniform along the beam, Eq. (7-la) specializes to


(7-lb)

p(x).
dx'

Here we have used z

as the vertical coordinate [Fig. 7- 1(a)]

and w

as the

corresponding transverse displacement.

We now
Step

1.

follow the various steps of

element formulation.

Discretize and Choose Element Configuration

With the one-dimensional


[Fig. 7- 1(b)],

172

finite

with the rigidity

idealization the

Flumped

beam can be

at the line.

The

replaced by a line

idealized

beam

is

now


Beam Bending and Beam-Column

Chapter 7

Cross

173

pT

section

(a)

fTTp( x

>

-*s
(b)

(c)

hv A

^.M,
u,.P

w i,Qi

,M 2

w 2 ,Q 9
(d)

Beam bending and beam-column,


and axial loads, (b) One-dimensional
cretized beam, (d) Generic element.

Beam

with trans-

Figure 7-1

(a)

verse

idealization, (c) Dis-

discretized into one-dimensional line elements [Fig. 7-l(c)].


is

shown
Step

A generic element

in Fig. 7- 1(d).

2.

Choose Approximation Model

In the case of (one-dimensional) column deformation (Chapter


dealt only with plane deformations.

continuity of such structures,

it

To

3),

satisfy the physical condition

was necessary

to satisfy interelement

we
of

com-

Beam Bending and Beam-Column

174

fulfill

problem by using only

[Fig. 7-2(a)]. It was


and mathematical requirements of the

nodal displacement

patibility at least with respect to

therefore possible to

Chapter 7

the physical

first-order (linear)

approximation. In contrast to the

plane deformations, for realistic approximation of the physical conditions in


the case of bending,

it is

necessary to satisfy interelement compatibility with

respect to both the displacements

As

of displacement [Fig. 7-2(b)].

and

slopes, that

a consequence,

is, first
it

derivative (gradient)

becomes necessary

to use

higher-order approximation for the displacement in the case of bending.


Since

it is

necessary to provide for interelement compatibility for slopes also,

we can add

slope at the

node

as

an additional unknown. This leads to two


vv and slope 6 = dw/dx, at each node

primary unknowns, displacement

hence, for an element there are a total of four degrees of freedom:

node

and

vv 2

at

node 2

wu

at

[Fig. 7-l(d)].

0A00A0

Local nodes
Elements

Subscript^
SuDerscript

node
element

local
=

A
(a)

(b)

Figure 7-2 Requirements of interelement compatibility, (a) Inter-

element compatibility for axial deformation (Chapter

element compatibility for

beam

3). (b)

Inter-

bending.

The commonly used

interpolation approximation model for displacement


any point s
x/l
(x
*i)//, where s = local coordinate, x = global
coordinate of any point, x = global coordinate of point 1, and / = length of
the element, is given by

at

w (x)

=Nw +N6 +Nw +

w(x)

N,0 2

(7-2a)

or
[N]{q}.

(7-2b)

Chapter 7

Here{q} r =[w! 6

polation functions

Beam Bending and Beam-Column

175

= [N N N N

the matrix of inter-

w2

t,

9 2 ] and [N]

4 ] is

1, 2, 3, 4,

N = - 3s + 2s
N = /s(l -2s + s
N = s (3 - 2s),
2

3
,

2
),

(7-3)

Nt
In Fig. 7-3,

we have shown

literature the functions

ls (s

1).

plots of

i9

1, 2, 3, 4.

are called Hermitian functions

In mathematical

[2, 3].

-^_N

5^*

'3

N4

_^-

Figure 7-3 Plots of

It is

worthwhile here to

it

1, 2, 3, 4,

illustrate that the

[Eq. (7-3)].

matrix [N] in Eq. (7-2) can be

derived by following the procedure outlined in Chapter

with polynomial functions. Here


w(x)

where

a,

= generalized

matrix notation as

= j

-f

we can
cc 2

3,

where we started

use the following cubic polynomial:

<x 3

x2

a 4 x-

(7-4)

coordinates. Equation (7-4) can be expressed in

Beam Bending and Beam-Column

176

w (x )

X2

[1

Differentiation of w(x) with respect to

^=Q =a

Chapter 7

X2]

(7-5)

leads to

2a 3 x

= [0
2x
= [*']{}.

3a 4 x 2

3x 2 ]{a}

(7-6)

Here the prime denotes derivative with respect to x. To express w(x) in


term of nodal values of w and 0, we use Eqs. (7-5) and (7-6) to first evaluate
their values at nodes 1 and 2 as
'

Wl (x

"10

= oy

*i(*=0)
w 2 (x = I)

[q]

0"

<x' 2

(7-7)

<

M* = 0.

_0

2/

a3

3/ 2 _

.4

or

{q}

(7-8a)

4]{a}.

Therefore,

{a}=[A]-'{q}.

(7-8b)

Finally, substitution of {a} into Eq. (7-5) leads to

w(X)

= [*][A]-'{q},

(7-9)

where

[A]

-3// 2

L
If

we perform

2//

the multiplication

Comment on Requirements

for

-2//
l//

3//
2

_1
[<J>][A]

2//

-1//

l//

we obtain

[N] in Eq. (7-2).

Approximation Function

The approximation function [Eq. (7-2)] is conformable since it provides


up to n 1 = 2 1 = 1 derivative of w,
that is, for both w and its first derivative [Fig. 7-2(b)], where n = 2 is the

for interelement compatibility

highest order of derivative in the potential energy function in Eq. (7- 14a)

below.

It

also satisfies the completeness criterion since

it

allows for rigid

Beam Bending and Beam-Column

Chapter 7

body motion because of


(slope), Eq. (7-6).

is

a, in Eq. (7-4)

177

and for constant

states of strain

Notice that the complete expansion [Eq. (3-3d)] up to

retained in Eq. (7-4).

PHYSICAL MODELS
It

could be very instructive to the student to explain some of the principles in

the finite element formulation by using simple models.

For

instance, the

concept of interelement compatibility provided by approximation functions


of different orders can be illustrated by preparing models of a beam. Details

of construction and use of such models are given in Appendix


Step

3.

3.

Define Strain-Displacement

and Stress-Strain Relationships

From beam bending theory

[1],

the relevant strain-displacement relation

d w
du _ ~
au
_ z a*\v
~ zw
= =
-izi =
dx
dx
2

(*> z )

where u

axial displacement

derivative. In

fn " im

r,

(7

is

1() )

and the superscript (two) primes denote second

bending problems the constitutive law

commonly

is

expressed

through the moment-curvature relationship

M[x)

= F(x}w"(x).

(7-11)

Differentiation of w(x) in Eq. (7-2) twice with respect to x, through use of


the chain rule of differentiation,

d
d
= -T7Tr
^
dx
ds
1

A
and

d1

d2

v
(1 " 10
12

^2=72-^2'
dx 2
ds
I

(7

leads to

"

where

[B]

/ v
(x)

d2W

= S? =
=

= transformation

differentiation.

d2W

d1

nvT-ir

/-,

-,

F SP = Frf? [N]{q}

-j

(7-13b)

[B]{q},

matrix;

(7 " 13a)

'

its

coefficients are obtained

by proper

For example,

= -j-[-6s -

6s 2

1(1

-4s+

3s 2 )

6s

l(3s

9,
2s)]

(7- 13c)

Beam Bending and Beam-Column

178

Chapter 7

and

d2w
dx 2

_}_d_(dw\
I

ds\ds)

JL[_6

-4/

\2s

6/j

12s

61s

21]

(7- 13d)

1*2

Step

4.

Derive Element Equations

ENERGY APPROACH

We now
derive

use the principle of stationary (minimum) potential energy to

element equations. The potential energy for the

assuming only surface traction loading p(x),

Up =
Substitution of w and
II,

where

F is

[*'

w" from

= \Fl JoV
=

Differentiating

IT^,

pw

element,

(7- 14a)

dx.

Eqs. (7-2) and (7-13d) into Eq. (7-14a) leads to

{q}W[B]{q}<fe

-/f

'

(7-14b)

\N\p{q}ds,

^0

assumed uniform, and from

the transformation dx

beam
[4, 5]

Xi

%F(w") 2 dx

expressed as

is

(x

x^/l,

we have

substituted

Ids.

w l9

with respect to

l9

u> 2 ,

and

term to zero, we obtain four element equations expressed

and equating each


in

matrix notation

as
1

Fl\ [BY[B]ds{q}=l\ [Wpds

(7-15a)

Jo

Jo

or

M{q}

(7-1 5b)

{Q},

where
\k]

= Fl C

[BY[B]ds

Ja

and
{Q}

which

[k]

in the

= F/f
Jo

expanded form
6(25 2(3* 6(2* 2(3* -

'

f
Jo

[NT/&,

is

\)ll*\
2)//
l)// 2
1)//

After the integrations,

p^JJ

2(^-2)

-6(2,-!)

ffi^iyj^

(? . 16a)

Beam Bending and Beam-Column

Chapter 7

ri2

[k] = ~~

6/

-12

6/

4/ 2

-6/

2/

12

-6/

sym.

179

(7- 16b)

4/ 2

and, assuming the surface traction

p
where

/7j

=(1 -

varies linearly as

^s/? 2

5)/?!

(7-1 6c)

are values of loading at nodes

and p 2

-3s - 2s
ls(l -2s + s
2

(I

and

2, respectively,

3
)

{Q}

= /f

1-5]

[S

- 2s)
(s - 1)

(7-16d)

s (3
ls

J>2

or
7/?,

3/? 2 \

^-(3/7,

-f-

2/7 2 )

(7-16e)

=2i<

-^-(2^

3/> 2 )|

DERIVATION USING GALERKIN'S METHOD


On the basis of the explanation in Chapter
element.

The

residual

is

we consider one

3,

generic

given by
(7-17)

Therefore, use of the Galerkin

S"

method

p Ntdx

Integration by parts of the

gives

/=1

'

2 3 4
'

(7-18)

'

first

term leads to

d*w dN,
dx 2 dx

w from

Substitution of

0.

(7- 19a)

Eq. (7-2) leads to

<**-**+'*

,d

w dN

dx 2 dx

= 0,

i=l,2,3,4, y=l,2,3,4

(7-19b)

or

I,

</x

</x

"**

Jxi

^'^

' </x 3

dxl dx

(7- 19c)

Beam Bending and Beam-Column

180

Chapter 7

Expansion of the term on the left-hand side leads to


~N'[ 2

N'i
J X\

Nim
N'm

N'iN'(
1

N'iNT

N'iN'l

aw

N?

sym.

w,

N'i

(7-20a)

>dx.

w2

e2

Here the double prime indicates second derivative with respect to


proper transformations for differentiations and integrations,
will yield the

same

x. After

this expression

results in Eq. (7-1 6b).

term on the right-hand-side will lead to the load vector in


Eq. (7-1 6e). The remaining two terms on the right-hand side yield (internal)
joint shear and moment forces. Let us consider

The

first

(7-20b)

iVj

dx'

Noting the properties of

we

(Fig. 7-3),

find that this reduces to

(f)\

(7-20c)

(4s)
which gives "joint" shear forces

two nodes.

at the
2

rd w

dNA

Similarly,

X2

dx 2 dx
leads to

K
(

dx 2 dx)\ xi

Fd

wdN \\ x
2

dx 2 dx)\ x

(F

d 2 wdN 3 \\ x
dx 2 dx)\ x

V
/

2
f d w dNA

dx 2 dx

Xi
\

)\ xu

(7-20d)

Beam Bending and Beam-Column

Chapter 7

181

or

m.
X

dx>) 2

(7-20e)

because

dx

dN2
dx

and x 2)

at x,

dx

dN4 _,

at

dx

x and x 2
x

and
at x>

dx

dN,_
dx
which

yield "joint"

there were
(7-20c)

and

no

moment

Sit

xu

two nodes.

forces at the

when

externally applied joint forces,

(7-20e)] are

assembled for

all

It

may

be noted that

if

joint load vectors [Eqs.

elements in the discretized body,

only the terms related to the two ends of the entire

beam

will remain,

and minus

the other terms will be zero due to alternating plus

whereas

signs.

The

nonzero end terms will denote the boundary conditions; for instance, for a
simply supported beam the end term in Eq. (7-20e), which denotes moment,
will vanish. Note that the element equations from both the energy and
residual procedures will be essentially the same. Futher details of applications

of the Galerkin and other residual methods are given in Appendix

1.

Steps 5 to 8

We

shall illustrate the steps

of assembly and computation of primary and

secondary quantities by using an example. Figure 7-4 shows a


length

20

cm and cross-sectional

area

= 2 cm

cm deep x

(2

beam with
1

cm wide)

kg/cm 2 It is subjected to a uniform surface traction p(x) =


100 kg/cm. The beam is divided into two elements of length / = 10 cm each.
Use of Eqs. (7- 16b) and (7-16e) leads to the following (local) stiffness

and

E=

10 6

relationship for both elements (with /


3

16

15

15-3
100

-15

cm 4 ):

15"

50

'

0,
<

-3

-15

15

50

-15

-15

w2

100_

e2

3'

1000

(7-21)

"

,-5,

Beam Bending and Beam-Column

182

p(x) = 100

Chapter 7

kg/cm

M Mi M tn
L

Ut*

L = 20
1

e
E . 10 kg/cm:

cm

cm

Global

Local

10cm

O
A

10cm

Nodes
Elements
Figure 7-4 Example for

By following

beam

bending.

the assembly procedure as described in Chapter

3,

we

obtain

the following global stiffness relation for the two elements:


3

15

15

100

-3

-15

15

50

15

-15

50

15

The beam

(Fig. 7-4)

3]

0i

15

w2

-15

50

-15

50

-15

(7-22a)

-15 I*
Wy

kJ

100_

-5j

supported at two unyielding ends; therefore, the

is

boundary conditions are w

(hV

200

16

0"

= w = 0. Introduction of these constraints into


3

Eq. (7-22a) leads to modified global equations as

0"

(w>r
f

15

100

-3

-15

15

50

-15

50

16

15

-15

50

w2
1

200

15

50

(7-22b)

<

r
W>3

-15

1
5

0i

100_

kJ

,-5J

Beam Bending and Beam-Column

Chapter 7

183

Solution of the four simultaneous equations gives

w,
0j

= 0.0000 (given),
= 0.0500 radian,

=
=

w2
2

0.3125 cm,

w3

0.0000 rad.,

0,

CLOSED FORM SOLUTIONS


From beam bending theory
w(x)

dw_

dx
which

yield the

same

computations. That

method

yields the

[1]

= 0.0000 (given),
= -0.0500 rad.

the closed form or "exact" solutions are

P x (T3
{L
24F

2Lx 2

p (fZ
24F K

6x 2

x 3 ),

(7-23)

4x 3 ),

(7-24)

results as those in Eq. (7-22c)

from the

with the cubic approximation model, the

is,

same

(7-22c)

results as the closed

finite

element

finite

element

form solutions insofar as the

displacements and slopes are concerned.

SECONDARY QUANTITIES
Let us now consider the secondary quantities: moments (M) and shear
forces (V). To find moment
we substitute relevant nodal displacements

into

d2w
dx 2

M
where d 2 w/dx 2 is defined
For element 7,

(7-25)

in Eq. (7- 13d).

0.0000

M(at$

0)

0.0500

-[-6

4/

-21],

>

0.3125
.0.0000

-' 25F

0.08334

\l

833.4 kg-cm

lilarly,

M(at

M
For element

(at s

0.5)

1)

- -0/250F -

15F

~f
2

-3333.3 kg-cm,

-5833.4 kg-cm.

2,

0.3125

M(ats

0)

= -^[-6

-4/

0.0000
6

-21}.

0.0000

-0.0500,

-5833.4 kg-cm.

Beam Bending
I
and Beam-Column

184

Chapter 7

Similarly,

M(2LtS

-0.05F

0.5)

M(aXs

0A25F
1)
I

The closed form

M=
which

at

//4

solution for

moment

3333.3 kg-cm,

-833.4 kg-cm.

is [1]

\2p
(-Lx
24

-Fw'Xx)

and x

x 2 ),

(7-26)

1/2 gives

m(x =

-j\

M (x = -L\ =
Figure 7-5(a) shows the bending

-3750.0 kg-cm,

-5000 kg-cm.

moment diagrams from

the finite element

computations and from the closed form solution.


Figure 7-5 Comparisons for bending

two elements,

(a)

moments and shear

Bending moment diagrams,

(b)

forces:

Shear force dia-

grams.

~j^ 833.4

Finite element solution


(a)

Closed form solution

500 kg

Finite element solution

(b)

Beam Bending and Beam-Column

Chapter 7

Next we compute shear force

is

given in closed form as


(7-27)

yields

V (x

^=

0)

= y) =
V (x = L) =
(*

From

which

d3w
dx 3

V
which

V,

185

-1000.0

kg,

0.0kg,
1000 kg.

the finite element computations,

dx

^-[12

Therefore, for element

ds

6/

-12

(7-28)

6/]

/,

[0.00001

K=-^[12

6/

-12

0.0500
500.25 kg,

6/]

0.3125

0.0000

and for element

2,

= 0J5F = + 500.25 kg.

Figure 7-5(b) shows plots of shear force from closed form and

finite

element

computations.

Comments. As noted previously, both methods

yield the

displacements and slopes. However, for moments, the

finite

same values of
element method

whereas the closed form solution


shows continuous distribution. Also, introduction of only Wj
in
wthe finite element procedure does not yield zero moments at the ends, which
is required for the assumption of simple support in the closed form solution.
Overall, the magnitude of the moments from the finite element computations

yields a bilinear distribution [Fig. 7-5(a)],

is

satisfactory.

In the case of the shear forces, however, the

finite

element computations

show a wide disparity as compared with the closed form results [Fig. 7-5(b)].
At the ends the difference between the two results is high.
As discussed in Chapter 3, there are two possible methods by which we can
improve computations of bending moments and shear forces: (1) refine mesh
and/or

(2)

choose different (higher-order) approximation models.

MESH REFINEMENT
The beam is now

divided into four elements of

Fig. 7-6. Substitution of E,

/,

/,

cm, as shown in

and/? into Eqs. (7-16b) and (7-16e) leads to

the following general element equation

30

12

10

125

100

X 2
x 3

-12
-30

30"

50

100_

20

cm

Figure 7-6

0i

w2

250

kJ

-2500/12,

cm

cm

Mesh

25
f

2500/12

-30

12

ym

cm

refinement for

beam

cm

bending.

Assembly of the four element equations and introduction of the boundary


conditions w,

lead to the followi ng modifie< J ass embk igee qui itions:

"10

0~

30

100

-30

-12

-30

24

30

50

50

-12

30

200

-30

50

-12

-30

24

30

50

-12

30

-30

50

0.00\

<Wi^
0i

208.33

w2

500.00

02

0.00

W3

500.00

333.33

200

-12

-30

30

50

200

-12

30

-30

50

50

-30

w4

500.00

04

0.00

W5

30

0.00

24

100_

,051

0.00
,

-208.33y

Solution of these equations gives

Wi
vv

e,

= 0.0000,
= 0.0000,
= 0.0500,
= 0.0500.

w2

0.2227,

H> 3

0.3125,

H> 4

0.2227,

0.0344,

03

0.0000,

-0.0344,

Figure 7-7 shows comparisons between bending

moments and shear

forces

from the finite element analysis and those from the closed form solution. It

186

Beam Bending and Beam-Column

Chapter 7

cm

cm

187

cm

cm

208-^

Closed form
Finite element
a

243

(b)

Figure 7-7 Comparisons for bending

moment

four elements, (a) Bending

moments and

shear forces:

diagrams, (b) Shear force dia-

grams.

can be seen that the computed values are


than the comparisons

now

closer to the exact solutions

two-element approximation.

in Fig. 7-5 for the

HIGHER-ORDER APPROXIMATION
As the next step from the cubic approximation, we can adopt

a fifth-order

approximation as follows:
w(x)

=
=

ol

[1

a2x

X2

a3
A'

.\-

J
cc X
B
4x

a 5 x4

a 6x5

(7-29)

where
{a} 7

The

first

and second

"

[a,

%2

%4

a.

a.]-

derivatives are obtained as

= [0
=
ax

3x 2

4x 3

5.x ]{ai

(7-30)

Beam Bending and Beam-Column

Chapter 7

and

d 2w
dx 2

w
where

vv"

at the

two nodes leads to

\2x 2

6x

[0

20x 3 ]{a},

(7-31)

the second derivation or curvature. Evaluation of w, 6,

is

"10

'Wt

"

10

w"

{q}=<

and w'

>

IIP

w2

2/

/5

/4

4/

6/

12/ 2

3/

_0

(7-32a)

5/

20/ 3 _

or

{q}

(7-32b)

[A]{a}.

Therefore

{a}=[A]-{q},

(7-33)

where
"

~
9

[A]"

-10/ 6

-6/ 7

15/

8/

-6/

-3/

/2

-3/ 8 /2

10/ 6

-15/

3/ /2

-/

6/

/2

- -4/ 7
7/

- -3/ 5

P/2

-P
-P/2_

Hence,

uir)=[<D][A]-Hq}=[N]{q}.

As
i

before, multiplication of [<D][A]

1, 2,

6,

-1

(7-34)

leads to the interpolation functions

ti

given by

N = t

10s 3

N = l(s -

6s 3

Ni

=
N =
N,
5

P
(s

10j

/(-4s

+
+

- 6s
- 3s

155 4
8s 4

3s 3

5
,

),

3s 4

- 6s
- 7s - 3s

s 5 ),

(7-35)

\5s*

),

-V

Use of
patibility

these

comand curvature w". This would provide

interpolation functions will enforce interelement

of displacement w, slope

6,

Beam Bending and Beam-Column

Chapter 7

189

an improvement in the approximation to the continuity of the deformed


beam. The shape functions are plotted in Fig. 7-8.
Problem 7-9 refers to derivation of equations for the beam bending
problem with the fifth-order approximation. We have given partial results
and comments with Prob. 7-9.

BEAM-COLUMN
If in addition to the lateral

load

[Fig. 7-l(a)].

effect is called a

To

then the

load the

beam

beam

is

subjected to a constant axial

also acts as a

column, and the combined

beam-column.

simplify the problem,

we assume

that the axial load

is

small in corn-

Figure 7-8 Interpolation or shape functions for fifth-order model.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Beam Bending and Beam-Column

190

Chapter 7

Figure 7-8 (cont.)

parison with the lateral load and that the problem

assumptions,
Step

As

we can superimpose

the bending

and

is

linear.

Under

these

axial load effects.

before, the

beam

is

divided into one-dimensional line elements.

Step 2

Now

with the cubic model

we have

of the element: w, 9 for bending [Eq.

three degrees of freedom at each end

(7-2)]

and u

for the axial deformation

Beam Bending and Beam-Column

Chapter 7

[Fig. 7- 1(a)].

191

Thus the total number of degrees of freedom for the element


(w $ u u x h< 2 ,0 2 ,w 2 ). As in Chapter 3, we choose linear

[Fig. 7-l(d)] is six

approximation for axial deformation (denoted here as u instead of as v in


Chapter 3); accordingly, in terms of local coordinate s,

u(x)

[\

(7-36)

[NJ{qJ.

s]

For the bending part, we use the cubic function [Eq.


model can be now written as

The combined

(7-2)].

interpolation

["Ml

[NJ

'

(1

"

[0]

2)

(1

fe.J

4)

(7-37)

w(x)

(1

where [NJ and

[NJ

[0]

[q b ] are the

same

2)

(1

and

as [N]

4).

{q} in

Eq. (7-2); here

we have used

the subscript b to denote bending.

Steps 3 to 5

The strain-displacement and


bending problem are given

stress-strain relations for the axial

in Eqs. (3-1 2a)

and

(3-15)

and

and the
and

in Eqs. (7-10)

(7-11), respectively.

The

potential energy

bending [Eq.

II,

=/

and

(7-14)]

is

i-F(w") 2<fc

Here we assume

given as essentially the

axial

deformation [Eq.

-I

energies due to

Hence

wpds

Pu.

a concentrated uniform axial load; the term

is

the potential of load P.

Substitution for

E@ffds

Al

sum of the

(3-21)].

[Eq. (7-2)],

[Eq. (7-13)], u [Eq. (7-36)],

(7-38)

Pu denotes
and du/dx

[Eq. (3-12)] leads to

II,

= {q

T
b}

[B b Y[B b ]ds{q b }

AEl

{q a F[BJ

[BJ^{q

fl

-/['[NJfqJ^-PINJfq.}.
Here the subscripts a and b denote

Now we

axial

(7-39)

and bending nodes, respectively.


w l ,0 l ,u i ,w2 ,02 , and w 2 and

differentiate TL P with respect to

equate the results to zero to obtain a

set

of

six linear algebraic equations.

After proper integrations and rearrangement of terms, and assuming

uniform

in

to be

Eq. (7-1 6e), the resulting element equations can be expressed as

Beam Bending and Beam-Column

192

Y1E1

6EI

\2E1

/3

/2

\ o
jo

rv

12

Chapter 7

f
j

Pl

4EI

6EI

6EI

12EI
/

0,

ft

12

W2

pi
2

e2

pi*
12

j.
J

sym.

EA

EA

EA

(7-40)

P
"2

I
1

We

have arranged the element relations such that the superimposition of

The top relation is the same


and the bottom relation as in Eq. (3-28).
The remaining steps of assembly and introduction of boundary conditions are essentially the same as before. The reader should at this stage
the bending and axial effects can be seen clearly.
as in Eq. (7-15)

undertake solution of Prob.

7-8.

COMMENT
What happens

if

the load

between the bending and axial

is

not small and

effects

if

the coupling or interaction

needs to be considered ? For instance,

we ignored the bending effects that may be caused by P.


Under these circumstances, the problem can no longer be considered as linear,

in the foregoing,

and the principle of superimposition usually does not hold. One of the
is buckling. This type of nonlinearity falls under the category of geometrical nonlinearity and is beyond

manifestations of such a nonlinear problem


the scope of this text.

OTHER PROCEDURES OF FORMULATION


we shall discuss the use of the complementary energy and
mixed principles for the derivation of finite element equations. Only a simple
introduction to these procedures is included for advanced study the reader
may consult texts and publications referenced elsewhere in the text, e.g., in
Chapters 3 and 15.
In this section

Complementary Energy Approach

For generating a

stable structural element for

eliminate rigid-body degrees of freedom. For the

beam bending, we need to


beam element [Fig. 7- 1(d)],

such degrees of freedom are those relevant to the conditions of displacement

Beam Bending and Beam-Column

Chapter 7

and rotations

beam

at

one of the nodes. This leads to an element


Then the complementary energy

or cantilever (Fig. 7-9).

193

like a built-in
is

expressed as

a special case of Eq. (3-55c) as


i

f*

fw,l

(7-41)

n\N^

l
Q-|

Figure 7-9 Stable element for bending.

The

variation of

M along x can be expressed as


M = xQ -M

moment

(7-42)

-1]

[x

Note

that the

moment replaces

-\}dx

(7-43a)

-3/

2
2/
21

(7-43b)

-3/
3/1-3/
relation between the

flexibility

is

[x

The

primary unknown. The

stress as the

matrix according to Eq. (3-58a)

IT

end nodal forces

is

given by
(7-44a)

=IQ X

Mu

(7-44b)

or

-1

ia

2,

-1

[G]

(7-44c)

(7-44d)

M,

Now

the inverse of

[f ] is

LJ
/

By using Eq.

(7-45)

b/

2/2

(3-62), we obtain the element stiffness matrix [k], which will be


same as that in Eq. (7- 16b). If the variation of moment were different
from linear, we would obtain a different stiffness matrix.

the

Beam Bending and Beam-Column

194

Chapter 7

As stated earlier (Chapter 3), use of the stress function approach can be
more straightforward and easier. It is illustrated in Chapter 11.
Mixed Approach

Here the formulation involves two primary unknowns: displacement w


and moment (stress) M. We assume the same cubic function [Eq. (7-2)] for w
as
and linear variation for

M
M=

- s)M + sM

(1

= [N m ]{M

(7-46)

n ],

T =
where
is the moment at any point, [M n }
[M
2 ] is the vector of nodal
and
matrix
of
interpolation
functions.
moments,
[NJ is the
A special form of Eq. (3-63) for beam bending can be written as
t

n,

= - i- j" MMdx +

J"

M^dx - {Qf{q},

(7-47)

where

{Qf=[e,

2]

and
r
{q)

Substitution of

=K

<?,

e1i

>v 2

M and w" from Eqs. (7-46) and

(7- 13d), respectively, into

IIr leads to

{M.w
AJ

lf[

+ {Mr

r [NjqN'Trfxfa}

- {QF{q}.

(7-48)

For a stationary value, we take derivations of


and (w l5 6 l9 w 2 9 2 ). Thus

II* with respect to (Af,,

2)

fe = -A [Njr[N-k/x{M
+
^g|

P [N J

[N"]rfx{q}

-1

(7-49a)

0,

= o + j" [NJ[N"]V.x{M} -

{Qj

0,

(7-49b)

Beam Bending and Beam-Column

Chapter 7

195

or in matrix notation,

'

~~[6fJ
1

(7-49c)

<

M,

0,

w2
e ll

Mi,

where
[k Tr ]

= -^J"[N m F[NJ^
2

_/_

fTd-^)
Wl - J)

ds

IF Jo

'2

J_
6F

fcJ

o2

1"

(7-49d)

= rtNJTN"]^
s

-Hi'
i//

[-6 + 2^-

-4Z + 6/5

6-125

6ls-2l]ds

-1

1//

0'

(7-49e)
i//

1//

Hence the element equations are


_/_

IF

6F

J_

6F

J_
3F

_J_

j_

-1
_1_

l_

e,
(7-50)

<

o.

Qi

4-

As an
elements.

illustration,

we

The assemblage

consider the

r
beam

relation for the

in Fig. 7-4, divided in

two elements

is

two

obtained by

Beam Bending and Beam-Column

196

Chapter 7

observing compatibility of nodal moments, displacements, and rotations;


this gives

-1

3F

6F

(f

W\

-1
1

11

6F

3F

6F

Mi

H'2

QT-Qr
M - .\/f

3F

(7-5 la)

l)

02

6F

>

M<?

0i

Q\

M,

1
/

QT

H'3

0_

03'

2)

or

[K]{r]

Here the superscripts

By

(1)

and

(2)

(7-51b)

denote element numbers.

example of

substitution of the numerical values as in the

Fig. 7-4,

we

have, for uniform loading,


5

106

10

-1

2.5

106

10

"10

Hi

10

-1

01

2.5

10 6

10

10
106

2.5

10

10 6

10

fo

H'2

~I6

|= 1000

02

2.5

10 6

10

106

10

10

fo

'

Af3
\_

H'3

We

o_;

>

can see that at the midsection the slope 9 2

boundary conditions

(7-51c)

10

= 0.

10
1

12

Introduction of the

into Eq. (7-5 lb) leads to

\r

Beam Bending and Beam-Column

Chapter 7

2.5

-1

10

10

-^:

10

10

-l

01

12
2.5

10
10*

10*

10

2.5

~I6

~I6

10

10

Mz

10

:.

)= lOOOj

r-5id>

0:
2.5

~w

10

Vi

10
1

Note
Chapter

12/

that the assemblage matrix in Eq. (7-51) has zeros

such a system

If

10

3.

we

is

shall

on the diagonal.

solved by using Gaussian elimination, as discussed in

need to resort to partial or complete pivoting.

Solution of Eq. (7-5 Id) leads to

Af

*,
d.

= -833.33 kg-cm.
= 0.00 cm (given).
= 0.04999 rad.

M_ = -5S33.33 mg-cm.
w z = 0.312495 cm.
=0.0000

0-

-V 3

-833.33 kg-cm,
0.000 (given).

-0.05100

rad.

rad.

method. The
unknowns, whereas in the
displacement approach they were derived from computed displacements and
slopes. The results are the same because we assume linear variation for
moments. This is a rather elementary problem included simply to illustrate
the mixed procedure.
These

difference

results are the


is

same

that here the

as in the case of the displacement

moments

are primary

PROBLEMS
7-1.

(a)

Derive

stiffness

matrix

[k] for the

beam bending problem

if

EI

= F varies

linearly within the element as

f = y-.F - \ F
= s. s = (x x
:

where
(b)

V =
:

Derive

[k] if

and

\"
;

the area of the

beam

A = A \
:

7-2.

I.

varies linearly within the element as


:

- A \z
z

Derive the load vector due to uniform body force


with the approximation function for w as in Eq.

Z acting in
"-2
.
I

Solution:

4 i2'

1 1

the z direction,

Beam Bending and Beam-Column

198

7-3.

Derive
if

by a

matrix

stiffness

beam

the

is

the

[k] for

supported on an

Chapter 7

beam bending problem

elastic

(with uniform F)

foundation which can be represented

of linear elastic springs with uniform spring constant k f (F/L)

series

(Fig. 7-10).

k f = foundation spring constant

Figure 7-10

Part/a/ solution:

The

tion spring support

Beam on

elastic foundation.

additional term in 11^ [Eq. (7-14a)] due to the founda-

given by

is

(k f w)wds

Jo

folTNTMNHq}*,
which

will lead to the

stiffness

following contribution by the spring support to the

matrix

+ 2s
ls(l - 2s + s
s (3 - 2s)
Is (s - 1)
+ 2s
[1 - 3s
1

3s 2

i*-*/ J.
Jo'

)
>

13

11/

35

210

70

/a

13/

105

420

ls(l

- 2s

s 2)

s 2 (3

--

ls 2 (s - -

2s)

Ws

13/"

420
I

140

k
sym.

13

11/

35

210
/2
1( >5_

7-4.

_1
in Eq. (7-9) and show that the result yields
Multiply [0][A]
4, in

7-5.

Eq.

Perform

it i

1, 2, 3,

(7-3).
all

(7-1 6a) to

steps of multiplications

Eq.

(7-1 6b)

and from Eq.

and integrations
(7-1 6c) to (7-1 6e).

for going

from Eq.

Beam Bending and Beam-Column

Chapter 7

7-6.

Derive {Q}

if

199

the load varies as

p{s)

+N

NiPi

p2

+N

where

= 2s 2 - 35
N = s(2s -1)
No = 45(1 -s)
Ni

denotes the node at midpoint of the element.


7-7.

end fixed against movement and rotation and the other end
settles by 0.05 cm.
7-8.

One
One end

Solve the example problem in Fig. 7-4 with the following conditions (a)

By using two elements

and with constant

(Fig. 7-4)

free, (b)

axial load

P=

50 kg,

perform the entire process of formulation, assembly, introduction of boundary conditions, and solution of equations. Assume boundary conditions
Wj
7-9.

VV 3

Compute

0.

and load vectors due

the stiffness matrix

face traction

p with

body force

to

Z and

sur-

the fifth-order approximations [Eq. (7-34)].

Solution:

K=Fl Jof[BF[B]<fr
- 1205
/(-365 + 965 2 - 6O5
2
2 IO5
(1 -95 + I85
2
6O5 - I8O5 + 1205
/(-245 + 845 2 - 6O5
2
(35 - 125 2 + IO5
-6O5

1805 2

4-

where

[BF

=
/

3
)

Partial results:

= 17.41 F// k l2 - 175.24 F// 2


^ 4 = 8.57 F// 2 k 33 = 0.088 F/l
body force Z and surface load p, the load vector
3

kit

For constant

is

'

{Q}

= AZ

Compute
tion

(7-34)]

/120

>

/10

/120

Pi
1/2

-/ 2 /10

-/ 2 /10

the load vector for the

model [Eq.

/10

1/2

,/

7-10.

H2

'

111

'

/120.

/120,

beam bending element with

and surface loading varying


p(s)

(1

s)pi

sp 2

the interpola-

linearly as

Beam Bending and Beam-Column

200

Chapter 7

Solution:

Pi

Wt +

-Pi

*lft+A<*-*>]
Jpj_

Pi

280 >)

*120^
V12C

{Q}={

{% + Y4<J>2-Pi)]

[3
1

(Pl.

V120

+ Pi 210
_|_

Consider two elements (Fig. 7-4) with different values of E:

7-11.

Element

Element

1:

E=

10 6

2:

2 x 10 6 kg/cm 2

kg/cm 2

Assemble and solve for displacements, slopes, moments, and shear forces
for the other properties as for the example of Fig. 7-4. Comment on the
distribution of these quantities around the junction of the two elements.
7-12.

Assume

the area of the

beam element

to vary linearly as, see Prob. 7- 1(b)

- N A + N2 A 2
X

and derive element equations for uniform loading. Assuming the area to vary
shown in Fig. 7-11, compute displacements, slopes, movement,
and shear forces for the loading and properties of the example in Fig. 7-4.

linearly as

Note: Since area varies, corresponding variation

in

moment

of inertia should

also be considered.

10

cm

cm 2

10

-*

cm

cm 2

3 cm'

Figure 7-11

REFERENCES
[1]

Timoshenko,

S.,

Strength of Materials,

Van Nostrand Reinhold, New York,

1956.
[2]

[3]

Strang, G., and Fix, G. J., An Analysis of the


Hall, Englewood Cliffs, N.J., 1973.

Finite

Element Method, Prentice-

Prenter, P. M., Splines and Variational Methods, Wiley,

New

York, 1975.

Beam Bending and Beam-Column

Chapter 7

[4]

[5]

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Crandall,

S.

H., Engineering Analysis,

to the Finite

McGraw-Hill,

201

Element Method, Van

New

York, 1956.

ONE-DIMENSIONAL
MASS TRANSPORT

INTRODUCTION

number of problems in various disciplines of engineering involve the


phenomenon of mass transport. This can include transport through diffusion
and convection of chemicals, pollutants, contaminants, and dissolved salts
we shall treat mass transport for simple one-di-

in water. In this chapter,

mensional idealizations.

The

differential

be stated as

where

Dx

equation governing one-dimensional mass transport can

[1]

is

the dispersion coefficient, c*

(concentration of pollutant or dissolved

is

salt),

the

vx

unknown
is

or state variable

the velocity or convection

x and t are the space and


same as the first term in the
flow equation [Eq. (5-9)] and represents the phenomenon of diffusion. The
second term denotes the process of transport by convection.
parameter,

is

the applied source or sink, and

time coordinates. The

FINITE

first

term

is

essentially the

ELEMENT FORMULATION

In this and the following chapters,

main

steps, while

labels.

202

some of

the

we

common

shall often detail

and

label only the

steps will be discussed without explicit

One-Dimensional Mass Transport

Chapter 8

203

H
T
\

\
\

\
\
\
\

'a)

(b)

A A A
k

L=+1

L = -1

(c)

Figure 8-1 One-dimensionai mass transport, (a)


diffusion

and convection,

(b)

Mass transport by

One-dimensional idealization,

(c)

Dis-

cretization.

The domain of mass transport (Fig. 8-1) is idealized as one-dimensional.


Use of the line element and the linear approximation gives the approximation
for c as

i(l

L)c

i(l

+ L)c

(8-2a)

or
c

where

{q}

[cj

c 2 ],

[N]{q]

and

is

= E Vet,

4.

U2,

(8-2b)

and
medium.

the approximation to the concentration c*

denotes local coordinate; in Fig. 8-1,


Step

it

also denotes length of the

Derivation of Element Equations

For derivation by Galerkin's method, the

residual

is

(8-3)

One- Dimensional Mass Transport

204

Weighing

with respect to

TV,-

yields

|r(S
As explained
ever,

Equation

in

Chapter

3,

we

(8-4)

is

NjCj) N dx = 0.
t

shall treat the elements

understood that the procedure

it is

Chapter 8

one by one. How-

applied for the entire medium.

is

expanded as

[ilD>i&

Nrt- wJSEg*
dx

W - jf(LNJ>)Ntdx = 0.
Integrating by parts the
stant,

(8-4)

first

term

Eq. (8-5) and assuming

in

Dx

(8-5)

to be con-

we have

JT^-icEW }N dx = D

x N,$i:

dn

*/

X\

The first term on the right-hand side in Eq. (8-6) denotes Neumann-type
boundary conditions and can be specified as a known value of the normal
derivative dc/dn. The term D x (dc/dn) j is the (known) flux on the end
boundaries.
Substitution of Eq. (8-6) into Eq. (8-5) leads to element equations in

matrix form as
[E]{q}

[E,]{q}

(8-7)

{Q},

where {q} = [dcjdt dc 2 /dt], [E] and [E ] are the element property matrices,
and {Q} is the element forcing parameter vector. These matrices are defined
r

as follows:
[E]

= r([BYD x [B]dx +

[BYvJS\dx)dx

(8-8a)

[E,]= r'[NF[N]</x,

(8-8b)

and
{Q}

p [WWdx - D N^
x

(8-8c)

Here [B] is the usual transformation matrix [Eqs. (3-13) and (4-4)] obtained
by taking the proper derivative of c [Eq. (8-2)]. In writing Eq. (8-8a) we
assumed, only for simplicity, that v x
second term in Eq. (8-5)

is

is

constant with respect to x; hence the

not included.

We

note that

it is

not necessary to

make this assumption because that term can be easily included.


The evaluation of [E], [EJ, and {Q} is as follows:

One- Dimensional Mass Transport

Chapter 8

[-1

[E]

- /[_]
-

We

ij

205

\}dL

+ 2[

i_

DJl - vJ2 -DJl - vjT


DJl + vJ2_\
-DJl - vJ2

(8-9)

note here the important characteristic that the element matrix [E]

nonsymmetric ;

mass transport.
[E,]

this

property

is

is

contributed by the convection part of the

Now

2
"2

[1

-L

L]dL

11
(8-10)

This matrix

[Eq

is

from the time-dependent term

similar to the matrix arising

(5-21)].

Wl\

(8-11)

1]

(4

The first term in Eq. (8-11) indicates that applied sink or source quantity is
lumped equally at the two nodes. As discussed in previous chapters, only
the terms at the end nodes remain when the second part of {Q} is considered.
Step

5.

Assembly

Consider the three-element mesh (Fig.

Combination of element

8-1).

matrices for the three elements leads to the assemblage equations as

D x /l -Dx ll +

v x /2
v x l2

- D x /l -

2D x /l +

-D x + v x /2
/l

v x /2

-D x \l-v x \2
2D x jl +

(c

-D x /l-v x /2

-D x + v x /2

D x /l +

/l

10 0"
14 10
14

"2

v x /2j

)c 3

dd/dt

daldt

<

dajdt

2_

dctldt,

Wl
= E{
2

2
,

(8-12a)

One- Dimensional Mass Transport

206

The contribution of

the boundary terms

Chapter 8

not included because

is

we assume

that

|(A0 =

(8-13a)

and

c(0,t)=c.

(8-13b)

In matrix notation, Eq. (8- 12a) becomes

[K]{r]

where

and

[K]

[K,] are the

KSjt)

{R},

(8-12b)

assemblage property matrices, {R}

blage forcing parameter vector, and

{r} is

is

the assem-

the vector of assemblage nodal

unknowns.
Solution in

Time

number of

integration schemes can be used to solve the matrix equa-

For

tions in the time domain.

form as

difference

instance, Eq. (8-1 2b) can be expressed in finite

[2]

- 0)W,)
+

RK{rU +

+ o-^W-^{*L
We

where 6 is a scalar.
ing on the value of 6.

= ^,

we

(8 - 14)

obtain various implicit and explicit methods depend-

If

1,

the fully implicit

method

is

obtained, and

if

obtain the Crank-Nicholson scheme. These schemes possess dif-

ferent mathematical properties such as convergence

and

stability

which can

influence the quality of the numerical solution. Detailed discussion of these

aspects
If

is

we

beyond the scope of this

text.

use the following scheme to approximate the

first

derivatives in

- ^[K ]){r} r

(8-16)

Eq. (8-14),

then Eq. (8-14) becomes (for 6


([K]

^[K,]){r} f+A

~ 2{R} f+A -

Since the values of {R} f+A and


,

{r},

([K]

known, we can compute {r} r+Ar at the


The procedure starts from the first step

are

next time step by solving Eq. (8-16).

when the values of {r} at t = are prescribed as initial conditions.


The boundary conditions are prescribed as known values of {r}
nodes. Equation (8-16)

is

at given

modified for these conditions before the solutions

are obtained at various time levels.

CONVECTION PARAMETER v x
For the solution of Eq. (8-16), the value of the convection velocity v x
should be known. It can often be obtained by using available data or formulas.

One- Dimensional Mass Transport

Chapter 8

One

of the ways to compute

it is

207

to first solve the flow or seepage equations

which allow computation of v x


Comments. In most engineering problems the assemblage matrices are
symmetric and banded. In the case of mass transport, the matrix [K] in Eq.
(8-16) is banded but not symmetric. Consequently, as discussed in Appendix
2. we have to store in the computer core the coefficients on the entire band
(2B-1). This aspect will involve relatively more computer time in solving the
equations as compared to the time required for symmetric and banded systems. The nonsymmetric property of [K] can be considered to be a characteristic of the non-self-adjoint nature of the problem.
The behavior of the numerical solution can be influenced significantly
due to the existence of the convection term. If the magnitude of the convection term is relatively large, the system [Eq. (8-1)] is predominantly convective.
(Chapter

4),

This can render the solutions more susceptible to numerical instability.


the other hand,

On

the diffusion term predominates, the numerical solution

if

can be well behaved.

Example

To

8-1

illustrate a

few time steps

in the solution of

Eq. ($-16). we adopt the following

properties:

D,=
*=
1=

(i-

(L

[L\

W=
At
Here L denotes
illustrate the

length.

may

Tk

{MTU)

Tk

M mass and T

procedure, and

n.

time.

These properties are chosen simply to

not necessarily represent a

field situation.

Boundary- conditions.

aO.n =

(ML

1.0.

3
)

at

node

1.

>

0. (Fig. 8-1).

Initial conditions:
ci.v.

At time
i

-i

-1

At,

from Eq.
"2

0"

-i

-4

-4

_3

0)

(.8-16).

0.0.

we have
0~\ 'ci

14 10
14

J_
j

l_

_0
1

= -

i -i
-4-

!_

c:

c3

2_|/ [C4

Ar

o"

2-1

-i
\

r<0.

0-4

2
1

-f
4_

x
x

~2

:_

_0

"\

\C\

c2
Cl
ie4

One- Dimensional Mass Transport

208

Because c

c2

c3

C4

0at

o,

0"

Cj

_7

20

Cl

=r

>

37

20

-i

Chapter 8

6
1

13

,c A

Introduction of the boundary condition c x

<

>

c3
Ar

gives

Cl

20

J
1

20

13

-i

C2

<

|c 3

[oJ

- lc 4

Solution by Gaussian elimination leads to


c,

c2
c3
c4

=
=
=
=

1.00 (given),

509 x 10" 4

260 x 10" 5

200 x 10~ 6

The next stage, t = / + A/ = 2Af, can now be performed by using the values of c
computed at the end of t = At as initial conditions and so on for other time steps.
Example 8-2

A solution to Eq. (8-1) was obtained by Guymon [3] using a variational procedure.
We present here his results, which compare numerical predictions with a closed
form solution for a problem
in

idealized as one-dimensional.

terms of the nondimensional quantities x/L, c/c

the total length of the one-dimensional

The

initial

The results are presented


and v x l/D x where L is

v x t/L,

medium.

conditions are

c(x, 0)

and the boundary condition

0,

is

c(0,f)

That

is,

Co

(0,0

Table 8-1 shows a comparison between the numerical solutions and the closed form
solution. The former are obtained for two conditions: v x l/D x = 0.5 and v x l/D x =
0.25 at a time level v x t/L = 0.5. The closed form solution is obtained from [3]

where

erfc denotes error function.

One- Dimensional Mass Transport

Chapter 8

TABLE

209

Comparison Between Finite Element Predictions


and Closed Form Solution at v x t L = 0.5 [3]

8-1

Closed

Finite

Element Solution

Form

Dx =

Dx =

x/L

Solution

1.000

1.000

1.000

0.1

0.976

0.975

0.972

0.2

0.928

0.927

0.923

0.3

0.851

0.850

0.845

0.4

0.-45

0."43

0.738

0.5

0.616

0.614

0.606

0.6

0.478

0.475

0.463

0.7

0.345

0.341

0.324

0.8

0.230

0.228

0.209

0.9

0.142

0.145

0.132

1.0

0.080

0.060

0.080

Comment. The

vx l

0.25

vx l

0.50

magnitude of the convective


less accurate. This term has
significant influe: s on numerical solutions of the one-, two-, and threedimensional problems involving diffusion-convection phenomena.
term vj

results indicate that as the

D., increases, the

numerical solutions are

REFERENCES
[1]

Bear,

J..

Dynamics of Fluids

in

Porous Media. American Elsevier,

New

York,

1972.
[2]

Richtmeyer. R. D.. and Morton. K. W.. Difference Methods for Initial-Value


New York. 1957.

Problems. Wiley-Imerscience.
[3]

Guymon. G.

L..

Finite

Element Solution of the One-Dimensional Diffu6. No. 1. Feb. 1970,

sion-Convection Equations." Water Resour. Res.. Vol.


pp. 204-210.

BIBLIOGRAPHY
Amend.
fate

cal

Contractor. D. N.. and Desai. C. S.. "Oxygen Depletion and Sulin Strip Mine Spoil Dams." in Proc. 2nd Intl Conf. on NumeriMethods in Geomechanics. Blacksburg. Ya.. C. S. Desai fed.). June 1976.
J.

H..

Production

ASCE. New York,

1976.

Cheng. R. T.. '"On the Study of Convective Dispersion Equations." in Proc. Intl
Symp. on Finite Element Methods in Flow Problems. Swansea. U.K., Jan. 1974,
University of Alabama Press, Huntsville, 1974.

One- Dimensional Mass Transport

210

Chapter 8

and Contractor, D. N., "Finite Element Analysis of Flow, Diffusion


Water Intrusion in Porous Media," in Proc. U.S. -Germany Symposium
on Theory and Algorithms in Finite Element Analysis, Bathe, K. J., Oden, J. T.,
and Wunderlich, W. (eds), M.I.T., Cambridge, Mass., Sept. 1976.

Desai, C.

and

S.,

Salt

L., Scott, V. H., and Hermann, C. R., "A General Numerical SoluTwo-Dimensional Diffusion-Convection Equation by the Finite Element Method," Water Resour. Res., Vol. 6, No. 6, Dec. 1970, pp. 1611-1617.

Guymon, G.
tion of

Proceedings, First International Conference on Finite Elements in Water Resources,

Gray,

W.

G., and Pinder, G. F. (eds.), Pentech Press, London, 1977.

Segol, G., Pinder, G. F., and Gray,

W.

G.,

"A

Galerkin Finite Element Technique

for Calculating the Transient Position of the Salt


Res., Vol. 11,

Smith,

I.

No.

2,

Water Front," Water Resour.

April 1975, pp. 343-347.

M., Farraday, R. V., and O'Connor, B. A., "Rayleigh-Ritz and Galerkin

Finite Elements for Diffusion-Convection Problems," Water Resour. Res., Vol.


9,

Wu,

No.

3,

1973, pp. 593-606.

S., and Contractor, D. N., "Finite Element Procedure for


Water Intrusion in Coastal Aquifers," Report No. VPI-E-76-23, Department of Civil Engineering, Virginia Polytechnic Institute and State University,

T. H., Desai, C.

Salt

Blacksburg, Va., July 1976.

ONE-DIMENSIONAL

OVERLAND FLOW

INTRODUCTION
Overland flow can include problems

in hydraulics

and hydrology such as

runoff due to rainfall, flow in shallow open channels and in rivers in flood
plains,

and flood routing. The problem

is

three-dimensional, but

be approximated by one-dimensional idealization, and for


treatment,

we

shall consider the

this

it

can often

introductory

one-dimensional case.

The equations governing overland flow

and

consist of the continuity

momentum equations developed by Saint- Venant. The equations for gradually


varying flow in open channels can be assumed as follows [1-5]
Continuity equation:

(i)

dQf.dA
+ dt
dx
(ii)

Momentum

(9-1)

0.

equation:

dQ

*+&5) -**-*>-*

<>

dx~

Here

Qf

is

the discharge in overland flow (Q f0 ) or channel flow (Q fc ) r is


excess rainfall in overland
;

the lateral inflow per unit length of flow plane

flow

bed

(r )

and net inflow

slope,

in the

for overland

channel caused by overland flow

and Sc

for channel

the area of overland flow or channel flow

Sf
is

is

(r c )

the friction slope

is
;

the

is

the coordinate in the flow

211

212

One- Dimensional Overland Flow

Chapter 9

Ac
Section C-C

0-0

Section

(a)

Lateral overland flow

due to

I from overland flow,

rainfall, r

(b)

Channel inflow
I

rr

(c)

Figure 9-1

Representations

direction

is

overland

for

Overland and channel flow,


rainfall, (c) Channel flow.
(a)

(b)

and channel

flows,

Overland flow due to (excess)

the coordinate in the depth of flow direction

and

is

the

time [Fig. 9-1 (a)].

Various assumptions lead to the so-called kinematic wave approximation


[6],

according to which the

momentum equation reduces to


S

Sf

(9-2b)

Equation (9-2b) can be approximated by using Manning or Chezy equations.


use Manning's equation, given by

We

One- Dimensional Overland Flow

Chapter 9

K=!^/y
n

213

S 1/2

(9-3a)

or

Q f = h^.RpS 1/2 A
V

where

is

the velocity of flow,

Rh

is

(9-3b)

the hydraulic radius

area/wetted

and n is the Manning roughness coefficient. As a consequence of


the use of Manning's formula, the finite element procedure essentially
perimeter,

involves solution of Eq. (9-1).

APPROXIMATION FOR OVERLAND


AND CHANNEL FLOWS
Figure 9- 1(a) shows an arbitrary domain with a channel. The rainfall excess

Q f0 which discharges into the channel, resulting in the


Q fc A number of investigators have considered finite element

causes overland flow

channel flow

solution for general


[7-11].

As

In the

first

and one-dimensional

a simple approximation,
stage, the overland flow

we can

idealizations for overland flow

solve the

due to excess

problem

rainfall, r

in

two

stages.

[Fig. 9- 1(b)], is

computed by solving Eqs. (9-1) and (9-3) for the overland flow. The results
of flow from the first stage are considered as the input flow r c in the channel
[Fig. 9-l(c)]. Then Eqs. (9-1) and (9-3) are solved for computing the channel
flow. The same finite element formulation can be used for both stages.

ELEMENT FORMULATION

FINITE
Step

Discretize and

1.

The

Choose Element Configuration

domain [Fig. 9-2(a)] is idealized as shown in Fig.


domain is divided in two parts: overland part [Fig.
9-2(c)] and channel part [Fig. 9-2(d)]. The overland part is divided into a
suitable number of zones depending on the physical characteristics such as
the bed slope and topography. Each zone of the overland domain and the
9-2(b).

arbitrary flow

The

channel

is

Step

2.

idealized

replaced by one-dimensional line elements [Fig. 9-2(e)].

Choose Approximation Functions

The area A and

the discharge

Qf

can be assumed as unknowns and

represented by using linear approximation

A(x9

t)

= (l-s)A (t) + sA 1
l

(t)

[N]{A(r)}

= jlN Afi)
t

(9-4)

214

One- Dimensional Overland Flow

Chapter 9

Divisions idealized on the basis

of topographic
and other properties

(a)

r-

~~

Overland

0)

.c

--*

- <>,--<

^--

-*

>1

L_

(b)

(d)

(c)

= x//

(e)

Figure 9-2

Finite

(b) Idealization, (c)

element

discretization,

Overland flow,

(d)

(a)

Actual

Channel flow,

(e)

region,

Generic

element.

and

Qf (x,

t)

= (1 - s)Q n {f) + sQn {t)


= mQfjm
= t N,Qfl
{t),

is

= [A

the vector of nodal areas and [Q fn } T


the vector of nodal discharges.

where {A} T

A 2]

is

(9-5)

= [Qfl Qf2

One- Dimensional Overland Flow

Chapter 9

215

In view of the use of Manning's equation, the values of

we

various time levels, and as

area

as the time-dependent

Qf

are

known

at

shall see later, the solution involves only the

unknown.

Derivation of Element Equations

Step

4.

We

use Galerkin's residual method to derive the individual element

equations. Minimization of the residual of Eq. (9-1) leads to

(9-6)

i
Use of Eqs.

(9-4)

and

(9-5) gives

(9-7a)

[BHQ,.}

and

^=
=
When

[1

s]{A]

[N]{A}

[N](^>

(9-7b)

we obtain

Eqs. (9-7a) and (9-7b) are inserted into Eq. (9-6),


X

n[NHN]rfjr{A}+
J x\

JfX\

'[N] r [B]{Q /n }rfx= ["

[Wrdx

(9-8a)

J X\

or

[15

s]ds {k " ]

-r[

(-2/1
Ir

After performing integrations,

V2

11

'

'

Qn - Q ;-

Ir
;

dA
|_1

2j [dt

(9-8b)

rrrh-

we have element equations

[dAA
dt

- Qf i)ds

'

<

(9-9a)

>

7
J

or
[k]{A}

Step

5.

(9-9b)

{Q}.

Assembly

Consider a mesh made of three elements of lengths


9-3(a)].

The element equations can be assembled

as

/,,

/2

and

/3

[Fig.

One- Dimensional Overland Flow

216

~2h

Chapter 9

0" dAA

/,

dt
(21,

2/ 2 )

dA 2

/,

dt

(ll 2

l2

dA
21,)

dt

Symm

dA 4

Qh Qh [Qh ffi>.

dt\

2/ 3 _

Qh
Qh + Qh - Qh
Qh + Qh - Qh
Qh

(Vi

+4

(9-10a)

<

hr 2

hr z

IVs

or in matrix notation
[K]{A}

{R g }

{R,}

{R}

(9-10b)

where the superscript on Q f denotes an element and {AN } = [A A 2 A 3 A A ].


Here we have used {A^} for the asemblage vector in place of [r] to avoid
T

confusion with

r that is

used for applied inflow.

Figure 9-3 Discretization and timewise solution, (a) Three-element


discretization, (b) Finite difference

approximation for

first

deriva-

tive.

(?) Globa node


I

A A

(b)

(a)

+ At

Local node

/l\ Element

One- Dimensional Overland Flow

Chapter 9

We

use the most simple time integration, often called Euler formula to

approximate the

derivative in Eq. (9-10), [Fig. 9-3(b)].

first

jA^jf

iAy]r+Ar
lAtfJ

If

we

(9-11)

At

substitute Eq. (9-11) into Eq. (9-10), the result

^[K]{A^J f+Af

[R Q }

now

Equation (9-12) can

The

+ JptKKA*},.

(9-12)

be solved by using the "marching" process in

solution starts at time

= 0, when we can assume that

A (x,0)=0]
i

where

{R r }, +Af

is

CONDITIONS

INITIAL

time.

217

the upper

Q fi (x,0)=0\
= number of nodes. We
boundary

(or

i=

further

an end node)

N,

1,2,

(9-1 3a)

assume that discharge

at all times

that

2/(0,0=0.
The values of excess

rainfall for

is

zero at

is,

(9-13b)

overland flow and resulting inflow due to the

excess rainfall for the channel flow are prescribed as

= F(x,

r(x, t)

(9-13c)

t),

where the overbar denotes known quantity.


Solution in

With

the conditions in Eq.

J.

0"

"2/,
/,

Time

1,

2/,

6At

2/ 2

2/ 2

.0

/,

2/ 3

2/J

Mil
A2

Si

+ hh

At

is

(9-14)

Ay

UJ

I3P3

t = At. These values


by using Manning's equation
in which the bed slope S and coefficient n are prescribed, and the knowledge
of areas and prescribed values of average widths permit computation of R h
At the next time step t = At + At = 2At, the vectors {R Qit* and
(\IAt)\\L]{K N }to are now available from the solutions at t = At, while {FN } is
known at all time levels hence

Solution of Eq. (9-14) gives values of areas of flow at

are

now used

to find nodal quantities of flow

[K]{Aj

The procedure

is

2Af

,}*

{R,h*

+ ^[kha*}*.

repeated until the desired time level

is

reached.

(9-15)

Step

Solution for {\ N }

6.

Example

As a simple

9-1

procedure

illustration of the solution

in time,

we adopt

the following

properties:

/=
r =
At =

Element length,

Flow due

to excess rainfall,

Time increment,
Here and

in the

following

1.0m;
10

may

simple properties in order to illustrate

not necessarily represent a

Substitution in Eq. (9-14) yields equations at


"2
1

0"

x 0.166

At

A2

A3

2_

A4

field situation.

as

At

_0

assumed constant

0.166 hr.

we have chosen

the procedure; these properties

m 2 /hr,

-1

1.0

x 10"

2
,

2
2
1

At

or

0~

~2

A3

2_

A4

0.5

'Ai

A*

(9-16)

'

_0

Comment.

'

.0.5. At

At

We note here that the matrix differential equations [Eq.

sent an initial value problem.

The

the stiffness matrices (Chapter

3),

resulting matrix [K]

is

and can be solved as expressed

solution by Gaussian elimination gives the areas of flow at

A = A2 = A3 = A4 =
l

0.166

(9-9b)] repre-

not singular, in contrast to

in

Eq. (9-16). The

At:

2
.

Now we use the Manning formula [Eq. (9-3)] to find nodal flows Q f by assuming
the following data:

= 0.3,

5=

0.1;

Average width of flow


Hence,

at

nodes

at global

= 0.5 m,

= 0.6 m,

w3

= 0.75 m, u> 4 =

-0.166/0.5

Q n =^(o.33)- 667 (0.1)

218

w2

= 0.330 m,
=
Rh2 =0.166/0.6 0.277 m,
Rh3 =0.166/0.75 = 0.220 m,
RM =0.166/1.3 = 0.128 m,
nodes at time = At as

R hl

which gives flows

x 0.166

0.125

/hr,

1 .3

m.

One- Dimensional Overland Flow

Chapter 9

1.49

Qf:

(0.277)- 667 (0.1)

x 0.166

0.110,

x 0.166

0.095,

0.3

Q f3 =

1.49.

^y(0.220)- 667 (0.1)


1.49,

(0.128)

Qf

To

219

667

(O.l) 05

x 0.166

0.066.

0.3

evaluate the

first

part of the forcing vector in Eq. (9- 10a),

necessary to

it is

convert the above flows at global nodes into flows at local nodes, Fig. 9-3(a). If

we

assume that the flow at a global node is divided equally among the common local
nodes of two adjoining elements, and that only half of the flow at an end node is
that node, then Q} = 0.0625, Q} 2 = 0.055; Q fl = 0.055,
=
0.0475, and Q} 2 = 0.033.
Q}
time = A/ + A/, we now have, using Eqs. (9-10a) and (9-15),

effective

At

at

0.0475;

0~
1

'Ai

0.0075

A2

0.0150

0.5
1.0

>

Q} 2

x 0.166

A3

2_

A,

0.0220

1.0

0.0145

2 At

'

0.5

Af

(0.1667

0.1667

(9-17)

x 0.166

0.1667
10.1667

.0

Solution of Eq. (9-17) leads to nodal areas at


for the required

2A/.

The procedure

is

repeated

number of steps.

Example 9-2
field problem solved by Ross [10] using a finite element procean area of 136 square miles around a portion of the South River
in Virginia. Field observations of overland and river flow were available from a
gaging station at the base of the area considered. The region was divided into three
subregions upper South River, Back Creek, and lower South River containing

Figure 9-4 shows a

dure.

It

consists of

three channels.

The

finite

element formulation

is

applied independently to three

overland flows and three channel flows and the results are added to obtain computed
flow at the gaging station. Special provision was

modate

Figure 9-4 also shows one of the


[10];

it

made

in the

formulation to accom-

existing flood retention structures [10].


finite

element discretizations used by Ross

contains three overland flood plains (OFPs) that are idealized as 24 one-

dimensional elements. The elements are labeled as IA1, meaning


idealized strip(s) in flow plane,

flow planes,

strip,

and element

and elements are shown

first

flow plane,

in the strip. Detailed properties of the


in

Table

9-1.

The

three channels were

discretized similarly; the properties are given in Table 9-2. Values of excess rainfall

220

One- Dimensional Overland Flow

Chapter 9

Legend
Overland flow plane boundary
Surface element boundary

Channel element
Primary channel

Secondary channel
Flood detention structure and reservoir

IIB1

Figure 9-4 Finite element

watershed

mesh

(24 elements) for upper South river

[10].

based on the rainfall during Hurricane Camille, the night of August

which

total rainfall in the region

ranged from about

3 to

1 1

in.,

19, 1969, in

were used as the

in-

put for overland flow.

Figure 9-5 shows a typical comparison between the computed and observed

(hydrograph) values of discharges during the period of the hurricane. For these
results a value of time

Comments.

We

increment At

900 seconds was used.

have presented simple problems only as an introduction.


factors such as multidimensional effects and physical

number of other

One- Dimensional Overland Flow

Chapter 9

TABLE 9-1

221

Surface Geometry and Properties of 24-Element Mesh

[10]

Average

OFP

Strip

Element

No.

No.

No.

A
B
2

D
2

E
F
2

G
H
A

II

B
2

C
2

D
2

E
F
2

III

TABLE

9-2

Length
(m)

Width

Slope

Manning

(m)

(m/m)

2334
4023
4023
2334
4023
4023
2334
4023
4023
2334
4023
2736
2736
2855
2655
2736
2736
2655
2655
2736
2655
2655
2655
2414

9012
2374
7121
5177
7215
6196

0.028

0.297

0.016

0.343

0.010

0.336

0.020

0.323

0.065

0.316

0.025

0.309

5981

0.024

0.244

2982
5825
5338
2843
4855
5552

0.103

0.292

Element

No.

No.

Length
(m)

properties of the flow


tions. Also,

0.340

0.006

0.338

0.041

0.286

0.024

0.266

1475

0.108

0.300

4426
3018

0.043

0.300

0.111

0.280

5968

0.071

0.291

4493

0.111

0.300

5432

0.047

0.286

2950
4734
5512

0.011

0.231

0.066

0.308

0.019

0.294

1931

0.003

0.264

3058

0.057

0.121

[10]

Slope

Manning

(m/m)

9495

0.013

0.100

5686

0.004

0.085

5686
5686
5901

0.001

0.085

III

0.310

0.017

Geometry and Properties of Channel

Channel

II

0.027

0.001

0.105

0.039

0.120

5901

0.010

0.115

5901

0.005

0.115

3862

0.002

0.130

domain may require additional advanced considera-

numerical characteristics such as convergence and

require special attention for

more complex problems. Hence,

stability

may

the derivations

One- Dimensional Overland Flow

222

Chapter 9

600.0

500.0

400.0

o
*

Calculated

Recorded

\\

-o

\\

300.0

n*--

- "*<^

200.0
/

'

100.0

//

J/
0.0 SdfeHBE
0.0

10.0

5.0

20.0

15.0

Time

25.0

30.0

35.0

40.0

(hr)

Figure 9-5 Comparison of observed and computed discharges of

gaging station, At

and

results presented herein

900 sec

[10].

should be treated simply as illustrations for the

one-dimensional idealization. For practical applications and multidimensional


problems, the user should consider other relevant characteristics.

REFERENCES
[1]

"Numerical Solution of Flood Prediction and River Regulation


I Derivation of Basic Theory and Formulation of NumerMethods Attack," Report No. IMM-200, Institute of Mathematical

Stoker,

J. J.,

Problems, Report
ical

Science,
[2]

[3]

[4]

New York

Abbott, M.

University,

New

York, 1953.

and Ionescu, F., "On the Numerical Computation of Nearly


Horizontal Flows," /. Hyd. Research, Vol. 5, No. 2, 1967, pp. 96-117.
B.,

Brutsaert, W., "De Saint-Venant Equations Experimentally Verified,"


Hydraulics Div. ASCE, Vol. 97, No. HY9, Sept. 1971, pp. 1387-1461.

/.

J. A., and Woblhiser, D. A., "Difference Solution of ShallowWater Equations," /. Eng. Mech. Div. ASCE, Vol. 93, No. EM2, April 1967,

Liggett,

pp. 39-71.
[5]

Prince, R. K., "Comparison of Four Numerical Methods for Flood Routing," /. Hydraulics Div. ASCE, Vol. 100, No. HY7, July 1974, pp. 879-899.

One- Dimensional Overland Flow

Chapter 9

[6]

Lighthill, M.

Movement

in

and Whitham, G.

J.,

Long

B.,

223

"On Kinematic Waves,

Rivers," Proc. R. Soc. London Ser. A, Vol. 229,

I:

Flood

May

1955,

pp. 281-316.
[7]

[8]

Judah, O. M., "Simulation of Runoff Hydrographs from Natural Watersheds


by Finite Element Method," Ph.D. thesis, Virginia Polytechnic Institute and
State University, Blacksburg, Va., Aug. 1972.

Al-Mashidani, G and Taylor, C. "Finite Element Solutions of the Shallow


Water Equations Surface Runoff," in Proc. Intl Symp. on Finite Element
Methods in Flow Problems, University of Wales, Swansea, U.K., Jan. 1974,
Oden, J. T., Zienkiewicz, O. C., Gallagher, R. H., and Taylor, C. (eds.), Uni.

versity of
[9]

[10]

[1 1]

Alabama

Press, Huntsville, 1974, pp. 385-398.

Taylor, C., Al-Mashidani, G., and Davis, J. M., "A Finite Element
Approach to Watershed Runoff," /. Hydrol, Vol. 21, No. 3, March 1974.
Ross, B. B., "A Finite Element Model To Determine the Effect of Land-Use
Changes on Flood Hydrographs," M.S. thesis, Virginia Polytechnic Institute
and State University, Blacksburg, Va., Nov. 1975.

Taylor, C, "A Computer Simulation of Direct Run-off," in Proc. Intl Conf.


on Finite Elements in Water Resources, Princeton University, Princeton, N.J.,
July 1976, Gray, W. G., and Pinder, G. F., (eds.), Pentech Press, London,
1977, pp. 4.149-4.163.

ONE-DIMENSIONAL
STRESS WAVE PROPAGATION

INTRODUCTION
5, 8, and 9, time-dependent problems of heat and fluid flow, mass
and overland flow were considered. Although these problems were
represented by different mathematical equations, the finite element solutions

In Chapters
transport,

in time involved first-order derivatives with respect to time.

Now we consider

a different kind of time-dependent problem which involves second-order time


derivatives in the finite element equations.

When a time-dependent force caused by factors such as an impact, blast,


and earthquake loading impinges on a medium, it is transmitted through the
medium as a (stress) wave. Generally such waves propagate in all the three
spatial directions. Under certain circumstances and assumptions, it is possible
to idealize the

medium

as one-dimensional.

Consider a homogeneous bar of uniform cross section (Fig. 10-1). A timedependent force Px (t) acting on the bar causes vibrations in the bar, and a
(stress)

wave propagates

to

and

fro in the bar.

equation for the one-dimensional case, often


is

given by

where a x

is

[1,

The governing

known

as the

differential

wave equation,

2]

the axial stress in the

material of the bar, and u

is

direction,

is

the mass density of the

the axial displacement. Equation (10- la)

is

statement of the dynamic equilibrium at an instant of time and can be


224

One- Dimensional Stress Wave Propagation

Chapter 10

225

(a)

= +1

(0

(b)

Figure 10-1 One-dimensional wave propagation, (a) Bar subjected


to time-dependent load, (b) Idealization

and

discretization,

(c)

Generic element.

derived by using Newton's second law.


the

force;

acceleration

term p(d 2 u/dt 2 )

and

Px (i)

is

= pii

derivatives,

is

the elastic

is

is

the

assumed to be

is

=E^.
L
dx

(10-2a)

we obtain
fo x
dx

where

law

Ee
Lx

ox
first

left-hand side denotes internal

the external force. If the material

linearly elastic, the stress-strain

Taking

The

denotes inertia force, where u

=E
dx

modulus and e x

(10-2b)

is

the gradient of u or axial strain.

Substitution of Eq. (10-2b) into Eq. (10-la) gives


dh4.

dx 2

dt 2

+ PM.

(10-lb)

The quantity */E/p = c x is called the velocity of elastic wave propagation.


A number of other physical phenomena such as propagation of waves of
sound and vibrations of strings are also governed by Eq. (10-1), which in
mathematical terms

FINITE

classified as a hyperbolic

is

equation

[3, 4].

ELEMENT FORMULATION

For the body idealized as one-dimensional, we use


and linear approximation as

= N u, + N u = [N]{q],
N = s, and s = (x x
u

where N,

5,

t )/l.

line

elements (Fig. 10-1)

(10-3)

Step

Derive Element Equations

4.

This step can be achieved by using the variational (Hamilton's) principle


and the principle of virtual work [4, 5]; these two are essentially statements
of the same phenomenon. Here we shall illustrate the use of the virtual work
principle [4], which is somewhat easier to understand for the introductory
treatment of the dynamics problem.
Equilibrium of the dynamical system at an instant of time requires
satisfaction of the virtual

{x f8{*x

JJJ

work equation

W=

[4, 5]

{F x Yd{u}dV

JJJ

JJJ

{Px Yd{n}dV,

where S denotes a small virtual change or perturbation and Fx


lent body force per unit volume caused by the inertial effect,

Fx
The

(10-4a)

=-P dt

is

the equiva-

(10-4b)

quantities required in Eq. (10-4a) can be evaluated as

6x

dx

(10-5a)

1]

or
{}

[B]{q]

duy
dt

du 2
\dt)
(10-5b)

[N]{q],

and
{d

uA

dt

d u2
[dt 2

=
since the

Af-

(10-5c)

[N]{q};

are functions of space coordinates only, the time derivatives

apply only to the nodal displacements.


Substitution of

(JJJ

{e}, {u},

and

{u} into

[BF[C][BJ^){q}

-W
+

226

Eq. (10-4a) leads to


r

(JJJ

{<5qf

JJJ

P[W[^W)W
[N]{P^K.

(10-6)

One- Dimensional Stress Wave Propagation

Chapter 10

Since {dq} represents arbitrary virtual changes,

JJJ

[BF[C][B]^K{q}

-JJJ

p[W[K]dV{q]

we have

{X\ T

JJJ

F*W

(10-7a)

227

or

M{q}

We

[m]{q]

= {Q}.

(10-7b)

we could have obtained these results by using Hamilton's


and by properly differentiating the associated variational function with respect to the components of {q} and equating the results to zero.
The terms [k] and {Q} have the same meaning as before [Eq. (3-28)]. The
additional matrix [m] is called the element mass matrix. For the line element
with uniform area A and constant P r at nodes, the matrices are
note that

principle

[5]

-1

M = AE

(10-8a)
1

AIP X (Q

{Q(01

(10-8b)

and

w-^f'f/ki

s]ds

pAl

(10-8c)

The matrix

in Eq. (10-8c)

is

called consistent because

it is

consistent (variational) principle; in other words, the mass

may

Often,

it

Apl

divided equally

among

Because the lumped matrix

On

the

two nodes:

Apl

"1

0"

(10-8(1)

2
is

diagonal,

it

can offer computational advan-

the other hand, the consistent matrix can be

mathematical analyses. Detailed consideration of


scope of this
Step

5.

more accurate
is beyond

this aspect

for

the

text.

Assemble Element Equations

The element equations can now be added such

that interelement con-

tinuity of displacements (and accelerations) are ensured at

Then

distributed to

first

tages.

is

term on the right-hand side of Eq. (10-7a).


be convenient to approximate it as lumped, where the total mass

the nodes consistent with the

is

derived from the

for the three-element discretization (Fig. 10-2)


[K]
(4

{r}

4) (4

1)

[M]
X 4)

(4

(f)

(4

1)

{R(0},
x 1)

(4

common

nodes.

we have
(10-9)

One- Dimensional Stress Wave Propagation

228

L = 30

Chapter 10

cm
j|1

kg
t

A A A

Figure 10-2

Mesh

for one-dimensional

medium.

where [M] is the assemblage mass matrix and the other terms have the same
meanings as in Eq. (3-32).
Equation (10-9) represents a set of (matrix) partial differential equations
and is the result of the discretization of physical space in the first phase. The
time dependence is contained in {?} = d 2 {r}/dt 2 and the next phase involves
,

discretization in time in order to find solution in time.

As

in the case

of Eqs. (5-21),

(8-7),

and

(9-9),

Eq. (10-9) represents a time-

dependent phenomenon. The difference is that here we need to consider


second derivatives instead of the first derivative in the other three equations.

Time Integration

we assume that the acceleration {r} varies


between a time step At from time level / to t + At [Fig. 10-3(a)]. As
a result, the velocities will be approximated as quadratic and the displacements
as cubic [Figs. 10-3(b) and (c)]. As presented in Refs. [5] and [6], this assumpAs

a simple approximation,

linearly

L near approximation

Quadratic

/-r

n
t

1t

+ At

(b)

(a)

t
(0
Figure 10-3 Approximations for time integration, (a) Acceleration,
(b) Velocity, (c) Displacement.

One- Dimensional Stress Wave Propagation

Chapter 10

229

tion leads to finite difference discretization of Eq. (10-9) as

M'U*

(10-10a)

{},-*,

where
[K]

{R},.^

= [K]-^L[M].

(10-10b)

- ^[M]({r} -

{R},.*

Ar{f} r

-^(Ar) 2 {r}

(10-10c)

).

starts from time t = 0, {r}


{r}
and {r)
boundary (displacement) and initial (velocity and
acceleration) conditions/ Thus the vector {R}a.. for the first time step t =
At = At is known since {R} is known at all time levels. Thus we can solve
Eq. (10-10a) at t = At and compute displacements at At as {r}^. Once the
{r} Af are known, velocities and accelerations at t = At are found from [5, 6]

Because the time integration

are

known from

l*U

= 5<M* -

m =

Wo)

is

2{f}

^{f}

(10-1 la)

^(M* - Wo) - 5 Wo - m*

The procedure can now be continued


nAt, where nAt

the given

(io-i lb)

for subsequent time steps 2At, 3 At,

the total time period for which the solution

is

desired.

and t At may not be


the best procedure for the time integration. We have used it mainly for a
simple illustration. Alternative and mathematically superior schemes are
available and used for practical problems ([4] and the bibliography).

The assumption of

Example

linear acceleration

10-1

Consider the propagation of an


(Fig. 10-2).

between

Assume

wave

elastic

in a bar discretized in three

elements

the following properties:

Element length.

10cm;

/ ==

cm :

==

==

Density,

= = 10"

Wave

Cx

Element area.
Elastic

modulus.

velocity,

lOOOkg'cm 2
11

= Jf

These simple properties are chosen only to

cm 4

kg'sec :

109

cm

illustrate the

sec

procedure; they

may

not

necessarily represent field situations.

An

approximate

size

of the time step can be found as

At

=^=
10
c
7

10-6 sec

(10-12)

This size

is

often called the characteristic time step. Substitution of these

into Eqs. (10-8a)

and (10-8b) and an assemblage

[Eqs. (10-9)

and

numbers

(10-10)] leads to

One- Dimensional Stress Wave Propagation

230

10
_

x 10~ n

[M]

2-1
0-1 2-1
0-1
"2
10 0"
x 10 14 10
14

-1

x 100C

[K]

0"

-1

Chapter 10

>

1.

>

_0

2_

and hence
1

-1

-1

-1

-1

"

[K]

10 2

-1
2

-1

-1

INITIAL

is

zero at

all

The

initial
/

2_

10 0"
14 10
14

10 2

(10-13)

_0

1_

is

2_

that the displacement at the fixed end of

times; that
Wl (0,

zero at

CONDITIONS

The boundary condition


(Fig. 10-2)

0"

"2

-1

-1

_0

0"

BOUNDARY AND

-1
1_

-1
10 2

6 x 10"
x 10
10" 6 x 10~ 6 x 6
11

-1

"2

0"

node

is,

t)

(10-14a)

0.

conditions are that displacements, velocities, and acceleration are

0:
v(x, 0)

u{x, 0)
ii(x,

0)

=
=
=

0,

(10-14b)

0,
0.

An initial condition defines the state of the body at the start or initiation
ing in terms of the displacement and/or

An external

force

is

its

of load-

derivatives.

assumed to be applied

at the free end,

node

4, as

a constant

force

F4 (t) =
With

(10-14c)

kg.

these conditions, the load vector {R}

is

~2

(R}r + Af

[R}f+Af

6 x lO" 11 x 10
10" 6 x 10" 6 x 6

_0

10- 6 {f}, +-i-x

10- 12 {r},).

0"

2_

(M,

(10-15)

One- Dimensional Stress Wave Propagation

Chapter 10

Using Eqs. (10-13) to (10-15), we can write Eq. (10-10)

10

At

At as

(10-16)

<

ir 3

12.

.0

lw 4

Ar

introduce the boundary condition

u { (At)
in

10 ur
14 10 w? =
14
1

Now we

0"

2
10

at

231

Eq. (10-16), which leads to

T600
600

(10-17)
300.

Solution of Eq. (10-17) by Gaussian elimination (with u

prescribed) gives

u'
1

(10-18a)

U3

uj

Ar

Substitution of Eq. (10-1 8a) and the

and

(10-1 lb) allows

300 }

initial

computation of

conditions [Eq. (10-14b)] in Eqs. (10-

velocities

and accelerations

at

a)

At as

|o\

dt

()

fl

du 2
dt

du 3

>

-{

>

>

1Q- 6

2<

(10-18b)

dt

du 4

\300

dt

At

ol

/ o

and
d 2u

dt*
\

d 2 u2
dt 2

d 2 u.

6
)

10" 6 x 10" 6

J=-A
10" 6

>

-2<

>-(10-18c)

dt 2

d 2 uA

300

dt 2

Ar

Equation (10-10a) can

now

be solved for time level

/o

2At, 3At, and so on.

We note here that the foregoing numerical calculations are meant only as an illustration of the procedure

and not necessarily

For the latter, one


and temporal meshes.

as acceptable solutions.

needs to program the procedure and select

optimum

spatial

Example 10-2

Now we

present results from a problem in one-dimensional

solved by

Yamada and Nagai

The

[7].

wave propagation

properties of the problem (Fig. 10-4) are as

follows:

L or

Total length of bar,

Number

of nodes

Number

of elements

500

mm

Area of cross section assumed

arbitrarily

elasticity,

Density,

Wave

= 51;
= 50;
= 10mm;
= mm 2
= 20,000 kg/mm 2
p = 0.008 kg msec 2 /mm 4
c x = 5000 mm/msec;
At = ljc x = 0.002 ms.

Length of each element,

Modulus of

velocity,

Time increment.

=:

500

mm

*-

mm/ms

Figure 10-4 One-dimensional wave propagation

[7].

BOUNDARY CONDITIONS
t)

(A, t)

u(h,t)

u(0,

du

0.

mm/ms.

or

^-(h,t)

t.

Ox

The
10/5000

results in

is,

after

analytical solutions [Figs. 10-5(a)

mations, respectively.

mass

sistent

at time / = 0.08 ms (for At = I/c x =


40 time steps, are compared with closed form

terms of particle velocity

0.002 ms), that

It

and

(b)] for consistent

and lumped mass approxi-

can be seen that for the characteristic time step, the con-

yields exact solution,

whereas the lumped mass formulation

is

not that

accurate.

The

size

of the time step can have significant influence on the numerical solu-

shown in Figs. 10-5(c) and (d) for both approximations for At =


Again the lumped mass formulation is not close to the analytical solution.
On the other hand, the consistent mass formulation shows oscillations and inaccuracies, particularly in the vicinity of the wave front. The aspect of numerical
stability and comparison of lumped and consistent masses is wide in scope, and the
reader may consult the references and the bibliography.
tions.

0.5l/c x

232

This

is


2.0

Analytical solution

"""Numerical solution

ms

= 0.08

1.0

HMXXWJSOOO.

21

11

31

41

51

41

51

41

51

41

51

Node number
(a)

Analytical solution

000 Numerical solution

2.0

- 0.08 ms

:
o

1.0

......

oo

o
a

11

21

31

Node number
(b)

Analytical solution

000 Numerical solution

2.0

t= 0.08 ms
1.0
<

.a

,E

oo

11

31

21

Node number
(c)

Analytical solution
~E

000 Numerical solution

2.0

>

= 0.08

ms

1.0

>

I
-

o
11

21

31

Node number
(d)

Figure 10-5 Results for wave propagation in bar

mass: At
At = 0.5

=
I

l'c x

cx

(d)

[7]. (a)

Lumped mass: At = I c x (c)


Lumped mass: At = 0.5 Hc x

(b)

Consistent

Consistent mass:

233

Damping

Most natural systems possess damping. It can be easily included


dynamic equations; the resulting element equations will then be

[k]{qj

where

[c]

material
([4]:

damping

the

is

is

[c]{q}

[m]{q}

(10-7c)

[Q(t)\.

matrix. Determination of

and

a very important topic

discussed

is

in the

damping properties of
various publications

in

the bibliography).

PROBLEMS
10-1.

Compute

mass matrix for the quadratic approxima-

the element consistent

tions [Eq. (3-40)]:

where

\L(L -

),

( 1

- L2

WL

uQ

)w 2 ,

denotes the middle node.

Solution:

Apir\

[m]

UL

J.,

L(\

Apl

10-2.

Compute

1)1
2
)

\[L(L

(1-L

1)

L{L-\)]dL

2
)

1)J

8-4"

16
8

64

L-4

120

-L

(1

8
16.

the element consistent

mass matrix for cubic (Hermitian function)

u:

(1

3s 2

-5 2 (3

2s 2 )u

ls(s

l) 2

^
OX

-2s)u 2 +ls 2(s

- 1)^-

Solution:

[m]

10-3.
10-4.

22

54

22

13

-3

-22

Apl
420

54

13

156

-13

-3

-22

Perform computations for three time steps

in

4.

Example

10-1.

Consider a single-element discretization which can be replaced by a spring-

mass system.

It is

subjected to a forcing function

Specialize Eq. (10-10a) for this single element

time steps.

234

-13"

156

Assume

properties as in

Example

Px

as

shown

in Fig. 10-6.

and obtain solutions for two

10-1

Compare numerical

results

One- Dimensional Stress Wave Propagation

Chapter 10

235

P.tlii
m

Figure 10-6

with the closed form solution from


U{t)

where p 1

k/m, where k and

respectively.
10-5.

The

initial

=p
y(l -

cos pt),

are the stiffness

conditions are: u(x, 0)

and mass of the

it(x,

0)

spring,

0.

Prepare a computer problem based on Eq. (10-10) and solve Probs. 10-3 and
10-4.

REFERENCES
[1]

Love, A. E. H.,

Treatise on Mathematical Theory of Elasticity, Dover,

New

York, 1944.
[2]

[3]

Timoshenko,
York, 1970.

Carnahan,
Wiley,

[4]

New

Desai, C.

S.,

B.,

and Goodier,

Luther, H.

[6]

N., Theory of Elasticity, McGraw-Hill,

and Wilkes,

J.

New

O., Applied Numerical Methods,

York, 1969.

and Christian,

S.,

Engineering, McGraw-Hill,
[5]

A.,

J.

Numerical Methods
York, 1977.

J. T., (eds.),

New

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

to the Finite

in

Geotechnical

Element Method, Van

Wilson, E. L., "A Computer Program for the Dynamic Stress Analysis of
Underground Structures," Report No. 68-1, University of California, Berkeley,
Calif., 1968.

[7]

Yamada,

Y.,

and Nagai,

Y., "Analysis of

One-Dimensional Stress Wave by

the Finite Element Method," Seisan Kenkyu, Journal of Industrial Science,

Univ. of Tokyo, Japan. Vol. 23, No.

5,

May

1971, pp. 186-189.

BIBLIOGRAPHY
Archer,
Div.

J. S.,

"Consistent Mass Matrix for Distributed Mass Systems,"

ASCE,

Vol. 39,

No. ST4, Aug.

Clough, R. W., "Analysis of

Structural Vibrations

and Dynamic Response," in


Yamada, Y., and

Proc. U.S. Japan Seminar, Tokyo, 1969 (Gallagher, R. H.,

Oden,

J.

T. eds.), University of

/. Struct.

1963.

Alabama

Press, Huntsville, 1971.

One- Dimensional Stress Wave Propagation

236

Chapter 10

Desai, C. S. (ed.), Proc. Symp. on Appl. of Finite Element Methods


Waterways Expt. Station, Vicksburg, Miss., 1972.

Desai, C. S.

(ed.),

Proc. Second Intl Conf. on

burg, Va., 1976, Vols.

Desai, C.

S.,

and Lytton, R.

Parabolic Equation,"
Idriss,

I.

I, II, III,

M., Seed, H.

B.,

Num. Methods

Num. Methods

Geomech., Blacks-

of Finite Element Schemes for

Eng., Vol. 9, 1975, pp. 721-726.

Seriff, N., "Seismic

Finite Elements," /. Geotech. Eng. Div.

in

Geotech. Eng.,

1976.

L., "Stability Criteria

Int. J.

and

ASCE,

in

Response by Variable Damping


Vol. 100, No. GT1, 1974, pp.

ASCE,

1-13.

Kreig, R. D., and Key, S. W., "Comparison of Finite Element and Finite Difference Methods," Proc. ONR Symp. Num. Methods Struct. Mech., University
of

Illinois, Sept. 1971, (Fenres, S.J.,

brich,

W.C.

eds.)

Academic

Press,

Perrone, N., Robinson, A.R., and Schno-

New

York, 1973.

Newmark, N. M., "A Method of Computation


ASCE, Vol. 85, No. EM3, 1959, pp. 67-94.

of Structural Dynamics," Proc.

Nickell, R. E., "Direct Integration Methods in Structural Dynamics,"


Mech. Div. ASCE, Vol. 99, No. EM2, April 1973.

/.

Eng.

TORSION

INTRODUCTION
Until

now we have

considered problems that can be idealized as one-dimen-

They involved only

sional.

unknown

only one

at

elements and, except for

line

beam

bending, had

each node. Next, we advance to problems that can be

idealized as two-dimensional.

To

We

start with, let

choose torsion

unknown

us consider a problem in stress-deformation analysis.

first

problems (Chapter
the torsion problem.

field

We

because

at each point (node).


12), are

shall illustrate here a

it

involves only one degree of freedom or

number of other problems,

often called

governed by equations similar to those for

number of formulation procedures

displace-

and hybrid and mixed, which are based on the


energy principles. For all these formulations, we approximate the behavior
of a bar [Fig. 11 -1(a)] subjected to torsion by considering essentially the
ment,

stress or equilibrium,

behavior of the cross section of the bar [Fig. 11 -1(b)]. In the semi-inverse

method of Saint- Venant,

it is

assumed that the twisting of the bar

is

composed

of the rotations of the cross sections of the bar as in the case of a circular bar

and of warping of the cross


sections

[1, 2].

sections; the latter

As a consequence, no normal

is

constant for

stress exists

all

cross

between the longi-

is no distortion of the planes of cross


and hence the strain components e x yi and y xy vanish, and only
pure shear components y x2 and y y2 will remain.

tudinal fibers of the bar. Also there


section,

237

238

Torsion

Chapter 11

y, v

(b)

(a)

Warping

(c)

Figure 11-1 Torsion of bar. (a) Bar subjected to torsion, (b) Cross
section of bar. (c) Warping.

For the two-dimensional approximation, we

shall consider discretization

involving triangular elements (Fig. 11-2); quadrilateral elements are treated


in

Chapter

12.

For the triangular

known can be adopted,

case, a linear

approximation for the un-

while for the quadrilateral isoparametric elements,

a bilinear approximation can be chosen (Chapter


first

two

12).

This constitutes the

steps in the finite element formulation.

TRIANGULAR FINITE ELEMENT


Before considering subsequent steps,
ties

we

shall first detail

some of the proper-

of the triangular element.

As

discussed in Chapter

coordinate systems for

3, it is

finite

advantageous to use the concept of local

element formulations. For the triangular

ele-

ment, the local or area coordinates are often defined in terms of component

Chapter 11

239

Torsion

(x 3 .y3

(0,0,

P(x, y)

(L,,L 2 .L3

1)

u=o

U =0

Figure 11-2 Triangular element.

areas,

A l9 A 2

and A 3

local coordinates

(Fig.

L l9 L

Then

11-2).

-\-

L2

-\-

L3 =

nondimensional form, the

Since

in the

and L are defined as

1,

(11-1)

1,2,3.

there are only

two independent

corresponding to the two global coordinates x and

between the global and the local coordinates

y.

local coordinates

The

relationship

given by

is

(11 -2a)

=L

3
x

+L

+L

y3

= 2= Ajn
i

*
where [N]

is

= m*&

(ll-2b)

the matrix of interpolation functions

the local coordinates

{x n } T
is

it

which are the same as

these are plotted in Fig. 11-3.

[x M

x2

x3

y2

y3

the vector of nodal coordinates. Equation (11-2) represents linear variations

for coordinates at

any point P(x,

y).

Figure 11-3 Distribution of interpolation functions

1, 2, 3.

240

Chapter 11

Torsion

The

inverse relation corresponding to Eq. (11-2)

is

N =L =

2^(^23

M+

a iy)>

N =

2
2^( ^3i

b2x

a 2 y),

J_(2A l2

b3x

a 3 y),

L2

AT 3

2A

(ll-3a)

or
~

AY

"2^23

6,

2^3

62

flj

'

ATa

A ,.
7

where 2A

=a

b2

of the triangle;

and the

origin

A 23

_2^ 12

= ^63

a2b3
is

A3

3.

#3^

=a

'

(ll-3b)

.*.
2

the area of the triangle

a b2
x

whose

and so on; and a and b are given


{

is

the total area

vertices are

nodes

2, 3

as differences between

various nodal coordinates

a2
a3

=x - x
=x -x
=x - x u
2

is

b3

b2

The unknown, represented by a

where [N]

=y - y
=y -y
=y - y
2

ly

general symbol

= A^w, + N
= [N]{q},

u2

+N

(H-3c)

w,

can be expressed as

u3
(ll-4a)

the matrix of interpolation functions and yields linear variation

of u over the triangle. Since [N] in Eqs. (11-2) and (11-4)


call this

element an isoparametric triangular element.

FINITE

ELEMENT FORMULATION

is

the same,

we can

Displacement Approach
First

torsion

we

consider the displacement approach. According to Saint- Venant

[1, 2], in

body forces, the differential (Laplace) equahomogeneous and isotropic bar can be expressed

the absence of

tion governing torsion in a

as

dx 1

dy 2

(11-5)
'

where x and y are the global coordinates in the plane of cross section of the
bar (Fig. 11-1). The warping function, y/, is related to the displacement w
in the z direction, and is assumed to be constant along the length of the bar

Chapter 11

241

Torsion

and function only of x and y

(Fig. 11-1):

w=6y/(x,y)
where

is

moment,

(ll-6a)

>

the twist of the bar per unit length under the applied twisting
[Fig.

l-l(a)].

The warping function denotes a measure of warpThe other two displacement

ing of the cross sections of the bar [Fig. 11 -1(c)].

components, u and

v,

are given by

= 6zy,
= Qzx.

u
v

The boundary condition along the

(11 -6b)

surface of the bar

is

given by

(->)-( + *)=
where

is

measured along the surface or boundary. Equation (11-7) indicates

that the shear stress normal to the boundary, r z


Step

(11-7)

is 0.

Gradient-Unknown Relation and Constitutive Law

3.

By using Eq.

the gradient or strain-warping function relation

(11-6),

is [2]

'dw

dx
[J

where

{e}

[y xz

du

dw
y

yz ]

+
is

(% - >)

dz
>

,dy
r

<

dv
dz

>

(11-8)

<

+
*)l

the vector of (shear) strain components.

In the general three-dimensional problem, there are six nonzero

compo-

However, for the foregoing two-dimensional idealization,


there are only two nonzero shear-stress components, x xz = r zx and r yz = tzy
corresponding to the two shear-strain components. The constitutive rela-

nents of stress

[2].

tion can be expressed as

r xi

Gy X2

Gy yz

.0

(11-9)

where

{<r}

[t xz

r yz ]

is

GJ|

the vector of (shear) stress

components and

is

the shear modulus.


Step

4.

Derive Element Equations

For the displacement approach, the potential energy function correspondis given by [3, 4]

ing to Eq. (11-5)

where h

is

the length of the bar, and

denotes area.

242

Chapter 11

Torsion

According to the general expression [Eq.


y/,

(11 -4a)] for linear variation

of

we have
y/(x, y)

= N y/ + N
= [N r ]{q r
1

2 y/ 2

+N

y/ 3

(ll-4b)

},

where

y/{x, y) is

the warping function at any point (x, y) in the triangular

element and {q^} r

[y/ i

y/ 2

y/ 3 ] is

the vector of nodal values of the warp-

ing function.

Requirements for the Approximation Function

As

discussed in Chapter

the approximation function in Eq. (ll-4b)

3,

should satisfy continuity within an element, interelement compatibility, and


completeness. Equation (11 -4b)

form of

linear

y/

which

a,

the transformed version of the polynomial


y/

as

a2x

a 3 >>,

(ll-4c)

continuous within the element.

is

The

is

approximation for

highest order of derivative in the energy function [Eq. (11-10)]

is 1;

hence the approximation function should provide interelement compatibility

1=0, that

warping functions. The linear approximawarping functions across interelement boundaries. This is illustrated in Fig. 1 1-4. Since y/\ y/\ and y/\
yt\
and since only one straight line can pass through two points, the variations
of warping function along the edge AA of element 1 and BB of element 2
must coincide. That is, at the common boundary, compatibility of y/ is

up

to order

is,

for

tion indeed provides for compatibility of

Figure 11-4 Interelement compatibility.

^2 = warping

at

node 2 of element

1,

and so on

Chapter 11

Torsion

Note

fulfilled.

243

that for linear approximation this does not necessarily imply

interelement compatibility of higher orders, such as the

first

derivatives.

body
motion (term a ) and constant state of strains or gradients of y/ (terms a : .v
and a : v\ it is complete. Notice that in this problem, there exist two such
states: dy/ dx and dy/ dy. Furthermore, it contains all terms up to the order
Since the approximation function in Eq. (ll-4c) provides for rigid
:

polynomial expansion for the two-dimensional problem. This

in the

idea can be explained by using the polynomial expansion represented (by


Pascal's triansle) as follows:

order or degree

(ri)

of polynomial

n=0
x

To

x zy z

*r

approximation

see that for the linear

and including n

to

xy 2

x 2y

x~

Wc can

xzy

x3

up

XV

constant

Linear

Quadratic

Cubic

Quartic

in Eq. (1 l-4c), all the three

terms

are provided.

find the values of the derivatives in Eq. (11-8),

we use

the following

rule of differentiation:

_
^
sbL
" dx
dx
dx dX

2Ll

du_

dy

dX-

dy/

dy

dX

_ dX

dy

3L

dX

dx

$2i

dy/

_ dX

dX

dy

For example, by differentiation of


and (ll-4b). we obtain

(/

1, 2,

^L
dX

dx

iiai
Ui-na;

dy/

(1 1-1

dN

lb)

3)

and

y/

with respect to x

in Eqs. (11-3)

= Jx.~T4 i2A:
-

>

{2a ^
jx ta
d_

dx J2(2A ::

=
Since

h*>

A Z3

b X
--

~ " V)

^*^ ~ N** ~
dk (Nii ~

b*x

b,x

- a.yi^iX-U: - X

azy)1

+&> + &>

etc.,

y/ t

(i

y/ :

NiWi)

-X

y/ 3 )

(1M2a)

are constants, they

differentiation. Since the

Nzz

*****

do not contribute to the results after


3) are values at the nodes and not

1. 2.

244

Chapter 11

Torsion

functions of

and

they also do not contribute to the differentiation.

(x, y),

Similarly,
dy/

Y1
-ar.+a*
2A Y2
fy~2A
'

Equations (11-1 2a) and

(1 1-1

2b) are combined in matrix notation as


-

>

dy/

(ll-12b)

2A YZ

'

2A

~b,

b2

b3

J*\

a2

a3_

i
<y/ 2

Vs.

dy,

(ll-12c)

Here

[B]

is

the strain-warping function transformation matrix.

Hence Eq.

(11-8)

becomes

-By

OX

to =

(ll-13a)
1

Ox

dy\

By
(ll-13b)

Ox

Now
n,

Eq. (11-10) can be expanded as

Gh6 2

pas' -*+']+ [+*+*]*


(ll-14a)
dy/

GhO 1 [[
2

+b

JJ

dx

[dy/

dyf]

Idy

dy] dy

dy/

JSx

dy/

dy

w
(ll-14b)

\dxdy.

x]

Substitution from Eqs. (11-12) and (11-13) into Eq.

n,

Ghd

r
{q,} [B]^[B]{q,}

(ff

[y m

xm ]

)dxdy,

(1

l-14b) gives
t

-y>

2{qwY\Bf

(11-15)

Chapter 11

= (*j +

where x m

mean

Torsion

values. It

x2

= {y + y + ^

x 3 )/3 and y m

make

not necessary to

is

245

this

3 )/3

assumed

are the

assumption, however.

We

can

x and y from Eq. (11-2) and pursue the derivations, which will
be somewhat more involved. Now we invoke the principle of stationary
(minimum) potential energy hence
substitute for

dU p

*n,

dn p

(11-16)

dy/ 2

dUp
dy/ 3

which leads to

GhO 2

[B]

Note that the

[B]dxdy[%}

term

last

Equation

results.

JJ

is

JJ

[Bfi

in IT^, being constant,

(11 -17a)

_M

dxdy.

(1

l-17a)

does not contribute to the

can be written in matrix form as

= (<U

fcjfar}

where [kj

GhO 1

(ll-17b)

the element stiffness matrix:

[kj

Ghd 1

[B]

JJ

[B]dxdy

(ll-17c)

and {Q^}

is

the equivalent nodal load vector

m=
They can be evaluated

Ghe>\\{w\_

6,-

(ll-17d)

\dxdy.

as follows
~b t

a,

o2

a2

6,
L^3

Since a and

ym

a
"3.

LM

b,

dxdy.

are constants, the integral equals the area


[k r ]

(ll-18a)

hence

Gh0 2 A[B] T [B]

or

G/z0

fc

6|

4^
sym.

a\

b2b 3
b\

+
+

a 2 3
a\

(ll-18b)

246

Torsion

Chapter 11

and
b

>

0\

b2

a2

GhO-

[Q,

jy

y,

dxdy

x,

(ll-19a)

a3.
a

Gh0 2 A
2A

y,

b2

x,

b3

b,y

Gh6

Tl

b 2 y*
[b 3 y r

a xm
l

a 2 xm

a,x m

We note that the term GhO

is

(ll-19b)

common on both

lem

governed by the Laplace equation [Eq.

is

Step

and can
and the prob-

sides of Eq. (1 l-17a)

be deleted, because of the assumption that the bar

is

isotropic

(11-5)].

Assembly

5.

Example

Torsion of Square Bar:

11-1.

Warping Function Approach


Equations

(1

l-18b) and

(1

l-19b) can

now

be used to generate stiffness matrices and

For instance, Fig. 11 -5(a)


shows a square bar, 2 cm x 2 cm, discretized in four elements. Figure 11 -5(b)
shows the four elements with their local and global node numbers. For this example
assume the following properties
load vectors for

all

the elements in a discretized body.

G =
h

The area of each


illustrate the

The

A =

triangle

=
=

cm

N/cm 2

cm,

rad/cm.

2
.

These properties are chosen

in

order to

procedures; they do not necessarily refer to practical problems.

quantities a t

and b for the four elements are as follows


{

Element 1:

x2 =
- 1 = -1
2 = -1,
b = y2 - y 3 =
= -1,
a2
62=^3-^1=1-0 = 1,
xi x =
- = 0,
= 2,
b = vi - y 2 =
03 .= x 2 Xi =2
x m = (0 + 2 + l)/3 :l,^=(0+0 + l)/3=|.
a

=
=

x3

(ll-20a)

and

Element 2:

=
a2

a3

and x m

(1

0)/3

h =0 b2 = 1 - 2
1

1,

= -1,
= 0,

= i, ym =

^,3=2
(2

l)/3

(11 -20b)

Chapter 11

Torsion

0~(O,

247

(2,0)

0)

Global node

(a)

Local node

dement

(b)

Figure 11-5 Torsion of square bar. (a) Square bar and mesh,
(b)

Elements with local and global nodes.

Element 3:
a,

a2
tf 3

and * m

(2

= 1 -2 = -1,
= 2- 1 =1,
=2-2=
+ l)/3 = f,

b2

= 1 -0 = 1,

1,

(ll-20c)

= - 2 = -2,
+ 2 + l)/3 = 1.
63

0,
>>

bi

(0

Element 4:
fll

a2
^3

and x m

(2

= i -0=
=2 - =
1

1,

6j

1,

62

=2- =
= -2= -1,
1

1,

= - 2 = -2,
6 = 2 - 2 = 0,
+ l)/3 = 1, ym = (2 + 2 + l)/3 = f.
3

(ll-20d)

248

Torsion

Chapter 11

Substitution of these values into Eqs. (ll-18b) and (ll-19b) gives the following:

Element

1:

Global

(ll-21a)

Element 2:
Global
Local
1

^3

^1

y\ = vi

-2

L-2

yi-

(ll-21b)

Element 3:

^2

Global
Local
i

^1
2

(ll-21c)

L-2

Element 4:
Global
Local
i
1

(ll-21d)
3

-2

-2

we have shown the relations


between local and global node numbers. For the particular node numbering chosen
in this example the coefficients of the element matrices are the same. In fact, even if
a different numbering is used, if the elements are of equal dimensions, [k ] can be
generated once it is obtained for one element.
The assembly procedure is carried out by observing the fact that the values of
y/ at common nodes are compatible. Thus, by adding the local coefficients in

In Eq. (11-21) the superscript denotes an element, and

appropriate locations in the global relation,

we

obtain

249

7br^ K7/2

Chapter 11

>\

Global

"

(2

(-2 -

(-2

2)

(2

2)

(2

(2

_(-2-

2)

-2 -

- 2)
+ 44

(-2

2)

:-2-2) (-2-2)

2)

^2

2)

(-2-2)

2)

4
5

2)

(4

^4
4)J

|-i=o

'

+ f=4

(ll-22a)

-f + f =
+ 2-2 +

,-2

or
'

-4" Vii

-4
-4
-4

4
4
4

_-4

-4

-4

-4

16

V2
1

<^3

-16

(11 -22b)

>

^4

16_ Vs,

0,

or
:K]{r]

where

[K]

is

functions,

{R },

(11 -22c)

the assemblage stiffness matrix, {r}

and [R]

is

is

the vector of global nodal warping

the assemblage nodal forcing parameter vector.

Since the values of the warping function are relative, the boundary conditions

can be introduced by assigning datum value of y/ to one of the nodes. Then we


obtain relative values of nodal warping functions. For example, assume y/^

Then

the modified equations are obtained by deleting the

Eq.(ll-22b) (see Chapter


"

-4"

>

-4

-14

= cm

Step

6.

/rad (given),

y/ 2

-8
(ll-22d)

<

-4
16.

y/si

Solution of these equations yields the primary


y/ l

0.

in

8'

>2

-4

row and column

3):

_-4

first

shall

=2,

y/ 3

unknowns
-2,

as

y/ 4 =

o,

y/s

Secondary Quantities

The secondary quantities can be the shear stresses and twisting moment.
Equation (11-9) can be used to find the shear stresses in the four elements:

250

Torsion

Element

Chapter 11

1:
dy/

*xz
>

-y m

dx

G9<

>

G0<

dy/

*m

idyl

&1V1

b 2 y/ 2

a,y/,

a2^

6 3 ^:

(11-23)

^V:

Substitution from Eq. (ll-20a) then gives


\? X2

0+1X2
-1x0-1x2 + 2

if-lX

-i

HJH1
Element 2:

Element 3:
Txi
T...

Element 4:

The shear

stresses are plotted in Fig. 11-6.

TWISTING

MOMENT

The expression

Mt

for twisting

G6

moment

y
JI i~ fc

Often

M /G6
t

is

for the twisting

w ^ x2Jr

called the torsional constant

moment

bar

in the

is

given by

y2 dxdy

and

'

[2, 3, 4]

(11_24a)

is

used to express solution

in the bar.

we assume x m and y m as average constant values associated with the


two terms in the integrand and since dy/jdx and dy/jdy are constant for

If
first

Chapter 11

Torsion

251

the linear approximation, the integral in Eq. (ll-24a) can be approximated


as

- |, *. - |, [-y.A$l + XmA(gj

+ *. + ,)}

(,1.24b)

where [Eq. (ll-34a)]

xm

-7-(x\

-g-C^?

+ y\ + ^3 + 7l72 + ^2^3 + ^1^

*1

x\

XjX-2

*2* 3

Xi* 3 )

and

is

the total

number of

elements,

the contribution to twisting

3 ).

denotes an element, and

tm

denotes

moment by element m.

V'V
Stresses

shown

at

centroids of elements

cm

Figure 11-6 Plots of computed shear stresses: warping function

approach.

The
twisting

total twisting

moment can now be found

moments. Substitute

for dy//dx

as the

and dy//dy from Eq.

sum of element
(11-12),

coordinates of the nodes of each element lead to the following:

Element 1:
dy/

dx
dy/

i(-l

x0+lx2 + 0x0)=l,

^(-1 x

0-

^=-4x1 +

x (-1)

x 0)- -1

+ |= -f

and the

252

Chapter 11

Torsion

Element 2:

^L

-1x0-2x0] =

J[_l x (-2)

^ = }[lx (-2) M

tl

= -\

0]

1,

-1,

l+Jx(-l) + |=~t + f.

Element 3:

^ = J(lx2+lxO-2xO)=l,
= -1,
x + x
^ = J(-i x
2 4-

0)

Element 4:

* =
jS!

Afr

iKl

JI1

-1

X (-2)

= i[lxO-Mx
= 47 _

24
6

(-2)

= 1

+ X 0] =

1,

-1,

-1-

24.

0]

Therefore

= M + M + A/ + Mu
= _^(4 + 4 + 8 + 8) -f i(8 +
= _^ j_ M = _g + 10.6666
= 2.6666 N-cm.
(l

24

24)

COMPARISONS OF NUMERICAL PREDICTIONS


AND CLOSED FORM SOLUTIONS
Twisting

Moment

Based on the stress function approach presented subsequently, a closed


form solution for the twisting moment for a square bar is given by [2]

M, ~0.1 406(70(2 a) 4

(11-25)

>

253

Torsion

Chapter 11

or

Mi

where a

~ 0.1406(2*)*,

is

^-0.1406(2 x

~ 0.1406 x
~ 2.250 cm
The value computed from
the error

form value of the

the half width of the bar. Hence, the closed

torsional constant

l)

16
4
.

the finite element procedure

2.666; hence

is

is

em)r

inn
= 2.2500-2.6666 vXl0

12500

18.5%.

Shear Stresses

From

the stress function approach (see next section), the closed

form

values of the components of shear stresses, Eq. (11 -28a) below, can be

obtained from the following expression for closed form solution for stress
function in a rectangular bar

32G0c'-

9*

7l

rt

d(p

- dy'

=1,3,5,..

\6G6a
n1

[2]

^lL

n3

=]1.37s,...

cos h(nny>-2a) -\

CQS

cos hinnb 2d)\

sin^r2.]
^(-l)^
n
h{nnb 2d)
1

0S

cos

(nnx\
\ 2a /

W
\2a

(11 -26a)

(11 -26b)

_dg> _ \6G0a

r*

(7,X

7T^

!Lzl1

n=l,3,5,...

cos

h{nnyad)

~\

cos h(nnb;2a)\

/nnx\
^ \2a

(ll-26c)

where

a:

and y are measured from the center point of the

bar,

and a and b

denote half the dimensions of the cross section of the bar, respectively.
In the finite element analysis, we assumed linear approximation for the
warping function; hence, the distribution of shear stresses is constant at every

point in the element. This element can, therefore, be called the constant
strain or stress triangle (CST). If

we had used

a higher-order approximation,

shear stresses would vary from point to point in the element. In view of the

constant values for the shear stresses,


11-6, to attach

them

it is

often customary, as

at the centroids of the elements.

shown

in Fig.

For the purpose of com-

254

Chapter 11

Torsion

we compute

parison and error analysis below,

form or exact
must be understood
that this is only an approximation, since the computed shear stresses are
constant over an element as a part of the formulation, and are attached at
the closed

values of shear stresses at the centroid of an element.

It

the centroid only for convenience.

The

is important and wide in scope. Our purpose


by presenting comparisons between computed and
closed form solutions for some of the problems in this chapter. In making
these comparisons we have considered only typical elements and nodes.
For Example 11-1, let us choose element 3. The centroidal coordinates of
is

subject of error analysis

to introduce the topic

element are

this

= 0.667

Then use of Eqs. (ll-26b) and

and

= 0.000.

(ll-26c) gives closed

form values of shear

stresses as
tJ,

- 0.000 TV/cm

2
;

= 0.775 N/cm

r*

2
.

These and the subsequent closed form solutions are obtained by using n
in Eq. (11-26). The errors in the two components are then given by

25

error in xxx

and error

in z yz

= 0.000 - 0.000
= 0.00%,
Tyz r
Tyz
- r*
x

0.775

10

0.667

0.775

14.00%.

Comment: The foregoing comparisons show

that despite linear approxi-

mation, and the rather crude mesh, the computed results yield

realistic solu-

same order of magnitude. Here we used a coarse mesh mainly


to illustrate hand calculations. In general, however, one should use a finer
mesh, and the solution will improve significantly. For instance, with a finer
mesh, say with 200 elements, the computed value of the torsional constant
for the square bar would involve an error of less than 1 .0 %.
tions of the

STRESS APPROACH
The

torsion problem can be expressed by using the concept of stress function,

approach represents the same phenomenon as represented by


the warping function, but it is an alternative or dual representation for
<p

[1-7]; this

Chapter 11

Torsion

The governing

torsion.

geneous medium,

in

255

equation for a linear, isotropic, and homo-

differential

terms of (p, according to Prandtl

[1, 2], is

V^-W -^

where

is

26

(11-27)

the equivalent forcing parameter, which can be a prescribed

known

function of

(x, y).

The two nonzero shear

terms of

stresses in

>

Ty
<

between the

(ll-28a)

>

dcp

TL yz
relation

are

d(p

1 X2

The

cp

5iJ

stress function

and warping function

is

(ll-28b)

*-&+)
The energy function associated with Eq.

(11-27)

is

the complementary

energy, expressed as (see Chapter 3)

w =

n =u +
c

u -

w,

or
2

=JJi[ + (g)>^-JJe^'

*"*

or

dx

n =
2b

dxdy

'

Here
tial

/c is

the complementary strain energy,

of external loads, and

The unknown
T

[q s ]

[cp

cp

is

the

work of

for a triangular element

= Nl9l

(p

where

IT (20)prfjcrf>\

(11 -29b)

dcp

cp 2

cp 3 ].

-N
The

-N

2 cp 2

is

is

the complementary poten-

now

expressed as

[N s ]{q s ],

gradient-^? relation

(ll-4d)

is

dcp'

gj
{g}

2 <p 2

external loads.

>

dx
[B]{q,!

<

(11-30)

dcp

g,

Substitution of Eqs.

n =
c

(1

l-4d)

idyl

and

(1 1-30)

JJ {q s y[BY[D][B]{q s }dxdy

into Eq. (1 1-29) yields

jj (20)[NJ{q,}^,

(ll-31a)

256

Torsion

Chapter 11

where
1

P>]

is

the strain-stress matrix.

By invoking

the principle of stationary

(minimum)

complementary energy, we have


f'

sn c =

(ll-31b)

*S,-0.
3

or

JJ

[BY[D}[B]dxdy{q,}

(20)[N]Wy,

JJ

(ll-31c)

or

M{qJ =

(H-31d)

{Q,},

where [kj is the element property or flexibility matrix and [QJ is the
ment nodal forcing parameter vector. They are evaluated as follows

[kj

^jjlB7lB]dxdy

~b\

b\

4GA

^[BY[B]

+
+

b b2

a\

ele-

+a a
b + a a
bl + al

a a2

b b3

a\

b2

sym.

(11-32)

and
IN,

U-2JJ N
The

integrations in Eq.

formula

(1

1-33)

2 }

dxdy

(11-33)

can be evaluated

in closed

form by using the

[4]

a!jg!y!

N\N{N\dxdy
(a

++y+

2A,

(11 -34a)

2)!

where

denotes factorial and

|[ N.dxdy

IT

For example,

N\N\N\ dxdy
1!0!0!

(1+0 +

2A
2)!

=
3

2A
X 2~

A_
3

(ll-34b)

Chapter 11

257

Torsion

Hence

{<W

26A

(ll-34c)

which implies that the applied forcing function

is

distributed equally at the

three nodes.

Example

11-2.

Torsion of Square Bar:


Stress Function

We now
The

Approach

consider an example of torsion of a square bar 4

properties

G and 9

are as before.

Our

cm x

cm

(Fig. 11-7).

idea in choosing this and subsequent bars

of different cross sections for different approaches

is

to illustrate different discretiza-

by special characteristics such as symmetry of the cross sections. In


view of symmetry in the distribution of the stress function, we can consider and
discretize only one-quarter of the bar. In fact, as is done in the next example, only
an eighth of the bar can be discretized. The finite element mesh for the quarter of
the bar is shown in Fig. 11-7, and it is identical to the mesh used in Example 11-1.
Because the coordinates and the geometry of this mesh are the same as before, the
tions governed

(0,2)

(2,2)

\a/

a\a

/ A \

L_

(0, 0)

(2, 0)

_l

4< ;m

Figure 11-7 Torsion of square bar: stress function approach.

'"

258

Chapter 11

Torsion

element

stiffness

matrices are essentially similar.

The element load vector

is,

how-

by Eq. (ll-34c). Assembly of the four element matrices and

ever, different, given

load vectors lead to the global equations as


'2'

-41 9i
-4
r

-4

-4

-4

16_

<P*

.<Psj

.4,

(ll-35a)

or

[K]{r]

[R]

BOUNDARY CONDITIONS
According to Prandtl torsion, the boundary condition

(p

If

we choose

(p

we have

0,

is

or constant along the boundary.

(ll-36a)

(Fig. 11-7).

<Pi

<Pi

<^4

(ll-36b)

0.

Equation (ll-35a) is now modified by deleting the rows and columns corresponding to <p 2 (pi, and (Pi, to yield
.

L-4

\(p

16Jl^ 5

(ll-35b)

'

Solution by Gaussian elimination gives

?l

=| =

<p 5

<?l

2.66667 TV/cm,
1.33334,

and
?2
Step

6.

0.

Secondary Quantities

SHEAR STRESSES
The
[B]

stresses

can

now

be computed using the available values of

(p.

Use of the

matrix in Eq. (ll-12c) in Eq. (11-30) gives

dp
dx
dcp_

J_P>i
2A i_fli

b2

b{

a2

a2

(11-37)
.

[dy
Substitution of at and b t from Eqs. (11-20) leads to stresses in the four elements:

Element

1:
dip

dx
dip

dy.

r4
|J"-1

o-

- M<If

(H N/cm

2
.

259

Torsion

Chapter 11

we have

Therefore, from Eq. (ll-28a),

dcp

dy
>

N cm

2
.

d(p_

Txi

Element

dx

2.

N 'cm

Element 3:

N cm

i:i
Element

:
.

4.

eh>In view of the fact that the stress function [Eq.

and

l-4dj]

is linear.,

element are constant. Hence, as before, we can

stresses in the

the shear strains

element

call this

a constant-strain or -stress triangle (CST) element.

MOMENT

TWISTING

According to the

approach, the twisting

stress

||

tpdxdy

= f

(p

moment M,

m dxdy,

is

given by

[2]

(ll-38a)

'

where

m=

1, 2,

. . . ,

Af,

"a

M = number of

elements. Equation (ll-38a) represents

twice the volume under the stress function distribution over the cross section of the

bar (Fig. 11-7). For a generic element

(p

tu

we have

m dxdy

[Ntf -

N2f2 -

Nsfftbcdy

-yP?

Summing

(ll-38b)

(P\

over the four elements, we have

2x1/8

2 x
3

2 x

f-
0+3

4
3

0-0

0-0

2 x
3

2 (32

7.1111 N-cm.

(ll-38c)

260

Chapter 11

Torsion

Therefore, total

moment

for the bar

M, = 4 x
Step

l-8(a)

and

for the entire bar (4


linear

28.4444 N-cm.

(ll-38d)

show plots of computed stress

functions and shear stresses

7.1111

Interpretation and Plots

8.

Figure

is

approximation

and are attached

(b)

cm x
is

at the

4 cm) and the quarter bar, respectively. Note that because

used, the shear stresses are constant within each element

element centroids.

Figure 11-8 Results for torsion of square bar:

approach,

(a)

stress

function

Distribution of computed stress functions in entire

bar. (b) Shear stresses in quarter bar.

2.360

Exact distribution
of

\p

2.666

--?<:
Section

A-A
.2.666

3fc-.
1.333
Section B-B

(a)

y^yz
?

\
N

^4
t
3

jo

V
/\
A

shown

at centroids

/
L

Stresses

(b)

*- x. r.

COMPARISONS
The exact value of (p

Stress Function.

node

at

(x

0,

0)

from Eq.

(1

l-26a)

is

f
Therefore, the error

2.360

is

err0r

2.360-2.666
inn
X 10

1360

-13.00%.

Shear Stresses. Use of Eqs. (11 -26b) and


ical element number 1 {x
1/3) as
1, y

(11 -26c) gives the shear stresses in typ-

T = -0.246 AVcm 2
T

1.060.

Therefore, the errors are


error in x xz

error in x yz

= -0.246
= -25.00%

0.000

1.060-1.333

-26.00%.

=-gg

inn
x 100

Torsional Constant. According to Eq. (11-25), the torsional constant

^~
~
From

0.1406(2 x 2)*

36.0000

cm 4

computed

the foregoing, the error in the


error

is

= 36.0000-28.4444
=-7

Jo

torsional constant

is

inft
x 100

^21.00%.

We have used a coarse mesh so that hand calculations can be performed.


the errors are rather high.

a finer

mesh

is

The computed

Hence,

solutions can be improved significantly

if

used.

BOUNDS
Figure 11-9 shows the computed values of torsional constants from the warping
(displacement) and stress function approaches in comparison with the closed form

'Gd for the 2 cm x 2 cm bar of Ex.


by 16 as per Eq. (11-25). The displacement approach yields an
upper bound to the true value of torsional constant, whereas the stress function
approach gives a lower bound solution.
value.

For

this

comparison, the values of

11-1 are multiplied

Example

11-3.

Torsion of Square Bar with Finer Mesh:


Stress Function Approach

Figure ll-10(a) shows a refined mesh for an eighth of the square bar in Fig. 11-7.
In view of the symmetry,

it is

possible to reduce the problem to analysis of only an

261

262

Torsion

Note: The

number

Chapter 11

of nodes plotted refer to the specific problem only.

44.000
Warping function (Example 11-1)

42.000

40.000

38.000

36.000
8

10

Number

12

14

16

18

of nodes

34.000
Stress function

(Example 11-4)

32.000

Stress function

30.000

(Example

1-3)

Stress function

(Example 11-2)

28.000

26.000

Figure 11-9 Bounds in torsion problem.

eighth of the bar. Thus,

now we have

effectively eight elements for a quarter of the

bar instead of four as in Example 11-2. Figure ll-10(b) shows local node numbers
for the elements.

In the following

we

give brief details of computations of element equations for

the four elements (with values of

ax

Element

= 1,

0,

a2

a3

and 6 as before)

1,

2
1

b2

1,

b3

b2

\,

b3

=0;

0;

Xm

"J

sin

=i

ym

"J

0"

1.

Element 2:
ci\

1, a 2

0,

a2

1,

-1

-1

_-l

= i;

263

Torsion

Chapter 11

(1,0)

(0,0)

(a)

Y' T vz

(b)

Figure 11-10 Torsion of square bar: stress function approach, finer

mesh,

(a)

Mesh

for eighth bar of Fig.

1-7. (b)

Local node numbers.

Element 3:

ax

= 1,

a2

= 1,

= 0;

bx

= 0,

b2

1,

b3

= 1;

*m

=f

ym

=\\

|,

7m

0"

-1

-1

-1
1_

Element 4:

-1

-1

-1

-1, 62

= 1,

63

=0;

0'

-1
!_\{<p 6

tr

f;

264

Chapter 11

Torsion

Assembly of the element equations leads to the following global equations

1-1
-1

0"

4-2-1

0-2
0-1

0-2

2-10
0-2-1 4

(pi

03

<P*

-1

[*

-f

(ll-39a)

2
2

<Ps

1_

\<P6.

Introduction of the boundary conditions


g>4

<Ps

<Pe

0.0

leads to modified assemblage equations as

=h
+4p -2<p = $,
- 2<p 2 + 4^3 = ,

-<Pi -<Pi

-<Pi

(ll-39b)

solution of which gives


(p

<p 2

<p 3

(p*

(p 5

p6

The shear

stresses

= 2.333 TV/cm,
1.666,
= 1.500,
= 0.000'
= 0.000 prescribed.
= 0.000
>

according to Eqs. (11-37) and (ll-28a) are

follows

Element

1:
dq>

0.166

dy

N/cm

dy
dx

0.666

Element 2:
0.166

N/cm
f

1.666

Element 3:
0.000

)
I

N/cm

1.5000/

Element 4:
>

N/cm 2

now found

as

Chapter 11

Torsion

265

.1.666
2.333-

Along nodes 1-2-4

-1.500

2.333

Along nodes 1-3-6


(a)

y-^

Figure 11-11 Results for torsion of square bar: stress function


Distribution of computed stress func-

approach, refined mesh,

(a)

tions in quarter bar. (b)

Computed

shear stresses in eighth bar.

Figure 11-11 shows plots of computed values of

(p

and shear

stresses for the

refined mesh.

COMPARISONS
Torsional Constant.
(11-38)

torsional constant for the eighth bar according to Eq.

r/7

2 [(
L\ 3

=
Then

The

is

'

4J5
3 )

'

V3

3/"

'

3.8888 N-cm.

the total value of torsional constant

XM, = 8x

3.8888

is

31.1 100

N-cm.

"*"

3 J

266

Chapter 11

Torsion

Therefore, the error in the numerical solution

err0F

36.0000

is

31.1100

3^000

inn
X 10

~ 14.00%.
The computed value of the torsional constant from the finer mesh is shown
and yields an improved lower bound to the true value of MJG9.
Stress Function. From Eq. (ll-26a), the closed form value of (p at node

in

Fig. 11-9

(x

0,

0)

<p

hence, the error

found to be

is

is

e rr or

Shear Stresses. For element

2.360,

2.360-2.333

1.00%.

(x

3,

5/3,

x lnft
100

2/3),

use of Eqs. (ll-26b) and

(11 -26c) yields errors as

error int*,

error in x yz

= -0.194 = -19.00%

0.000

1.900-1.500

21.00%

For the foregoing comparisons,

j-^

inn
x 100

can be seen that the numerical solutions

it

with the finer mesh are closer to the closed form values. The error in the torsional
constant reduced from 21.00% to 14.00%, and that in the shear stresses from
25.00 to 19.00 and from 26.00 to 21.00% in r xz and T yz respectively. Note that the
,

comparisons of errors

in the shear stresses are

not rigorous, since they are not

same point; however, the general trend shows reduction

necessarily at the

in the

error.

The numerical

solution can be improved significantly with progressively refined

meshes. Such refinement should follow certain criteria for consistent comparisons;
these are briefly discussed in Chapter 13

Example

To examine

11-4.

Computer Solution

(Example

13-5)

for Torsion of

and

in Ref. 4.

Square Bar

the influence of a different (higher-) order element

on the numerical

we now consider a square bar divided into four (square) quadrilateral elements [Fig. 11 -12(a)]. Note that the size of the bar here is 2 cm x 2 cm, while in
solution,

the previous examples with the stress function approach the size of the bar

cm x

was

4 cm.

The computer code FIELD-2DFE


used to solve

this

rilateral element,

problem. This code

which

is

covered

tribution of the stress function as

in

and Appendix 4 was


based on a four-node isoparametric quadChapter 12. This element has a bilinear disdetailed in Chapter 12

is

compared

to the linear distribution within the

Chapter 11

267

Torsion

(1.D

(0, 1)

A
@o,o

G=

e =

A
d,o)

(5)

(a)

(0.00)

(0.00)

(0.00)

0.1

^
0.87

(0.48)

(0.00)

(0.38)

^J0.24
0.24

(0.62)

^J0.39
0.39

(0.48)

J0,7
0.1

(0.00)

[b]

[c]

Figure 11-12 Torsion of square bar with quadrilateral element,


(a) Finite

element mesh for quarter bar. (b) Computed nodal stress

functions in parentheses,

(c)

Computed

shear stresses.

is an improvement in the
assumed approximation within the element.
Figure ll-12(b) shows the computed values of stress functions at the node
points. Computed values of the shear stresses i xz and t yz are plotted in Fig. 11 -12(c).
The computed value of the torsional constant is 2.046.

triangular element used in this chapter. Thus, there

COMPARISONS
Torsional Constant.

constant

The

error in the

computed and exact values of the

is

error

2.250
2.046
x 100
2.250

^9.00'

torsional

268

Torsion

The value of

M /G9 = 2.046

x 16

32.74

Chapter 11

is

shown

(p

(Eq. 11 -26a) at node

in Fig. 11-9

and

yields a

lower bound to the exact solution.

0,

Shear Stresses. The shear stresses are compared for two typical elements,

The

Stress Function.

0)

is

closed

form value of

(x

found to be

<p

hence, the error

0.590,

is

error

0.590-0.620

^-^

inn
x 100

= -5.00%
and

2:

Element 1

(x

0.25,

0.25)

errorinr,^

error in t yz

Element 2

(x

0.75,

^- ^ !^

238

1.00%,

0.238-0.240

-1.00%.

fyrvi

24

1AA
x 10

0.25)

- (-0.100)
^-^

-0.104

error in i X2

'-

-4.00%,

0.853-0.870

-2.00%.

error in t yz

nfcTa

Use of the 4-node isoparametric


in the solution for the values of

<p,

inn
x 100

x inn
^^0

quadrilateral provides an overall

improvement

shear stresses, and torsional constant.

For better accuracy one should use finer meshes with a computer code.
With about 100 elements one can expect results of acceptable accuracy for
all practical purposes. The cost of such computations with a computer code
is

not high.

REVIEW AND COMMENTS


In the displacement approach,

we assume approximate displacement

or warp-

ing functions. These functions are chosen such that they satisfy physical

continuity of the

body or continuum up

mation. The potential energy


functions,

and

its

Up

is

to a certain degree or level of approxi-

then expressed in terms of the assumed

stationary value yields approximate equilibrium equations.

In other words, for the assumed compatible function, equilibrium of forces


is

satisfied

only in an approximate sense.

Chapter 11

Torsion

269

In the case of the stress approach, the assumed stress functions that

complementary energy

satisfy equilibrium are substituted into the

II C

Its

stationary value leads to approximate compatibility equations.

both cases, for the assumed unknowns, either the equilibrium or


is fulfilled only approximately. For example, in the case of the
displacement formulation for torsion, the stress boundary conditions over

Thus

in

compatibility

the surface are not necessarily satisfied. That

is, as required by Saint- Venant's


normal to the boundary computed from the
displacement procedure may not be zero. Figure 11 -13(b) shows the variation of t along the boundary AB of the discretized region of a triangle [Fig.
11-1 3(a)] subjected to torsion [8-10] computed from a displacement and a
hybrid formulation. It can be seen that t is not zero as required by the theo-

torsion, the shearing stress t

retical

We

assumptions.

can overcome some of these deficiencies by using hybrid and mixed

procedures.

HYBRID APPROACH

A wide variety of procedures is available for formulating the hybrid approach


One can assume displacement

and stresses on the


assume stresses inside
the element and displacement along the boundaries. Here we shall consider
[4, 8-11].

boundaries of the element. Conversely,


a special case of the former

Since

it is

inside the element

it is

possible to

[9].

found that the major error

in the results

from the displacement

we

shall choose the elements on the boundary, shown shaded in Fig. 11-14, for the hybrid formulation. For the elements in the interior the displacement formulation will be
used. Since the displacement approach has already been formulated, we give
now details of the hybrid approach for the elements on the boundary.
Consider a generic boundary triangular element (Fig. 11-14). The stress
function is assumed within the element as

formulation occur in the stresses at the boundaries,

<P

=N
=

q>

+N

2 <p 2

+N

2 q> 2

(ll-4d)

[NJ{q,},

where the subscript s denotes stress function. The general expression for the
warping function in the element is defined as
y/

= N.y/, + N
= tN]{q r

2 y/ 2

+Ny
3

(1Mb)

}.

If

we denote

the warping associated with the three sides of the element as

Vi3> Vsi* an d ^i2> respectively, specialization of

y/

in Eq. (ll-4b) along the

rad/mm

'a)

Hybrid

v^
-

4-4

-1

[,

A.
i

!/

\/\/]/\/\/V

-2 L
Displacement

(b)

Figure 11-13 Torsion of triangular bar with hybrid approach


(a) Finite

edge AB.

270

[9].

element mesh, (b) Distribution of normal shear stress on

111

Torsion

Chapter 11
Y

Boundary element

731

^Bou

W 7
CD

~^W
Inner element

=-

",2

and boundary elements

Figure 11-14 Discretization

in

hybrid

approach.

sides of the element leads to

y23
si

^12

Here we have used

^ +N

relation

y + (1 - N )y/ + N y/
= N ll +
y2 + (1 - tf )^
= (1 - ^ Vi + ^2^2 +
y/

definitions of

+N =
3

and the angle of

Nu N

and

(11-40)

as in Eq. (11-3)

and the

1.

As suggested by Yamada

et al. [9],

we

shall also

add the twisting moment


and follow

twist per unit length 9 to the formulation

their procedure.

Step

4.

Element Equations

In the case of the hybrid stress approach,

we

define a modified complemen-

tary energy expression, TI ch per unit thickness of the bar as follows

[9, 11]

u = ue + wph = u ch

h,

"-AJ['+(K)>*-*iA

= iHere

jj

{*}

\P]{"}dxdy

j {TYMdS

denotes the work of boundary forces, and

(11-41)

we have used

the subscript

h to denote the hybrid formulation. In very simple words, the procedure


in the case of stress

is

combines complementary strain energy Uc as


approach and potential of external loads
ph as in the

called hybrid, because in

ch it

case of the displacement approach.

In Eq. (11-41), {T} r

= [M

r 23 t 31 t 12 ]

and normal components of shear

stresses

is

the vector of twisting

on the boundary,

{\\f b

T
}

moment

[6 y/ 23

272

Chapter 11

Torsion

Wn

is

W\i\

the vector of corresponding angle of twist

with the sides of the element, {a}

defined in Eq.

is

and warping associated


and

(1 l-28a),

0"

[D]
1

which

the inverse of stress-strain matrix [C].

is

We

assume

linear elastic

behavior.

Uc and ph in IT cA we need to derive the following results


boundary element. For the triangular element, the stress vector {<r}
given by
To

evaluate

for the
is

Uxz

w=

dy

cp 2

(ll-42a)

lies

b,

dx

Because the element

a3

d(p '~2A

[*

boundary,

a2

01

on the boundary such that the side 2-3 is along the


which we assume as zero; then Eq. (ll-42a)

= constant,

<p 3

specializes to

f*l

f
"

(ll-42b)

<*>.

or
}

(ll-42c)

[PHK,

where
[P]

and

{P}

= {<p,} =

<p

is

2A\ b

just a scalar. Substitution of Eq. (ll-42c) into

[Eq. (11-41)] leads to

=m jj^ffow\dxd m
T

(ll-43a)

wmm
2

(ll-43b)

-<o

X 2A X 2A //
,

'a

_Ari (a] +
SA 2
G
where

l\ 3

side 2-3

= (a +
2

b\)

3U:H

b}) _lh(t>\

(ll-43d)

&GA

= (x
3

x2 ) 2

(11 -43c)

-f (y 2

2
3)

denotes the length of the

and
[H]

=
_

JJ

\PYtPmdxdy

*23

4G^

(11-44)

Chapter 11

Torsion

The second pan


tkms T over

in

273

Eq. (11-41) denotes the potential of the surface trac-

and can be

the entire surface of the element per unit thickness

expressed as
IV. ,

M.(

7r.Vr.dS-

'iiViidS
r

y,

'-^- dS

(ll-45a)

(ll-45b)

-;5.

Here r23 and so on are constants because of the linear interpolation function.
The components of {T} r = [M. t/ ;: u-. -.i\ are evaluated as follows:
.

{ymx

M.

(-y-a

f-4

xmx^dxdy

(11-46)

and by

referring to Fig. 11-14.


r ::

~.

~_

(a.b; b.a-.)p.
2Al ::

x: sin

r,,

cos a

'

'

'

_ (a-.b:

a b

imponant

'

It is

b a
:

(11-47)
l

b-a-

to note that the stress r23

)(p

0.

normal

to the

boundary

zero, as required by the torsion problem. In matrix notation,

\f

b;

identically

-xjt

y-.-i

is

we have

Ul

1 f3

(11-4&0

r::
*12-

hi

{f}

3 /::.

::

The

surface integrations relevant to

performed
Eq.

11-45

easily.

For

{\f/ b }

of Eq. (11 -45b) in

instance, substitution for

u :: from
.

W+

can be

Eq. (11-40) into

gives

because JS

di-4Sb)

[R;p

dN

l23

l-NJvt + N&JKhs

along side

2-3. Finally, integration

(ll-49a)

of the expression

274

in

Torsion

Chapter 11

Eq. (ll-49a) leads to

J2

The other two

y/ 23

= k^{y/i +

dS

(ll-49b)

Vi\

integrals are evaluated similarly.

Then Eq.

(11-45)

becomes

W = {tY\{y }dS
ph

(ll-50a)

(P}Wy

(ll-50b)

\hiiWi

2A g>1

^~ ai

b^

i)j

ymA

(b 2

-x m ^

(a 2

+
+

b 3 )/2

0,

a 3 )/2

(fl 3

+
+

&,)/2

(6 1

i)/2

(fl,

+6 )/2
+ a )/2
2

1^3
(ll-50c)

m [G]{%] = {P}W[L]{q,}.
T

(ll-50d)

Here

[G]

(ll-50e)

PT[L],

and

-2y m A

{L}

Here we used the

Now, we

relations a

substitute for

a2

and

b2

b3

a2

a3

(ll-50f)

2x m A

a,

a3

and

^ +

Z> 2

U = i{pf[H]{p} - {pf[G]{q,}.
differentiating U
with respect to {f} = p, and

By

Z>

= 0.
(1 1-51)

ch

to zero,

ph in Eq. (11-41) to obtain

ch

equating the results

we obtain

dUeh

[HP] - [GHqJ

0,

(ll-52a)

or
[H]{P)

(ll-52b)

[G]{q r ],

or

}=[H]-[G]{q r },
which expresses the relation between nodal
ing functions for the boundary elements.

(11-52C)
stress functions

and nodal warp-

Element

Matrix

Stiffness

We now

derive element stiffness matrix [kj relevant to the hybrid

approach. For

we

this,

express the strain energy as

P = tf*FhK*}where

{q w }

= [6

y/ x

y/ 2

the vector of generalized nodal displace-

is

y/ 3 ]

(H-53a)

ments. The complementary strain energy expression

e,

after substitution

of Eq. (ll-52c), leads to

m
=

u =
c

\n\m

Mq,}W([H]-y[H][H]-[G]{q,}
{qjqGF[H]^[G]{q
2"14^

Here, since [H] defined by Eq.

(1

1-44)

is

(ll-53b)

usually symmetric,

we have

([H]

_1 r
)

= [H]-'.
Equating the above two expressions for
stiffness

U and U we obtain the element


c,

matrix \k h ] as

[k*]=[GF[H]-i[G]
[Gf

AGA

(ll-54a)

[G]

Hz
A 2 (ym en + xm b\) 2

Al 223

A(y m ai
-A{y m a
(ym a\

23

y m ai

L-(y m ai

x m bi)y-^

+x m b

A{y m a\

-2 A

){^-)

x m b\) 2

+ xm bi
+ xm bi)

ym a\

a2

~2~ )

A(ym ai+x m bi)(2A)

j(-2A)(2A)

-j(2A)(-2A)

\(2A)(2A)

(ym ai +

x m bi

x m bi)~
(ll-54b)

-1

0-1
a2

2A\

xm b\)[

^(-2A)(-2A)

Here we used again the following


a

relations

= 0,

-f b 2

63

= 0,

and
a 3 b2

a b2

a2b 3

2A.

Inner Elements

We choose to use the displacement approach to define the stiffness matrix


for the elements inside the bar.

By following

the procedure outlined previ-

ously in this chapter, the stiffness matrix with warping functions can be
275

276

Torsion

Chapter 11

derived as

4A 2 (xl

w-&

yl)

2A{a\x m

b\

biy m )

2A(a 2 x m

a\

b\b 2

sym.

b\

+
+

2A(a 3 x m

b 2 ym)

a\ci2

b\bz

a\

b2b 3

b\

Both

[k A ]

and [kj express

relations

b 3 ym)

+ 01^3
+ a 2a
+ a\

(11-55)

between generalized forces and

dis

placements {q w } and have identical forms as

and

M{qJ =

{Ql

(11 -56a)

WW =

(Ql

(11 -56b)

where

Ay m

~^*m

&i

flj

{Q}

'M,'

^2
2

*3

*3

Ft

ft-

(ll-56c)

F2

F3
1

F ,F2 and F3 represent one-half of the resultant shearing forces on the


sides 2-3, 3-1, and 1-2, respectively.

and

Computation of Boundary Shear Stresses

For an element with


and (ll-52c),

side 2-3 along the

boundary, we have, from Eqs.

(11-42)

2A

_:;h

AL;;Hw
(0

a
l\i\_

{a

^- b x x m )
l ym

b {a y m
x

-b

xm)

0i

(11-57)
6,

3
Assembly

For elements on the boundary and inside the bar, we use element equaand (ll-56b), respectively. Since both are expressed in terms
of nodal displacements, the assembly follows the same rule that the displacements at common nodes are compatible.
In view of the fact that the hybrid approach satisfies the stress boundary
conditions, the results from the proposed procedure can yield improved

tions (ll-56a)

Chapter 11

277

Torsion

accuracy in the stresses at and around the boundary. In the following,

we

present an example using

puter solutions by

Example

We

Yamada

hand

calculations

and then

results

first

from com-

et al. [9].

Torsion of Square Bar: Hybrid Approach

11-5.

consider torsion of the square bar 2

cm x

cm shown

in Fig. 11-5. In this

example the local node numbering is changed as shown in Fig. 11-15 to comply with
the numbering for the boundary element in the foregoing hybrid formulation.

Global node
Local node

/j\ Element

(0, 0)

(2, 0)

Figure 11-15 Torsion of square bar: hybrid approach.

All the four elements

lie

on the boundary and hence are treated as boundary

elements in the sense of the hybrid approach.

boundary elements;
it is

necessary to use Eqs.

(1

incidental that all elements are

It is

both boundary and inner elements

in general,

l-54b) and

will occur.

Hence

for deriving the element

(1 1-55), respectively,

matrices for the boundary and inner elements. In the following are given salient
details of the

Element

element and assemblage equations

1:

Q\

a2
3

Here

A =

2,

b\

1,

b2

=y - ys = o,
=yi -y\ = -h

= X2 ~ X = 1,

b3

=y\ -yi =

= *3
=x
x

cm 2 and we
,

*i

x3

=
=

xm

1,

>23

2,

y m =i,

GA

have assumed that

G=

N/cm 2

i>

278

Chapter 11

Torsion

Use of Eq. (ll-54b) leads

to element equations:

2<

3<-

Mi = ~

Local

(0)

Mf,
f

}V1
2

-1

L-f
is

2-\

We assume that

Global

lj ly 3

uniform over the bar, and hence a special number

both local and global numbering


the twisting moment for element

is

assigned to this degree of freedom.

and so on.

common to

Element 2:
a,

=0,

br

a2

b2

1,

=2,

=
=

b3

-1,
-I.

=h
GA

/23

'23
I

1-

3-

id-

ri

Mtl

>

-1

u Wz\

<

[k] 2

Local

Element 3:
a\

*2
03

=
=
=
=

-2,

0,

6,

-1.

62

1,

63

= 1,
= 1,

.Vm

1,

GA
23

4<-

Global

3^-

W-

if

-*

-1

ion
3

Local

(6

M>

-1

1_ .Wsl

tl

denotes

Chapter 11

Element

Torsion

279

4:

xm

=
=
=
=

/'23
23

a\

ai
#3

-2,

6,

=0,

*3

1,
1,

"I,

y m =4,

1,

2,

'| 3

3*--

Global

3^-

^
[k;]4

10

Local

mu

01

1
"T

^
^

-1

-1

1_

Vs.

Combination of the foregoing four element equations by observing compatibility

of nodal

y/'s

leads to assemblage equations:

208

12

12
3

12
3

-1

-1

12
T

-1

W1

-1

4
0_

These equations can


for

the solution

now

m=l

tm )

1
1

-1

.s.

,
,

be modified for the boundary condition

y/^

0.

Then

1-1

with

is

= 1,
1

(prescribed),

2 =2,
s

*
s
These

=
=
=

-2,
0,
0.

results for y/'s are essentially the

the warping function approach.

We

same

as in the case of

have considered

Example

this rather

simple example

hand calculations. For acceptable accuracy and in order to


improved boundary stresses in the hybrid procedure, one
should use a finer mesh. In Example 1 1-6 we shall consider a computer solution for
the mesh in Fig. 11-13 where both boundary and inside elements occur. The results
in Example 11-6 will illustrate the advantage of the hybrid approach.

mainly to

illustrate

exploit the advantage of

SHEAR STRESSES
For the four elements

boundary, we use Eq. (11-57) to evaluate the shear

at the

stresses:

Element

1:

2(2 x

O-i

2"

0.

(0.666)

-2

J)

N/cm 2

lo.oooJ

Element 2:
1

0"

'
1

4 _2 x

WvJ

-2J -2

0.000)

N/cm 2

10.666/

Element

3.
1

"

Zxz

0"

" 4

lyz)

_-^
000]

-2

N/cm 2

2J

lo. 666)

Element

4.
1

_20

-2

ri--[

0_

-2
f

0.6661

N/cm
I

O.OOOJ

These results are the same as those in Example 11-1 (Fig. 11-6); they show that
boundary is zero. It was only incidental that with the
four-element mesh chosen the warping function approach also satisfied the zero
stress boundary condition. With a different and arbitrary mesh, the warping function approach, in general, would not satisfy this condition.
the shear stress normal to the

TWISTING MOMENT
The torsional constant can be found by
1//

into the

280

first

substituting the

of the above assemblage equations

computed

results for

281

Torsion

Chapter 11

M^ [fxl
=
This value

is

same

the

_ 4x2 + 4x( _ 2)]

1.7777 N-cm.

as that

from the

approach and has an error

stress function

of about 21.00%.

Example

Torsion of Triangular Bar:

11-6.

Hybrid Approach, Computer Solution


Figure 11-1 3(a) shows a bar in the shape of an equilateral triangle. Analysis for torsion for this bar using the hybrid

method

symmetry, only one-sixth of the triangle

is

is

presented by

discretized.

Yamada

The

et al. [9].

Due

to

properties of the system

are

Poisson's ratio,

E=
=

Shear modulus,

G=

Young's modulus,

2.1

x 10 4 kg/mm 2

0.3,

=0.81 x

10 4

kg/mm 2

Table 11-1 compares the values of warping along the side

AB

[Fig. ll-13(a)]

Angle of

twist,

>

2(]

1.194

'

..
fl

x 10~ 4 rad/mm.

obtained from the closed form solution, the displacement method, and the hybrid

TABLE

Warping Along the Contour

11-1

AB

[9]

Displace-

Node

Exact

Hybrid

ment

Solution

Method

Method

11

0.370

0.368

0.375

13

0.957

0.942

0.975

21

1.924

1.884

1.921

29

2.772

2.721

2.766

36

3.471

3.419

3.467

43

3.887

3.821

3.868

49

3.973

3.913

3.960

55

3.962

3.632

3.659

60

2.978

2.918

2.960

63

2.398

2.352

2.385

66

1.865

1.834

1.854

69

1.354

1.330

1.347

73

0.898

0.885

0.896

76
79

0.530

0.524

0.529

0.270

0.268

0.270

*h>

10 3

mm.

282

Chapter 11

Torsion

method. The two numerical results are close to the exact solution, without significant difference between them.
As noted before, Fig. ll-13(b) shows comparisons between computed values of
the shear stress normal to the boundary r n from the displacement and the hybrid
methods. In contrast to the solution by the displacement method, the hybrid
approach yields zero values of T, as required by the theory.
Figures 11 -16(a) and (b) show comparisons between shear stresses r xz and t yz
along

AB

as

computed from exact solution and the displacement and hybrid

methods. The

latter

boundary but

in the interior also; this aspect

improves computations of the shear stresses not only along the


is indicated in Table 11-2, which lists
the shearing stresses at the centroids of typical elements adjacent to side AB (Fig.
11-13). These results also show the improvement in accuracy provided by the hybrid

approach as compared to the

TABLE

11-2

results

from the displacement approach.

Comparison of the Shearing Stresses


AT THE CENTROID OF TRIANGULAR ELEMENTS ADJACENT
to AB [9]*

Element

48

81

99

122

-7.136

-6.354

-4.351

-2.259

-0.128

-7.168

-6.264

-4.117

-2.251

-0.125

-7.145

-6.884

-4.759

-2.563

-0.225

12.354

10.803

7.209

3.348

0.168

12.416

10.853

7.131

3.900

0.216

12.541

10.511

6.653

3.628

0.071

-0.003

-0.101

-0.164

-0.282

-0.027

0.000

0.000

0.000

0.000

0.000

0.083

-0.706

-0.795

-0.406

-0.159

Ixz

Exact
solution

Hybrid

method
Displacement

method

tyz

Exact
solution

Hybrid

method
Displacement

method

Exact
solution

Hybrid

method
Displacement

method

*Unit:

kg/mm 2

Exact solution

-2

Hybrid method
Displacement method

-4 -

t x2

(kg/mm 2

(a)

r yz

(kg/mm 2

14

+.*
12

10

Exact solution

Hybrid method
Displacement method

(b)

Figure 11-16 Comparisons for shear stresses


x xz along contour

AB.

(b)

Shear

stress xyz

[9].

(a)

Shear

stress

along contour AB.

283

MIXED APPROACH
In the mixed formulation, both the displacements (warping) and stresses
(stress function) within (including the

to be

unknown. Hence,

boundaries) the element are assumed

for a triangular element,

[N ,]{q r }

[N

N N N
2

MJfo.

N N N
t

[NJ
{q<xF

where

{<y}

x y2 ]

[t xz

is

The

* XZi

XZl

* XZ3

(ll-58b)

(q c
3

the vector of shear stresses at any point, [N^] and

are the interpolation functions, {q^}

= [i

(ll-58a)

3]

lyzt

Tyz*

*yz 3 ]

strain displacement relation

tutive relation for linearly elastic

is

is

is

the vector of nodal warping,

given in Eq. (11-8), and the consti-

and homogeneous material


1

and

the vector of nodal shear stresses.

is

0"

(11-59)
1

Step

We

Derive Element Equations

4.

[12, 13].

expressed as

n* =

Noor and
mixed approach can be

procedure presented by

dU

the

variational function for the

w w - *ww - JJJ^. -

is

[f xz

If

vm^ds,

(i

l-eo)

Si

the complementary energy density [Eq. (11-29)], given by

dU

{a}

essentially

[4, 12]

JIf
V

where

follow

shall

Anderson

fy2 ]

is

X2

TJ^J ^

*XM

dV

the shear stresses at the boundaries; {n}

is

the vector of

outward normal to the boundary; V is the volume of the element; [x} T


= [x y] and S 2 is the part of boundary on which tractions are prescribed.
Substitution for [a] in Eq. (ll-58a), {e} in Eq. (11-8), and dUc in n* leads
;

to

284

285

Torsion

Chapter 11

n*

{q.f[N.F[B]{q^K

|j{
V

{q.F[N,F[D][NJ{q}rfK

JJJ
V

-6

{q.F[NF[N,]{x}</K

JJJ
V

[[ {q,}
5

Here

{x}

x2 x 3 y y2 y 3 ]

[x x

[N r F{n}[NJ{qJrf5.

the vector of nodal coordinates, given by

is

x\VN N N
2

~~

yJ
and

[B]

is

Nt

(11-62)

N Nj
2

defined in Eq. (ll-12c).

Application of the variational principle to TIr leads to

The

(11-61)

its

stationary value.

variations or differentiations are performed independently

and simul-

taneously with respect to nodal stresses and nodal warping functions; thus

dUM
<5ru

Wq.

(11-63)

-<?{q

This leads to two sets of equations,

fc]

ftJ'

LfcrF

{Q.:

{q.

J {q

[0]

(ll-64a)

{0}

where, for the triangular element (Fig. 11-2),

fcj
(6

JJJ [NJ
(6

6)

PU = jjj [NJ r
(6x3)

(6

[B]

2) (2

^^z,

[NJ

[D]

2) (2

2) (2

(ll-65a)

6)

(ll-65b)

dtofy<fe,

3)

and

{Q.}=0jJ{N.f
(6

1)

(6

2) (2

^{4

[N]

6)

(6

(ll-65c)

1)

The last term in Eq. (11-61) represents boundary conditions. Here we


have assumed that the boundary conditions occur as prescribed shear

286

Chapter 11

Torsion

stresses

and that their components normal to the boundary vanish. The


components do not contribute when the variation, Eq. (11-63), is

tangential

performed, hence the

last

term in Eq. (11-61) does not appear in the calcula-

tion of forcing parameter vector {Q,,}.

Evaluation of Element Matrices and Load Vector

We now

computations of the element properties for the

illustrate

angular element. For unit length,


AT,

--*jj

N,

0"
3

Lo

N,

N
Ns

^2

fN,

AT,

//l
U.

N_
3

N,N
N,N

Mt/ft

"

Nl

Nl

hB

dxdy.

sym.

N,N2

If1*3

Nl

NN
2

Nl
The

integrations can be performed in closed

N\dA

2A

\\

form for
;

2!0!0!
(2

2)!

and

jy

NN
x

dA

2A

1!1!0!

(4)!

12

Therefore,

0"

"2

[kj

UG

2_

instance,

tri-

Chapter 11

Torsion

287

Now
~JV

0"

tf z

w-]J

b3

b2

bi

dA

2A

_0

^3_
r

N bi

Nb Nb
N b, N b
2

_N

a
3

and

N,a 2

N,a 3

N
N

N
N

-Air N a

N b,Nb
Nb

N,b 2

a2

3 a2

a3

a 3_

dA

"*,

b2

b3

*,

b2

b3

*,

b2

b3

*1

a2

a3

a2

a3

a2

a3_

0"

~N,

'*>'

Xl

w-fj

N
W
_0

AXi

6A

pv,

A^3

-\-

[_o

0"

AT,

x3

>'2

JV 3 _

^3>

X 2 ~ -x 3

x,

*1

+ x2 + 2x

2x 2 - - x 3

T2'

>

-y*
yi

2y 2

- ->' 3
2>'3,

xample

11-7.

To rsion of Squ are

The node numbers and details for


tion) in Fig. 11-4 are shown in Fig.

the

I lar

:Mixed Approach

problem of the square bar

(2x2

cross sec-

11-17. In the following are given essential details

288

Chapter 11

Torsion

of the computations of element and assemblage matrices based on Eqs. (11-64) and
(11-65).

Equation (ll-64a) can be written

in the

expanded form as

-2

-1

2b

2b 2

2b{

-1

-2

2b i

2b 2

2b 3

?XZ2

-1

-1

2b 3

Txz 3

A
\2G

2b

2b 2

-2

-1

-1

2a

2a 2

2a 3

-1

-2

-1

2a i

2a 2

2a 3

*yz2

-1

-1

-2

2a\

2a 2

2a 3

yz%

2b\

2bi

2b\

2a i

2a \

2a\

2b 2

2b 2

2bi

2a 2

2a 2

2a 2

_2b 3

2b 3

2b 3

2a 3

2a 3

2a 3

+ x2 + x
+ 2x 2 + X
x\ + x 2 + 2x
2yi + y 2 + y
\= QAhi + 2y 2 + y
121
y\ + yi + 2y
(2x\
X\

3\
3

W\

\V3l
(ll-64b)

Element 1:
i

a2

a3

x
13

-2

-1

-1

-1

-2

-1
1

14

=2,

= -1
= -1
=

-2
2

-1

-2
-1

-2
-2

-2
-2

= -1,
=h
=h
Global

-1

-1

b2

ym

-1

=0,

b3

-2

bi

-2
-2
-2
-2
-2
-2

2"

It XZ\ \
*>

-2
-2
-2

XZi

We

13
1

Txz,

tyz\

14

y \ =(
yz

Tyzt

Vi

0_

/4\

2
5

15

Wsl

have assigned global numbers by rearranging the unknowns as shown in


Fig. 11-17. That is, in the subsequent assemblage equations, the global numbers are
assigned such that the three unknowns at a given node appear consecutively. This is

done simply for convenience.


Element 2:
ax

=0,

bi

=2,

"2

=
=

1,

b2

-1,

b3

=
=
=

a3

Xm =i,

ym

-1,

-1,
1,

(7, 8, 9)
(

289

Torsion

Chapter 11

(10, 11, 12)

T xz3< r vz3< ^3

r xz4' r yz4' ^4>

(2,0)

(4,5,6)
(t xz ,,t

<

r xz2' r vz2'

iM

Global nodes

/l\ Element
Local nodes

(1, 2, 3)

(r xz1

T yz1

\p y

Global degrees of freedom

Figure 11-17 Torsion of square bar: mixed approach.

13

-2

-1

-1
-1

14

-2
-2
-2

-2
-2

-1

-1

-2

-2

Global
1

-2" /t

-2
-2
-2

f2\

-2

13

*l

***,

*jr*i

14

{*
-1

-2
-2

3<-

Tr*i

^
r

^1

15

^2

3 y

U)

Chapter 11

Torsion

290

Element

3:

=0,

= h
= 1,
=

#2

a3
v
13

10

-2

-1

-1

-1

-1

-2

14

11

Z>3

ym

-2

-2

Global

12*

2~

T\

6\

^*Z2

*,

10

Ty*i

Tr*

Ty*s

-2
-2
-2

-1

= 1,
=h

-4
-4
-4
-1

-4

-2,

15

-2

-4

2
i

0_

13

14
>

5
11

15

W3/

\0l

12

Element 4:

=
=
=
=

ax

a2
a3

x
13

10

14

11

-2,

0,

l,

63= -1

1,
1

.Vm

15

=4,
Global

12

-2

-1

-1

2"

-1

-2

-1

-1

-1

-2
-2

-1

-1

-1

-2

-1

-1

-1

-4

-4

-2
-4

-2

-2

-2

-4
-4
-4

r X2 \

'4\

13

~XZ2

10

r X23

ty*\

14

lyzt

11

Tyzi

>

<

12

l
0_ \3

15

<o,

The foregoing four element equations can now be assembled by observing

inter-

element compatibility.
It is difficult

hand calculations the (modified) assemblage equations


and it becomes necessary to use an (available) equation
of equations (see Appendix 2). In this elementary treatment,

to solve by

(involving 14 unknowns),
solver for large sets

it

291

Torsion

Chapter 11

will suffice

only to illustrate the mixed approach and the foregoing steps, because

considered beyond our scope. Solutions for this

detailed consideration of the topic

is

and the other relevant

Problems are

exercises in

left

to the inquisitive

and advanced

reader.

The foregoing completes our treatment of the torsion problem. Before we


proceed to the next chapter in which a 4-node quadrilateral element is described, it
is appropriate to present a brief description of the topic of (static) condensation
which

is

often used in finite element applications.

STATIC CONDENSATION
Often

it

may be

particularly
single

convenient and necessary to use nodes within an element,

when

element

is

it

is

required to use higher-order elements. Sometimes a

divided into subcomponent elements by using the inner

nodes. For instance, Fig. 11-18 shows a line and a quadrilateral element

each with one inner node. The

element has two end or external or

line

primary nodes and one inner node, and the quadrilateral has four corners
or primary nodes and one inside node.

Figure 11-18 Static condensation,

(a)

Line element, (b) Quadri-

lateral.

Inner node

'a'

292

Chapter 11

Torsion

The

finite

element equations

[k]{q}={Q]
are derived

on the

(ll-66a)

basis of unknown degrees of

freedom

at all the nodes.

Thus

the formulation includes improved (higher-order) approximation yielded

by the use of the inner node. In the case of the quadrilateral [Fig. 11-1 8(b)],
the element equations are often obtained by adding individual element equations of the four

unknowns

component triangles. Thus the total number of element


and 10 for the line and the quadrilateral, respectively.

in {q} are 3

These element equations can now be assembled to obtain global equations,


in which the degrees of freedom at the inner node(s) will appear as additional
unknowns, and to that extent the number of equations to be solved will be
increased.
It is

unknowns

possible to temporarily eliminate the

by creating an equivalent element relation


procedure of doing

this

in place

called static condensation. It

is

at the inner

node

of Eq. (ll-66a). The


is

possible to

do

this

because the unknowns at the inner node do not participate in the interele-

ment compatibility at the element sides; that is, the unknowns at the inner
node are not needed for the direct stiffness assembly procedure.
The procedure of static condensation involves solution of the unknowns
at the inner node in terms of those at the primary nodes. To understand it,
we write Eq. (ll-66a) in a partitioned form as
[q,}|

IK]

=
""

fCQJl

(ll-66b)

ItQJJ

{q,]J

{Q p } and {q,}, {Q,} denote vectors of nodal unknowns and loads


primary and inner nodes, respectively. The second equation in Eq.
(ll-66b) can be used to solve for {q,} as
where

{q p },

at the

{q,}

= [k,,]-'({Q,) -

Substitution of {q,} into the

dk,J

first

[k,,F{q,}).

equation of Eq.

frJfcJ-fcjFXq,}

= {Q,} -

(1

(1

l-67a)

(1

l-67b)

l-66b) leads to

fcJfcJ-'tQ,!

or

where

[k]

and {Q} are the condensed element

(stiffness)

and nodal load

vectors,

unknown {qj now contains the degrees


of freedom only at the external primary nodes thus when [k] and {Q} are
used to assemble element equations, we have reduced from the total equations
to be solved equations corresponding to the unknowns at the inner node.
respectively. Notice that the vector of

After the assembled equations are solved,

if

desired,

it is

possible to retrieve

Chapter 11

the

unknowns

static

293

Torsion

at the inner

node by using Eq.

condensation can be found in

many

(ll-67a). Further details of

publications, including Ref.

[4].

PROBLEMS
Invert the matrix [A] in Eq. (11-2) given by

11-1.

and obtain the matrix

in

x2

x3

yi

yy

EHB

Eq. (ll-3b).

mesh in Fig. 11-10 and treat the problem as torsion of a triangular


By employing the warping function approach and setting boundary
condition as y/ 2 = 0, compute nodal warping functions, shear stresses, and
Use

11-2.

the

bar.

twisting

With the mesh

11-3.
11-4.

moments.

in

With the mesh

Prob. 11-2, solve for torsion using the hybrid approach.

in

Prob. 11-2, derive element equations using the mixed

approach.
11-5.

Solve for torsion of the square bar in Fig.

1-19 by using the warping func-

tion approach.

\A
A\
a\

A
\A\A \
A\

(0, *)

'

\A

\*-,

*i

-*

(0,0)

G=

6 =

(2,0)

Figure 11-19
11-6.

Solve Prob. 11-5 by using the stress function approach.

11-7.

Solve Prob.
8 as

1 1-5 by using the hybrid approach. Consider elements


boundary elements and the remaining as inner elements.

4, 6,

and

This and some of the other problems will require use of an available equation solver once

the global equations are obtained.

294

11-8.
11-9.

Torsion

Chapter 11

Solve Prob. 11-5 by using the mixed approach.

Consider the mesh in Fig. 11-20 and solve for torsion by using the stress
it may be necessary to use a computer. The code
FIELD-2DFE can be used for solving the entire problem. Alternatively, the

function approach. Here,

assemblage matrix can be computed by hand calculations, and its solution


can be obtained by using an available code for solution of simultaneous
equations.

(2,0)

(0,0)

Figure 11-20

11-10.

For the mesh

in Fig.

1-21, use the

mixed procedure and derive the element

equations [Eq. (ll-64a)].

Figure 11-21

Chapter 11

11-11.

For the torsion problem in Fig. 11-7, treat the bar as having a cross section
cm x 2 cm. With boundary conditions (p = (p 2 = <p 4 = (p 5 = 0, compute <p 5 using the stress function approach. Evaluate torsional constant and
of 2

discuss

its

Answer:
11-12.

295

Torsion

accuracy.

<p 5

0.6667;

M /G9 =
t

1.777, error

21.00%.

For the circular bar of 2-cm diameter, compute nodal stress functions for
mesh shown in Fig. 11-22. Find shear stresses and the twisting moment
= 0.005 deg/cm.
by using a computer code. Assume G = 10 7 N/cm 2 and
the

Figure 11-22

11-13.

For the

bar a

elliptic

= 2 cm

stress functions, shear stresses,

code.

Assume G

and 6

and b = 1 cm (Fig. 11-23), compute nodal


and twisting moment by using a computer
1.

cm

Figure 11-23

11-14.

By

using the warping function approach, find shear stresses and torsional

constant for the triangular bar divided into two elements (Fig. 11-24).

Assume G

and

296

Torsion

Chapter 11

^
(0,0)

Figure 11-24

11-15.

Use

the stress function approach to solve for torsion of the triangular bar

inProb. 11-14.
11-16.

By

using the code

FIELD-2DFE

mesh
and obtain

or other available code, refine the

progressively (Example 11-4), say 1,4, 16,64, etc., elements,

<p and shear stresses. Plot the twisting moment


number of elements or nodes and study the convergence behavior of

results for stress functions

vs.

the numerical solution.


11-17.

Higher-order approximation. Use the quadratic interpolation function for


q>

with a triangular element (6 nodes) and derive the element matrix and

load vector. See Fig. 11-25.


Partial results:

Assume

W
Note

tyl

that this function includes

=
<P2

all

[N]{qJ,

9*

<P*

<P5

terms up to n

9s\

2 in the polynomial

expansion discussed under "Requirements for the Approximation Function."

Figure 11-25

Local coordinates, L

Node

(1,0,0)
(0,

1,0)

(0, 0,

(2

-. 0)
v;,
2

(O.I.I)
(1.0.1)

Chapter 11

297

Torsion

-1)1

[1.1(21,,

I 2 (2I 2
L 3 (2L
[N]

1)
1)

4I 2 I 3
4I 3 Ii

[BJ

(2x

[0]-,

[qj
.

[Blq,

&>]-

[0]

1)

where

t] = [(4Z.,

-1)6,

(4I 2 -1)^2

(4L3-1)63

-l)a 2

(4I 3 -l)fl3

4{IiZ)2-I:6i)

4(L z b 3

-L

b2)

4(Z. 3 6,

+,63)]

and
fc]

= [(4L,-l)a,

(4I :

4(I.,<i2

+2*1)

4d* 3 +X*2) 4(L3fli +JL,a3)J.

Computation of {Q,} will require integrations over element area and


vant boundary surfaces.

rele-

REFERENCES
[1]

Love, A. E. H., The Mathematical Theory of

Elasticity,

Dover,

New

York,

1944.
[2]

[3]

Timoshenko,
York, 1951.

Herrmann.

S.,

and Goodier,

N., Theory of Elasticity. McGraw-Hill,

New

L. R., "Elastic Torsional Analysis of Irregular Shapes," /. Eng.

Mech. Div. ASCE, Vol.


[4]

J.

91,

No. EM6, Dec.

Desai. C. S.. and Abel, J. F., Introduction


Nostrand Reinhold, New York. 1972.

[5]

Bach, C, and Baumann.

[6]

Murphy.

R., Elastizitat

1965, pp. 11-19.

to the Finite

Element Method. Van

and Festigheit. Springer.

Advanced Mechanics of Materials. McGraw-Hill.

G.,

Berlin. 1924.

New

York,

1946.
[7]

Yalliappan.

S..

Anisotropic Bars."

and Pltmano.
/.

Struct. Div.

V. A.,

ASCE.

'Torsion
Vol. 100,

of

Nonhomogeneous

No. ST1,

Jan. 1974, pp.

286-295.
[8]

Yamada,

Kawai,

T.. Yoshimura, N., and Sakurai, T., "Analysis of the


Problems by Matrix Displacement Method," Proc. 2nd Conf.
Matrix Methods in Structural Mechanics, Wright Patterson Air Force Base,
Dayton, Ohio, 1968, pp. 1271-1289.

Y..

Elastic-Plastic

[9]

Yamada.

Y.,

Nakagiri.

S..

and Takatsuka. K.. "Analysis of Saint-Yenant

Torsion Problem by a Hybrid Stress Model/' in Proc. Japan-U.S. Seminar on


Matrix Methods of Structural Analysis and Design, Tokyo, Aug. 1969.

298

[10]

Chapter 11

Torsion

Yamada,

Nakagiri,

Y.,

S.,

and Takatsuka, K.,

"Elastic-Plastic Analysis of

Saint-Venant Problem by a Hybrid Stress Model,"


Engineering, Vol. 5, 1972, pp. 193-207.
[11]

Int. J.

Num. Methods

Pian, T. H. H., and Tong, P., "Finite Element Methods in

Mechanics,"

in

Advances

in

in

Continuum

Applied Mechanics, Vol. 12, Academic Press,

New

York, 1972.
[12]

Noor, A. K., and Andersen,

CM.,

"Mixed Isoparametric Elements

Saint-Venant Torsion," Comput. Methods Appl. Mech. Eng., Vol.


pp. 195-218.
[13]

Noor, A. K.,

(private communication).

6,

for

1975,

OTHER FIELD PROBLEMS


POTENTIAL, THERMAL,
AND FLUID FLOW

INTRODUCTION
The problem of torsion (Chapter 11) and a number of other problems that
we shall consider in this chapter are often known as field problems. They are
governed essentially by similar differential equations, which are special cases
of the following general equation [1]:

U ) + U
k

The

U %) + =%-

(12 - la)

associated boundary conditions are


q>

kx

>%)

l + kjjgt, +
x

= p(t)

k^Ll,

<*(<p

on S

cpo)

(12-2a)

iff)

on S 2 and S 3

0,

(12-2b)

Here

cp

is

the

unknown

(warping, stress function, velocity potential, stream

head or potential); k x ,ky


and k z are material properties in the x, y, and z directions, respectively; Q
is the applied (heat, fluid, and so on) flux; c is specific heat or effective
porosity, and so on S is the part of the boundary on which (p is prescribed
and S 2 is the part of the boundary on which the intensity of flux q is prescribed.
In the case of heat flow a(^ <p ) is prescribed on 5 3 in which a is the
transfer coefficient and (p Q is the surrounding temperature. C xy C y and l 2 are

function, temperature, electrical potential, fluid

299

Potential, Thermal,

300

and Fluid Flow

Chapter 12

the direction cosines of the outward normal to the boundary, t denotes time,
and the overbar denotes a prescribed quantity.
In this book, we shall consider only two-dimensional steady-state problems; that is, the problem is independent of time and the right-hand side of
Eq. (12-la) vanishes. Also, for simplicity, only homogeneous materials are

considered; then Eq. (12-la) reduces to

kx %g

+ k*j* + Q =

and the boundary conditions

to

=f

q>

(12-lb)

on 5,

(12-2c)

and
kx

We note that Eqs.

/x

lx

+ k,fy, + q =

(1 1-5)

and

(1 1-27)

on S 2

(12-2d)

have the same form as Eq. (12-lb).

POTENTIAL FLOW
The

potential flow of fluids

is

governed by a special form of Eq. (12-lb),

dx 1

dy 2

or

V> =

(12-3)

0.

which is called the Laplace equation. Here V 2 is a differential operator. The


assumptions commonly made are that the flow is irrotational that is, fluid
particles do not experience net rotation during flow, the friction between the
fluid and surfaces is ignored, and the fluid is incompressible [2]. Some practical problems where this kind of flow can be assumed are flow over weirs and
through pipes (with obstructions).
Equation (12-3) is based on the basic requirement that the flow is con;

tinuous; that

is,

where v x and v y are components of velocity in the x and y direction, respectively. The flow problem can be represented in terms of either the velocity
potential

q>

or the stream function

that for the stress-deformation

the

same manner

respectively.

as

the

y/.

This dual representation

problem

in the sense that

(p

and

is

y/

similar to

are used in

displacement (warping) and stress functions,

Chapter 12

The

Potential, Thermal,

relations

and Fluid Flow

301

between the velocity components and

q>

and

y/

are given by

[2]

dx
(12-5a)

and

57
(12-5b)
tfy

"' =

Substitution of i\ and

p, into

Eq. (12-4) leads to the Laplace equation as in

Eq. (12-3).

Boundary Conditions

For flow through a cylindrical pipe (Fig. 12-1) a boundary condition is


and the wall have the same normal velocity. Hence, if the wall
stationary, we have

that the fluid


is

V.w =V..n =0,


where
n

is

and

the unit

Vw

(12-6)

are the velocities of the fluid

normal

and the

wall, respectively,

and

vector. Substitution of Eq. (12-5a) into Eq. (12-6) yields


d(P
n

-0

d(P
n

Sv

(12-7a)

S,-^'

Figure 12-1

From

Flow

in pipe.

Fig. 12-1.

dx
,

dn
dv

_
~

dy

(12-8a)

ds

dx

(12-8b)

Potential, Thermal,

302

Hence Eq.

dx
dx dn

is

Chapter 12

(12-7a) transforms to
dtp

which

and Fluid Flow

,dydy

similar to Eq. (12-2b). This

condition (Chapters 2 and

The

3).

the flow or

is

(12-7b)

0,

dn

dy dn

Neumann-type boundary
boundary condition

potential or Dirichlet

is
<p

onS

=<p

(12-8c)

Often both the flow and potential boundary conditions occur together, which
is

called the

FINITE

mixed condition.

ELEMENT FORMULATION

For two-dimensional

idealization,

we can

use a triangular or quadrilateral

element. Formulation with a triangular element will be essentially identical

one described for torsion in Chapter 11. Hence, we present a formulaby using a four-node quadrilateral element (Fig. 12-2). It is possible to

to the

tion

use either

(p

or

y/

(or both) for the finite element formulation.

sider a formulation with

We

first

con-

(p.

(x 4

y4

M.+1)

/T

@7( +1

(-1,-1)

<

+1

^^-^_~/^
(x 2

y2

>

(+1,-1)
~

Figure 12-2 Quadrilateral isoparametric element.

The velocity potential is a scalar and has one value at any point. For the
four-node quadrilateral, there are thus four nodal degrees of freedom, and a
bilinear

model

for

q>

at

any point can be written as

(p

= g, +

= mm*

CC 2

<x>

&*xy

(12-9a)

or
(12-9b)

where

[<J>]

x y xy] and {a} r = [<%! a 2 a 3 a 4 ] is the vector of generalized


Note that the term xy yields a bilinear distribution compared

coordinates.

Chapter 11;

by the triangular element discussed

in a given direction the distribution is linear,

pt
(Pi

^3
^4

in

however. Evalua-

nodes yields

at the four

<p

303

[1

to the strictly linear distribution given

tion of

and Fluid Flow

Potential, Thermal,

Chapter 12

= a, + a

+ y
+ 83^2
+a
+ a 3^4
<x 3

= + <*2*2
=a +ax
= ai + a *4
<*1

4 j> 3

+ aA x y u
+ a ^2>
+ax
+ xy
x

o^ 4

3 .y 3

(12-10a)

4i

or
{q,}

where

[A]

(12-10b)

[A]{a],

the square matrix of coordinates of the nodes

is

^4]-

[<P\ (pi <Pz

We

r
{q,,}

can solve for {a} as

{*}=[Ar{q,}
and

and

(12-lOc)

substitute the results into Eq. (12-9b) to yield

The product

mwr i%i = IMW = N&-

[<D][A]

(12-11)

gives the matrix of interpolation functions [N],

and

in terms of the local coordinates s

N
N
N

(Fig. 12-2) are given

=i(l-s)(l-t),

=#l+s)(l-t),

which

by

(12-12)

=i(l+s)(l+t%
N<=(l-s)(l+t).
3

Here,

s, t

forty,

The

are nondimensionalized local coordinates. Figure 12-3 shows plots

1,2,3,4.

global coordinates x,

at

any point

in the

element can also be

expressed by using the same interpolation functions

i9

S Nx
t

(12-13a)

= S N,y
i=l

or

x
\- ~N, N2 N N
y\N, N
r
T
where {x} = [x, x x x and {y} = [y, y
3

0~
{{*

#3

4]

<p,

are both expressed

(12-13b)

y 3 y4].

This approach in which the geometry, that

unknown

N _ \{y

is, x, y coordinates and the


by using the same interpolation functions

Potential, Thermal,

304

and Fluid Flow

Figure 12-3 Distributions of interpolation functions

is

called the isoparametric concept.

cept offers a

Chapter 12

number of advantages

As

in

1, 2, 3, 4.

stated previously, use of this con-

terms of easier differentiations and

integrations.

The approximation function

in

Eqs. (12-9) and (12-11)

satisfies

the

requirements of continuity, comformability and completeness for the flow

problem governed by Eqs. (12-1) and (12-14). It does not, however, include
all terms in the polynomial expansion, see Chapter 1 1
Step

4.

Use of

Derive Element Equations


either a variational or a residual procedure for the

problem gov-

erned by Eq. (12-3) will yield essentially the same results. We consider the
following variational function for the two-dimensional idealization

-JR[@' +()>*
We

note here that the terms

d Jt
d
\_ JP.
2 dx dx

and

i-^z
2 dy dy

(12-14)

Chapter 12

Potential,

Thermal and Fluid Flow

305

are similar to those in Eqs. (3-21) and (4-6) and can be considered to represent

a measure of energy.

The
of

tp

by taking partial derivatives

derivatives in Eq. (12-14) can be obtained

x and y

[Eq. (12-11)] with respect to

_
~

gx

dtp

dx

_
~

dtp ds
ds dx

_
~

dtp

dtp dt

~di

dx
(12-15)

dtp
g>

Since the

dy

ds

dtp^dt_

ds dy

dt

dy

are expressed in terms of local coordinates

s, t,

we

use the

following mathematical results based on the chain rule of differentiation in

order to find gx and gy

dN = dN

dx
dx ds

tr

ds

dNt== dN

dx

dx

dt

dN

dy
dy ds
{

6N

7
'

dy
}

dt

which as a general

rule, in

dy dt'

matrix notation, can be expressed as

d)

id]
dx

[*1
Ts

dx

dy

ds

ds

dx

[dt,

Tt

dy
d
Tt_ Idyl

>

The matrix

[J] is

dx
(12-16b)

[J]

d
Idyl

often referred to as the Jacobian matrix. Equation (12- 16b)

represents a set of simultaneous equations in which d/dx

unknowns. Solution by Cramer's

Id]
dx
>

[J]-i<

and d/dy are the

rule gives

>

dt

dy

dyl (d\

dt

ds

ds

dx

dx

(12-17)
dt

where the determinant

\J\ is called the


7

Jl

The terms

in

dx dy

~dJdI

{dt,

dx dy

(12-18)

dt ds'

Eq. (12-18) can be evaluated by using expressions for x and y

Eq. (12-13)] and

57

ds_

Jacobian,

[Eq. (12-12)].

For instance,

" t)Xl + * + s)(l " t)Xl


+ i(l -f s)(l + t)x + 1(1 + 0*J
= -1(1 - 0*i + i(l - t)x + 1(1 + 0*3 {X
l
fs *

"

5)(1

i(l

0*4

(12-19a)

306

Thermal and Fluid Flow

Potential,

Chapter 12

and so on, and

and so on. Then

-i(l

ds

(12-19b)

evaluated as

|/| is

ii-sitewf")-**")]
(12-20a)

By

setting

4 and j

1, 2, 3,

1, 2, 3,

4 for each

the

i,

summation

in Eq.

(12-20a) leads to |/| in matrix notation as

\-t -s +
-l+t
1+j
s-t -\-s
1-5
-1 t
S +

-l+s~

X
"gL-M

Xt

X,

X-i

-s-t

yi

l+t

(12-20b)

Expansion of /

in Eq. (12-20b) gives

= H(*i - x )(y - y - (x - x )(y - y


+ s[(x - x ){y, - y ) - (x, - x )(y - y
+ t[(x - x Xy - y - (x, - x )(y - y )]}
= i[^13^24 - *24.Fl3) + S(x 34 y l2 ~ *12j>34)
+ ^23^14 -^14^23)],
where x u = x Xj and y = y yj.
4)

4)

3)

4 )]

i}

Use of Eq.

(12-20c)

(12-17) allows computations of ds/dx, dt/dx, ds/dy,

and dt/dy

in Eq. (12-15) as
o

ds

dydk_

(iyis

dy

($*N

ai . 2 u\

Similarly,

dy'
dt

?i

from Eq.

(12-11),

fdN y \
= ~\T[\k~~SF
1

dy- \J\\M
Now

(12-21b)

Xi

\J]\&i dt
t

(12-21c)

(12-21d)

Xi )'

ds

we have

dN
% = -
dtp

m
and

dq>

dNim

f^-fff*

(12-22)

an4 Fluid Flow

Potential, Thermal,

Chapter 12

and (12-22)

Substitution of Eqs. (12-21)

307

into Eq. (12-15) finally leads to

d<p

dx
(12-23a)
dtp

&
The

indicial notation

instance, the
setting

first

and/

dN
m V(dN
x

+
,

When

h h i/'W ~df ~ ~5T ~J7)

(dN dN

1, 2, 3,

dN

6N,

(dN dN2

dN,

is

obtained by

dN

2\

dN dNA
~W) y \

(ON, dN,

W
x

added to the other three terms obtained by


4; i
3, j 1, 2, 3, 4; and i = 4, j = 1, 2, 3,

this is

1, 2, 3,

row

"

3\

\W ~dT ~ ~6T ~di) yz ^\d7Tt


x

first

4 as

dN, dN,\

\\T\\

Eq. (12-23) indicates double summation. For

in

term in the summation for the

Xi

setting
4,

(12-23b)

=2

we have

and j

d<p/dx.

Similar evaluation gives dtp/dy.

After relevant substitutions and rearrangements,


9\

#12

#13

#14

-"23

B"24

M{q,},

#11
B-,21
A

rr
-"22

(12-24a)

or
{g}

(12-24b)

where

#u

8|/|

0>24

-^34^ -y23*)>

#12

#13

8|J|

(-^13 +^34^+^140.

8T7t(
l

#14
8|/|

->

24

+y^s

-yut),

Ol3 -^12^+^230.
(12-24c)

#21

8|/|

8|/|

\X 13

X 34 S

8|7|

X 34 S

x \2 S

8~L7T
1

#24

B 22
B 23

(~ *24

13

*2 3 0,

^140

"T"

x l2 s

*140>

x 23 t).

Potential Thermal and Fluid Flow

308

Now, we

substitute {g} into

np =

Qp

[Eq. (12-14)] to yield

(12-25)

l\lBY[H[B]dxdy{q,l

{q,}

Chapter 12

where
[I]

0"

[C]
1

Taking derivatives of Q^, with respect to

en,
that

<p

we

obtain
(12-26a)

0,

is,

d<p x

dQ,

=
)==> sq p

(12-26b)

o,

d<p 3

dQ,

which leads to
r
{H} JJ[BnB]rf^{q,}

(12-27a)

or

\\lW[R]dxdy{%}

(12-27b)

or

Mi,} =
where [kj

is

(12-27c)

("I.

the element property matrix:

[kj

jj[BY[B]dxdy.

(12-28)

Numerical Integration

The

coefficients

k9ll

of [kj need integration. They look like

= ffah + Blt)dx4y
JJl(8|J

(j 2 4

-y 4S-y 23
3

t)

)2

</
|

(-^24

*34*

X 23

]^^

(12-29)

Potential, Thermal,

Chapter 12

and so on.

It

can be

difficult to

and

and Fluid Flow

309

evaluate these integrals in closed form, and

it

perform the integration numerically.


The idea of numerical integration in finite element analysis is similar to
integration by using well-known formulas such as the trapezoidal and
Simpson rules. In finite element applications, the Gauss-Legendre [3] formula
is

often convenient

is

often used. In general terms, numerical integration can be expressed as

efficient to

P F(x)dx = 2 F(

)W

Xi

(12-30)

t,

1=1

Jxi

where i denotes an integration point, m is the number of such points at which


the function is evaluated and summed [Fig. 12-4(a)], and the
are the
weighting function. That is, an approximate value of the integral in Eq. (1230) is obtained by finding the summation of the values of the function at a
number of points multiplied by a weighting function at each point. For the
stiffness matrix in the two-dimensional space of the quadrilateral, we have

[K]

f_JjBY[B]\J\dsdt
m

_n

^S E[B(*f
j=l j=l

y )F[B(*f,)]

|/(*,,/,)

W,0,

(12-31a)

and n are the number of integration points in the two coordinate


and
and Wj are corresponding weights. Note
that we transformed the integral to local coordinates by using \J\.
As an example, we can use m =2 and n 2 as shown in Fig. 12-4. Then
total integration points are 4, and we can write Eq. (12-3 la) as

where

directions [Fig. 12-4(b)]

W-t

[Bfe ORBfe,

The magnitudes of (s t t) and


which gives them in ready-made
{,

Example
It is

12-1.

t t)]\J(s t , t t)

(12-31b)

t.

can be obtained from available literature

tables

[3].

Numerical Integration

useful to understand the subject of numerical integration because

many

finite

element applications.

it is

used in

We illustrate here some of the steps for numerical

integration over a (square) quadrilateral [Fig. 12-4(c)]. In view of the fact that

hand

can be lengthy and cumbersome, details of


only one term will be attempted here further details on this element are also given
in Chapter 13. First we compute the common terms required for integration of term

calculations of

all

terms of matrix

[k,,]
;

ku

inEq.

From

(12-29).

Eq. (12-20c):

x3

=0- =
1

-1,

= 1 - = 1,
x4 = l -0 =

x4
*34

x3

1,

^24

= y 2 - y* =

yi3

=y\ -y$

yi2=yi -

= -i

=0- =
l

-l

=o -o = o,

F(x)

F(x)

III

II.

Si

s
s

2
3

sa

= ~a, t,
= + a, t :

+a U
#

- -a, t A

+a
+a

<b)

(1. 1)

(0, 1)

0, y,

2=1,72=0

(-1, 1)

(1,

'(3)

= 1.Y3 =

0,

v4 =

-f- Integration points

0,0

Point

(-1,-1)

2
3
4

(1,-D
*

-0.577
-0.577
+0.577
+0.577

(0.0)

-0.577
+0.577
+0.577
-0.577

1,0
(c)

Figure 12-4 Numerical integration, (a) Schematic representation


of numerical integration of a function, (b) Numerical integration

over quadrilateral,

310

(c)

Example of numerical

integration.

= x x2 = 1 = 1,
= x x3 = 1 1 =0,
= A _ x4 = 1 - 1 = 0,

A'12

x23
Xl4

and Fluid Flow

Potential, Thermal,

Chapter 12

>'

34

j>

14

=
=
=

y 23

-j

>'

vj

y2

311

- }'a =
- y4 =
-y =

=
=
=

0.

-1,

-1,

and
\J\

--{[(-D(-i)

where A

is

only the

first

(D(-i)]

unit 2

term contributes to |/| and


is

(-1X0)]

r[(0X-l)

- (OX-DD

the area of the square

square, which

sfdXO)

constant for

all

its

Note

that for a (square) quadrilateral

value

one-fourth of the area of the

is

points in the square.

Now, numerical

integration in

Eq. (12-29) can be expressed by using Eq. (12-3 lb) as

*,n

=S
r=

?ife, U)

Bi^s;. fJM J(*fc

tt)\

{,

1, 2, 3, 4.

From

the available numerical integration tables, such as in Ref.

[3],

we can

choose an appropriate number of points of integration and weighting function

t.

For a 2 x 2 or four-point integration with Gauss-Legendre quadrature, the points


of integrations are as shown in Fig. 12-4(c), and the weighting functions are W\ =
= W+ = 1. Here we have used the local coordinates of the integration
1V2 =
3
points only up to third decimal; for computer implementation, values with higher
decimal points are used. The computations for B :1 and B 2l are shown below in

tabular forms

B\\

B2i =

(?24

8|7|(- A'24

J34-S"

>'23^)

- x34S -

x 23 t).

Points of Integration

(-0.577, -0.577)

B 2i

(0.577,

-0.577)

(0.577,0.577)

1.577

1.577

0.423

1.577

0.423

0.423

1.577

Therefore,

(-0.577,0.577)

[Bh(s2i

t2 )

W,(*3, h)
[flf,(J4,

U)

- Bl {s 2
- Bl(s
+ Bl (s A
x

t 2 )]\J(s 2

t 3 )]\J(s 3

t 2 )\

t3

)\W3

U)]\J(s A U)\H>\
,

0.423


Potential, Thermal,

312

(-0.423) 2

^(9.9477

Other terms of
Step

[k 9 ]

+ (-

[(_1 577)2
-

(-0.423) 2

0.7157)

and the load

and Fluid Flow

577 ) 2

(-0.423) 2

= ^|^ =

(1-577)2

Chapter 12

(-0.423)2

(-1.577) 2 ]

0.667.

vector(s) can be evaluated in a similar

manner.

Assembly

5.

Equations (12-27) are assembled such that the potentials at common nodes

The

are compatible.

assemblage equations are

final

= {0}.

tKJ{r,}

Under
(12-32)

(12-32)

boundary conditions, Eq.


and then solved for nodal potentials.

the application of only Dirichlet-type


is

modified for given

(p t

Applied Fluxes

In addition to the flow caused by difference in velocity potentials, the

system can be subjected to a number of physical forcing parameters.

A fluid flux Q can be applied as a source or a sink, concentrated at node(s),


within the flow domain. Also, a fluid flow q can be applied at the boundary.

Then

the

boundary condition [Eq.


dtp

2?
dn
If

we provide
-

for

and

q,

(12-7b)]

(12-33)

0.

the variational function in Eq. (12-14)

becomes

Use of the

becomes

fl^ - J>*

<

12 - 34 >

stationary principle then leads to

WW =

C2-35a)

{Q3.

where
{Q}

JJ

[NY{Q}dxdy

[Wl$ds

(12-35b)

=
is

{Qi}

{Q 2 }

the nodal forcing parameter vector.


Evaluation of [Q]

This vector can be evaluated by using numerical integration. The expanded


forms of the two parts of {Q} are given below. The first part is

Chapter 12

Potential, Thermal,

{Qi}

\\lW{Q}dxdy

Ni(st

N (s h
N (s h
N4(s
Here we assumed

The second part

ii

[W{Q}\J\

ti

dsdt

)\w

t t)}

t t)

tt)

i9

313

= i:ms ,t )Y{Q}\j(s
i

and Fluid Flow

Q\J(s

,t t

)\w

(12-36)

t t)

to be a uniform flux.

relevant to the

is

boundary

only.

As an

illustration,

consider side 1-2 of an element (Fig. 12-5) subjected to components q x and qy


Then the second part for q x specializes to

N
N

{Q 2 }

(12-37a)

q x dS.

Figure 12-5 Boundary fluxes.

We

1,

i(l

note that along side

and

s varies

- s\ N =
2

{(W

(1

from

1-2,

w-

*)

*)

coordinate

has a constant value equal to

+1. Therefore, the N specialize


0, and NA = 0. Hence, we have

to

+ s), N =
10

qx dS =

K 1

N =
{

*)

fcl'IrfS.

to

(12-37b)

and Fluid Flow

Potential, Thermal,

314

Since s

= 2S/l

l9

we have
dS

where

is

lx

Chapter 12

lj-ds

\J\ds,

(12-38)

the length of side 1-2. Therefore,


fl

{Q 2 }

f,

-s]

(12-37c)

>q x ds,

which reduces to

(Q 2 }

(12-37d)

This indicates that the total flux on side 1-2 in the x direction
equally between nodes
the

x and y

and

2.

is

divided

Contributions of other applied fluxes in both

directions can be evaluated in an identical manner.

STREAM FUNCTION FORMULATION


Formulation by using stream function
with velocity potential

(p.

The

differential

g+$=

as an

y/

is

similar to that

vy = o,

or

and the corresponding variational functional

unknown

equation for flow with

y/ is

(12-39)

is

-am >*
+

(12-40)

Stream function

y/

can

now be

V=[N]{q}

By following

expressed as

N,.^.

(12-41)

a procedure similar to the one above, the element equations

are
(12-42)

[Mq,}={0],
where [kj

is

the element property matrix

and {%}

vector of nodal stream functions.

The boundary conditions

are expressed as
y/

where S

is

the part

on which

y/ is

y/

onS b

prescribed.

= WiWi

s V*]

ls

the

Secondary Quantities

The

velocities

can be computed by using Eq.

(12-5).

For the velocity

potential approach,
Pi

^11

#12

^13

#14

.-#21

-#22

^#23

#24j

fi

(12-43)

(Pa)

and

for the stream function approach.

V:
-#21

-#22

-#23

_
L -#11

""-#12

_ ^13

Vi

^24
~~

(12-44)

-#14.

Wa)

The quantity of flow Q f across

now found

can be

a section

A-A

in

an element

(Fig. 12-6)

as

QfM =AV

(12-45)

n,

where A is the cross-sectional area at section A-A and Vn is the velocity


normal to the section. Vn can be computed from the two components v x and
v y for an element as
vXt sin 9

where

denotes an element and 9

the horizontal.

The

is

Vy,

cos 9,

the angle between the section

total flow across a section

fj

(12-46a)

2,A V
t

where j denotes the cross section and

(12-46b)

A-A and

can be found as

is

the total

number of elements

across the cross section.


Figure 12-6 Computation of quantity of flow.

315

Example

Potential

12-2.

Flow Around Cylinder

shows the problem of uniform


around a cylinder of unit radius. The flow domain is 8 x 8 units. Due to
symmetry, only one-half of the domain is discretized, as shown in Fig. 12-7(b).
(a)

Velocity Potential Solution. Figure 12-7(a)

fluid flow

Figure 12-7 Analysis for potential flow around cylinder, (a) Flow

around

cylinder, (b) Finite element

(0, 0) etc.,

mesh

for half flow domain.

denote coordinates.

- J

4 y

t^
4^

D
+->

'c

00

'

L = 8 units
(a)

(8,4)

(8,0)
2)
^"^

Orinin for
fnr
Origin
finite

element

analysis

316

(b)

(4J

"- Origin for closed

(46)

(gl)

form solutions

Potential, Thermal,

Chapter 12

The boundary conditions


(p

(p

and Fluid Flow

317

for the velocity potential are

unit along the

upstream boundary,

nodes 1-5,

unit along the

downstream boundary,

nodes 51-55.

(12-47)

The computer code FIELD-2DFE

(for further details, see

Appendix

4)

was used

to obtain numerical solutions. Solutions for velocity potential are obtained by


setting

NTYPE =

2 in the code.

Figure 12-8 shows computed distribution values of velocity potentials at nodes


in half of the flow

tions

from

domain. Table 12-1 gives comparisons between closed form solu[2] and numerical results for cp at the nodes along the line y = 0.

Streeter

The formula

for closed

form solution
<p

where
[<p(x

is

the uniform undisturbed

0)

<p(x

8 [Fig. 12-7(a)],

= L)]/L = (1 r = (x 2 y
2

measured from x

[2]

+ y) cos 6 -

u(r

given by

is

0)/8

velocity

is

0.125,

(12-48)

0.50,

in

the negative

x direction
domain

length of the flow

the radius of cylinder,

and 6

is

=
=

the angle

axis.

TABLE

Comparison of Computed and Exact

12-1

Solutions for

Node

Computed

Exact

26

0.5000

0.5000

31

0.3743

0.3780

36

0.2500

0.2765

41

0.1875

0.2132

51

-0.03125

0.0000*

Prescribed.

Stream Function Solution. Specification of

(b)

2DFE

NTYPE =

3 in the

code FIELD-

permits solution for the stream function approach. For this solution, the fol-

lowing boundary conditions were used:

W =
y/

0.5 along the top

nodes 5-55,
(12-49)

0.0 along the

bottom nodes, 1-16-26-36-51.

Figure 12-9 shows the distribution of computed values of y/. Table 12-2 gives
comparisons between closed form solutions and numerical predictions for y/ along
the line of nodes 26-30.

For the
(0.5

line considered,

0.0)/4

0.125.

The closed form

solution

y/

= u(r -y)sin#.

90 deg, and

(x 2

is

given by

[2]

(12-50)

y2

y.

The value of

U=

'"

'

'

"

'

'

paquosajj

o
lo
o

1
1

//

i
J

o
LO
o

/o

/o
/

LO

r^-

LO

o
CO
CO

CO
r>.

1^

1
1

^
d

**

\o

V
\

w
\
\ o
V*"""

\
\^

fr

CO

\ o
\

CO
r^

r^
r^

"~"

0)
W)

00
00

\\ d
*

00
r-

Oi
00

o
o

o
o

o
o

,_

paquosajj

318

0)

Q.

^M

QX
OJ

\\

vv

'

o
o
"

T3

"

O
D

\o

00
00

o
a

C
-c

00
00

CD

j-

-C T3

s
CO
I

<L)

o
a

.2

>^

N
*

CTl

0)

**>

00
LO

o
o

JO

3
a
F3

+3

o Z a>
w o O

00
<4

</>

CO

a.

aj

re

(U

c c

o
c

o o

CD

h
3
00
to

'

1
i

05

n
o
N

D
n
n

a>
1

ol

/ / /^

r^
^

Id

^
o

"

<

^
\

^
*\ o

_-.

/
/

/in

s*
I

\ CN
\ d
\

as

-a

s^
CO
CO

o
d

\.

0)

-a

2:5

\^

\.

\.

O _
J^ c
CO
O
C C
o o
o o
c c
3 3
>4- Hb b
ra
<D

*-

a
<U

->

CO CO
i- CN

co
CN

"-"

"

?
CN

CN

p^

m
d

f^
CO

o
c

o
3

9s
ri

CO
CO

a
s

c
C

d
\
\

53
a>

'o

o
o

^
CN

CN

319

Potential, Thermal,

320

TABLE

and Fluid Flow

Chapter 12

Comparisons Between Computed


and Exact Solution for y/

12-2

Node

Exact

26

0.0000

0.0000*

27

0.1473

0.1529

28

0.2625

0.2746

29

0.3678

0.3892

30

0.4688

0.5000*

Computed

Prescribed.

Figure 12-10 shows comparisons between nondimensionalized values of the


x component of velocity v x along the section of nodes 26-30. The closed form value
is obtained from [2]

or
1

The numerical

+~

(12-51)

results plotted in Fig. 12-10 are the values of v x in element(s) adjoin-

ing the section.

The comparisons

in Tables 12-1

and 12-2 indicate

that the finite element predic-

The plots for v x /U in Fig.


and stream function approach give
potential approach yields the lower bound

tions yield satisfactory solutions for the flow problem.

12-10

show

that both the velocity potential

satisfactory comparisons.

The

velocity

Figure 12-10 Comparisons of velocities at midsection.

2.0

V*-Closed form
Stream function

i_
1.0

f
Velocity potential

i.i. i.i.i
y =

26

28

27

Nodes

29

30

Potential, Thermal,

Chapter 12

and Fluid Flow

321

and the stream function approach the upper bound of the exact velocity.
numerical solutions can be improved by using finer meshes.

If neces-

sary, the

THERMAL OR HEAT FLOW PROBLEM


In the case of the heat flow problem, the general governing equation
essentially the

are different.

is

same as Eq. (12- la) except that the meanings of various terms
The unknown can now be the temperature J at a point; k x
,

ky and k z are thermal conductivites; Q is the (internal) applied heat flux;


and q is the (surface) intensity of heat input. In addition, there is the possibility of heat transfer due to the difference between the temperature of the
medium, T, and the surroundings, TQ The differential equation for twodimensional steady-state heat flow then becomes
,

T
k**
+ k*
dx
dy
y

2 =

(12-52)

0,

and the boundary conditions are

T= 1

on S

(12-53a)

and
kx

tx

d~x

where a

The

is

>o\

C>

ol(T

-T )-q =

onS

and 5

(12-53b)

the heat transfer coefficient.

finite

element formulation with the quadrilateral element

will result

in equations similar to Eq. (12-35) with the addition of the following

term in

CI, [Eq. (12-34)],

i<x(r

which

- T Y-ciS.

(12-54)

[kj and {Q 3 } to the element

will lead to additional contributions

matrix and the forcing parameter vector, respectively:

For evaluation of

[kj

{<2 3 }

[k a ]

a[N] r [N]</S,

(12-55a)

*\Nr{T

(12-55b)

we need

]dS.

to integrate along the sides (boundaries)

of the element. As an illustration, consider side 1-2 (Fig. 12-5) as before:

= ^^lW[K]dS
c

[K]

-*

i(l

*)

>[(l-s)

i(l-s)

0]dS

Potential, Thermal,

322

i(l

i(l

* )

i(l

and Fluid Flow

Chapter 12

s )
s)

dS
0.

"W
olI x

o-

(12-56)

Computation of {Q 2 }

similar to that for q x in Eq. (12-37a).

is

Finally the element equations for heat flow will be

(M +
where

{q r

f - [^ T T

Example

As a

IW){qr}
2

12-3.

= {Q} = {6,} +

{Qi}

{Qil

(12-57)

rj.

Two-Dimensional Heat Flow

simple illustration, consider steady-state heat flow in a rectangular block of

The heat flow domain is divided


The boundary conditions are assumed as follows:

unit thickness (Fig. 12-11).

12 nodes.

=
7X3, y) =

7X0, y)

into 9 elements with

100 deg,
(12-58)

25 deg.

conductivities are assumed to be equal as k x = ky = 1


Table 12-3 shows the computed values of temperatures at various nodes ob-

The thermal

tained by using the code

FIELD-2DFE.

TABLE

12-3

Node

Numerical Results
for Temperatures
Temperature (deg)

100.0000*

100.0000*

100.0000*

74.9998

74.9998

74.9998

49.9999

49.9999

49.9999
25.0000*

10
11

25.0000*

12

25.0000*

'Prescribed.

Potential, Thermal,

Chapter 12

and Fluid Flow

323

3 units

Figure 12-11 Heat flow in two-dimensional body.

SEEPAGE
Seepage

is

defined as flow of fluid, usually water, through porous (soil) media.

The governing equation

is

similar to Eq. (12- la) except the

meanings of

various terms are

Unknown,

(p

k x ky k 2
,

the total fluid head or potential,

is

coefficients of permeability in the x, y, z

(12-59)

directions, respectively,

Q =
q

applied (internal) fluid flux,


applied (surface) intensity of fluid,

where p is the fluid pressure, y = density of water, and z is the elevation


head [Fig. 12-12(a)].
For two-dimensional steady-state confined flow, the element equations
are identical to Eq. (12-35) except for the meaning of the terms. Often seepage
is

defined as confined or unconfined

occurs through a saturated


ditions

medium

and does not involve the

[4].

so-called free or phreatic surface. In the case

of unconfined seepage, on the other hand,


surface.

In the confined category, seepage

subjected to prescribed boundary con-

Both categories can involve

we encounter

the free or phreatic

either steady, or unsteady or transient

conditions.

When
12-12(b)],

and surface of seepage


mixed boundary conditions on these surfaces:

the seepage occurs with a free surface

we have

the

q>

[Fig.

(12-60a)

and

**
911

= o.

(12-60b)

Potential, Thermal,

324

and Fluid Flow

Chapter 12

U
(a)

Free surface

Surface seepage

0,0,0

(2.0)

(0.0)
:

M: -:0
-

(2.0)

(0.0)

,a5,

(c)

Figure 12-12 Seepage through porous media, (a) Three-dimensional


seepage, (b) Free surface

and mixed boundary conditions,

(c)

Steady

seepage in porous medium.

At the free surface, the pressure is atmospheric; hence, p = and cp = z.


The second condition implies that the velocity of flow normal to the free
surface is zero. The two conditions [Eq. (12-60)] must be satisfied simultaneously. The mixed conditions render the problem nonlinear. Consequently, in contrast to the steady confined flow, which requires only one finite

element solution,

it is

often necessary to perform iterative analysis in the case

of free surface flow. This requires a number of


topic

is

beyond the scope of

finite

element solutions. This

this text; the interested reader

can consult

Refs. [4]-[6].

Example

12-4.

Steady Confined Seepage

Figure 12-1 2(c) shows a rather simple problem of seepage through porous (soil)
media with coefficients of permeability k x = ky = 0.005(L/r) and thickness =

Chapter 12

Potential.

The boundary conditions

unit.

q
O

=
=

Thermal and Fluid Flow

are

=
=

atx

at

The computed values of

at

<p

525

0.

(12-61)

2 units.

(12-61)

various nodes are shown in Fig. 12-12

parentheses.

The computed values of

velocities are

shown

in

Table 12-4. These values com-

pare closely with the velocities computed from Dairy's law:

TABLE

Computed Seepage Velocities

12-4

Element

0.005

-0.45

10"*

0.005

-0.1S

10-*

As expected, the

--:

=
S
dx

0.005

0.005.

\2

components of velocity are very small, almost equal


12-46b was found to be equal to 0.01 unit.

;.

quantity of flow [Eq.

Example

V,

'.,

to zero.

Steady Confined Seepage Through Foundation

12-5.

The foregoing problem is rather simple and can permit hand calculations.
consider a problem that is more difficult and needs use of the computer.

problem of steady confined flow through the foundation

(dam. sheet

pile)

shown

is

with different coefficients


is.

kx

k:

this

included in the

The

.'

we

of a structure

The foundation consists of two layers


of permeability. The soil is assumed to be isotropic, that
in Fig. 12-13' a).

assumption
finite

soil

Now

is

not necessary because anisotropic properties can be

element procedure. The structure

itself is

assumed

to be

impervious. The two-la>ered foundation rests on a material which has a very low
permeability; hence,
implies a zero flow
Stead;, rtuid
tively,

For

this

at

the

depth

natural or

heads o, and

ot"

'.'.

m.

impervious base

Neumann) boundary

:"-

act :n the

is

assumed. This

condition.

upstream and downstream

sides, respec-

problem, we have assumed

a =
g, =

m.
(12-62)

0.

which constitutes the geometric (or forced or Diriehlet) boundary condition. Because
the problem is linear, results for this boundary condition can be extended to any
multiple of fm and fd with proportional difference in the head. For instance, the
= 100 m and (p d = 50 m
subsequent results can be used to derive solutions for
by multiplying the computed nodal heads by 50.
The foundation medium extends toward "infinity" in the lateral direction.
Because we can include only a finite pan of this extent in the analysis, an approximation must be made to fix the discretized end boundaries. For instance, in

20

Impervious base (natural boundary condition)


(a)

<p

<

3<

Discretized end boundary

(b)

^z

Discretized end boundary

7^

V//////////////A

"
i

\
\
\
i

\
1

50

40 30

20

10
Percent

(c)

Figure 12-13 Steady confined seepage through layered foundation,


(a) Details of foundation, (b) Finite element mesh, (c) Solution
for nodal potentials: equipotential lines.

326

52"

and Fluid Flo*

Potential. Thermal,

Chapter 12

m (equal to the

en the end boundaries at a distance of 20

Fig. 12-

width of the stru::_:e Groin the edges of the structure;


that at that distar.ee the Bold r

ten! ia

ill

be app:

on the upstream and downstream

potentials

this involves

an assumption
:he applied

sides. Often,

it

may

be necessary to

perform parametric studies to find the distances at which such conditions can be
this topic appears in Chapter 13.

assumed. Further discussion on


Figure 12-13
is

needed

>r

at the interface

mesh m ith -5 modes and 72 elements. A nodal line


between the two layers. The geometric boundary condi-

a finite

upstream and downstream nodes are p. = 1 m and p d = m. resre:have dq> dz = 0: being the natural boundary
condition, it is satisfied automatically in an integrated sense. Consequently, in the
tions at the

me bonom boundary we

A:

fin te

element analysis,

problem
of the

is

we computed

fluid

heads at nodes

symmetric, we could have solved

medium

is,

al

mis bounda:;-

Since the

by considering mesh only for one-half

:
Then at the centerline we can assume a natural boundf/dx = 0. This is valid since at the centerline the rate of

[Fig. 12-13

dary condition; that

it

change of g van: sr.es


Figure 12-13 ; thowa the finite element computations for nodal heads ob:a ined
by using the code FIELD-2DFE. The equipotential lines are shown as a percentage
of the total head difference fB fd
.

From
on the

the results in Fig. 12-15

structure.

Moreover, the

we :an

finite

velocities in all or selected elements

find the seepage :~::.es

easing

uplift

element procedure allows computations of

and the quantity of flow

at selected secticr.s

AM

this

information can be used for analysis and design of structures founded on porous

soil

foundations through which seer-ge occurs.

ELECTROMAGNETIC PROBLEMS
In the case of steady-state electromagnetic problems the governing differential equation, in the

absence of

in

Eq. (12-la). reduces to the Laplace

equation

or

if

kx

A.\.

V-r =
where

is

(12-

0.

the electric or magnetic potential.

V:

is

the Laplacian operator.

and k x and k y are the material properties; in the case


electrical flow they
are the permittivities in the x and y directions, respectn ely.
The finite element formulations and details will follow essentially the same
procedures as in the :^se of the field problems covered in this chapter and
Chapter 11. For further details on this topic the reader can consult various
:"

publications

Thus,

[7, 8].

we once again observe

the similarities of various

phenomena

in

328

Potential,

Thermal and Fluid Flow

Chapter 12

and physics and how the


method provides a common ground for their solutions.
different disciplines in engineering

COMPUTER CODE

finite

element

FIELD-2DFE

problems and permits solution of torsion,


and heat flow. The user needs to supply
appropriate properties relevant to the specific problem and specify codes for
the type of problem as follows

This

is

a general code for

field

potential flow, steady-state seepage,

Problem

Code,

Torsion

NTYPE
1

Potential flow

Velocity potential

Stream function
Seepage
Heat flow

In

fact,

the code can be used easily for other steady-state problems such as

and magnetic

electrostatic

Appendix

flow. Further details of the

code are given in

4.

PROBLEMS
12-1.

Derive in detail
(see

12-2.

Chapter

[k]

and {Q}

for potential flow

by using a triangular element

11).

Find {Q 2 } [Eq. (12-37)]

if

q y and q x were applied on sides 1-2 and 2-3,

respectively.

12-3.

By using numerical

integration,

compute the

coefficients

k9tt and k 9li

for

the square element in Fig. 12-4(c).


12-4.

By hand

calculation, find the element matrix [k] for steady-state heat flow

in a square plate of unit thickness (Fig. 12-14) discretized in four equal

square elements. The boundary conditions are

T(x,y)

100 deg,

7X2,>0=Odeg,
and ky

ky

unit.

Assemble the equations for the four elements and


5, and 6. Hint: You may use the results

solve for temperatures at nodes 4,

and Fluid Flow

Potential, Thermal,

Chapter 12

329

t(9)

(0,2)
(2,2)

A,

A A

(0, 0)

(2, 0)

Figure 12-14

indicated in Eq. (12-3 la) and the numerical integration


12-1. Solution:

12-5.

Tat nodes

4, 5,

and 6

shown

in

Example

50 deg.

In Prob. 12-4, consider heat flux q


0.1 unit per unit length along the side
3-6-9 and compute temperatures at nodes 4, 5, and 6. Hint: You may use
the results in Eq. (12-37) for finding the forcing function vector {Q}.

12-6.

In Prob. 12-4, consider a source (or sink)


the temperatures at nodes, 4,

5,

and

6.

Q = 0.1

Hint:

units at

You may

node

and find

use Eq. (12-36) to

evaluate {Q}, which will need numerical integration similar to that used in

Example
12-7.

12-1.

Use the same domain


such as a

soil

as in Prob. 12-4 but

kx

and applied

fluid potentials

ky

(p{2, v)

(a) fluid potentials at

in all the elements;

12-8.

12-9.

(c)

=
=

nodes

10 cm,
5

cm.

4, 5,

Consider Prob. 12-7 but include


4, 5,

and

and

6; (b) velocities v x

and

vy

the quantity of flow across a cross section.

Consider Prob. 12-7 but include q = 0.1


and find fluid heads at nodes 4, 5, and 6.

heads at nodes
12-10.

and

cm/sec

0.1

equal to
(p{0,y)

Compute

composed of a porous medium

with

cm 3 /sec-cm

Q = 0.1 cm

at

along the side 3-6-9

node 6 and compute

fluid

6.

Consider Example 12-5. Prepare three meshes with discretized boundaries

and 40 m from the edges of the structure. By using a computer


compare the results from the three analyses and offer comments on
the effect of the distance of the end boundaries on the numerical predicat 10, 20,

code,

tions of the heads

under the structure.

330

Potential, Thermal,

and Fluid Flow

By using FIELD-2DFE or another

12-11.

Chapter 12

available code, solve for steady tem-

perature distribution in a composite material for the conditions

shown

in

Fig. 12-15.

Insulated

T=

T=

100 deg

0.0 deg

8 units
Figure 12-15

12-12.

By

using

FIELD-2DFE

or another available code, solve for seepage in the

foundation of the sheet pile wall shown in Fig. 12-16.

Figure 12-16

12-13.

Derive the

finite

element formulation for

node isoparametric quadrilateral

field

(Fig. 12-17).

problems by using an

eight-

and Fluid Flow

Potential, Thermal,

Chapter 12

331

(-1,1)
(1, 1)

(-1,0)

(1,0)
,

(-1,1)

(0

Figure 12-17

Partial results:

Assume

u
/

[A7"=<

where

[iV]{q},

- 5)(1 - /)(1 ~S ~t)\


-id s){\ - t){\ -s-t)
-1(1 - s){\ - r)(l - 5 - f)
-id - s){\ - r)(l -s-t)
id - **xi - t)
-i(l

i(l

-5)(1 -/*)

id -

{qf
/1(1

&"-<

Usee)f

u3

u4

u5

u6

t)(2s

JJ

-5)(2/-5)\

1(1

-5)(2r-5)

t)

i(l

-5)(2r-5)

i(l

pr

=<

t)
y

-/(1-J)

then leads to

[brc][B]^

and
{Q}

5)
)

etc.,

-id - *
"/(I ~ *)
2
id - *

t)

S )(2t

U B ],

/i(l

Eqs. (12 -16), (12- 17),

u-

t)

-t)(2s-t)

s(\ id -s{\ -id -

[k]

f)(25

u2

[i

-t)(2s-t))

#1 1(1

1(1

**xi

id -5)(1-^)

>>

\\[XY{Q}dxdy

[NF[-

REFERENCES
[1]

[2]

[3]

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Streeter, V.

L.,

to the Finite

Fluid Dynamics, McGraw-Hill,

New

Element Method, Van

York, 1948.

Abramowitz, M., and Stegun, I. A., (eds.), Handbook Mathematical Functions


with Formulas, Graphs, and Mathematical Tables, Applied Math. Series 55,
National Bureau of Standards, Washington, D.C., 1964.

[4]

Desai, C.

and Christian,

S.,

[5]

Desai, C.

J.

T. (eds.), Numerical

New

Engineering, McGraw-Hill,

Methods

Desai, C.

S.,

[8]

98,

Silvester, P., and Chari,

M.

Magnetic Field Problems,"

IEEE

in

Geotech. Eng.,

Fields by Finite Elements,"

/.

Soil

No. SMI, 1972.

V. K., "Finite Element Solution of Saturable

Chari, M. V. K., and Silvester,

PAS

FEM

"Seepage Analysis of Earth Banks Under Drawdown,"

Mech. Found. Eng., ASCE, Vol.


[7]

Geotechnical

"Finite Element Procedures for Seepage Analysis Using an

S.,

Isoparametric Element," in Proc. Symp. on Appl. of


Waterways Expt. Station, Vicksburg, Miss., 1972.
[6]

in

York, 1977.

PAS

Trans., Vol.
P.,

IEEE

89, 1970, pp. 1642-1651.

"Analysis of Turbo-Alternator Magnetic

Trans.

Power Apparatus and Systems, Vol.

90, 1971, pp. 454-464.

BIBLIOGRAPHY
Desai, C.

S.,

ments

"Finite Element

in Fluids,

kiewicz, O.

Doherty, W.

C,

P.,

Vol.

(eds.),

Methods

for

Flow

in

Porous Media,"

(Gallagher, R. H., Oden,

Wiley,

Wilson, E.

L.,

New

J.

T., Taylor,

York, 1975, Chap.

and Taylor, R.

in Finite Ele-

C, and

Zien-

8.

L., "Stress

Analysis of Axisym-

metric Solids Utilizing Higher Order Quadrilateral Finite Elements," Report


69-3, Struct. Engg. Lab., Univ. of Calif., Berkeley, Jan. 1969.

Ergatoudis,

I.,

Irons, B. M., and Zienkiewicz, O.

C, "Curved,

Quadrilateral Elements for Finite Element Analysis,"


4,

No.

1,

Int. J.

Isoparametric

Solids Struct., Vol.

1968.

Zienkiewicz, O. C, Mayer, P., and Cheung, Y. K., "Solution of Anisotropic


Seepage by Finite Elements," /. Eng. Mech. Div. ASCE, Vol. 92, No. EMI,
1966.

332

TWO-DIMENSIONAL
STRESS-DEFORMATION
ANALYSIS

INTRODUCTION
number of one-dimensional problems and two-dimensional
problems with only one unknown or degree of freedom at a point, we
are now ready to consider a different class of two-dimensional problems. This
class involves analysis of stress and deformations with more than one degree
of freedom at a point.
After studying a
field

Most real problems are three-dimensional. Under certain assumptions,


which can depend on the geometrical and loading characteristics, it is possible
to approximate many of them as two-dimensional. Such two-dimensional
approximations generally involve two categories: plane deformations and
bending deformations. In the case of plane deformation, we encounter
subcategories such as plane stress, plane strain, and axisymmetric; in the case
of bending, we deal with problems such as bending of plates, slabs, and
pavements.

PLANE DEFORMATIONS
Plane Stress Idealization

Figure 13-1 shows a (thin)

beam and

applied in the plane of the structure, that


small

compared

a plate subjected to loads that are


is,

in the x-y plane.

to the x-y dimensions of the body.

The

thickness

is

Such loadings are often


333

Two- Dimensional Stress- Deformation Analysis

334

y,v

Chapter 13

y,v

-p-J

(a)

(b)

Figure 13.1 Plane stress approximation, (a) Beam, (b) Plate.

membrane

referred to as in-plane or

(stretching).

Under

these conditions,

the assumption that the variation of stresses with respect to


the body,

is

constant,

it is

z,

that

is,

and

across

reasonable to assume that out of the six com-

ponents of stresses in a three-dimensional body [1-3] three of them, a 2 Tgx


and r yz can be ignored in comparison to the remaining three, a xi a y and
,

T xy This idealization
.

nonzero

stress

is

called plane stress

and involves only the following

components,

(13-1)

which are functions of the coordinates

x,

only.

The corresponding com-

ponents of strains are


*

(13-2)
[y,

a and denote components of normal stress and


strain and t and y denote components of shear stress and strain, respectively,
and {a} T = [o x a y T xy] and {e} r = [e x e y y xy] are the vectors of stress and
strain components. In view of the plane stress assumption, we need to consider only two components of displacements at a point, u and v, in the x and y
In Eqs. (13-1) and (13-2),

directions, respectively.
If

we

and isotropic materials, the

restrict ourselves to linear, elastic,

material behavior can be expressed by using the generalized Hooke's law for
the three

components of

Gv

and

stress

E
vE

V
+

vE

E
2(1

strain [1, 2].

v)'

Thus

,2'->'>

(13-3a)

Two- Dimensional Stress- Deformation Analysis

Chapter 13

335

or in matrix notation,

w=

rate]
1

],

(13-3b)

and v are Young's

-v
2

where [C] is stress-strain or constitutive matrix and


modulus and Poisson's ratio, respectively.
Plane Strain Idealization

is large compared to the x-y dimensions


and where the loads are acting only in the plane of the structure,
that is, the x-y plane, it can be assumed that the displacement component w
in the z direction is negligible and that the in-plane displacements w, v are
independent of z. This approximation is called plane strain, in which case the
nonzero stress components are given by

In cases where the thickness

(Fig. 13-2)

w=
and a z

= v(a x +

a y). The

(13-4)

stress-strain relationship for this idealization

is

expressed as

1-v

W=

v
1

[C]{]
(1

v)(l

-v

2v)

sym.

(13-5)

-2v

y, v

z,

y~^

x u
'

UL

(b)

(a)

Figure 13-2 Plane strain approximation, (a) Strip load, (b)

Long

underground tunnel.

Axisymmetric Idealization

Figure 13-3 shows a body symmetrical about

its

centerline axis

and

subjected to a load symmetrical about the axis. In view of the symmetry, the

Two- Dimensional Stress- Deformation Analysis

336

'

Chapter 13

'

>z,

Figure 13-3 Axisymmetric approximation.

components of stress are independent of the circumferential coordinate 6. As


a consequence, we have the following nonzero stress and strain components

[2]:

'Or

^r

\>

{}=rT

(13-6)

[yj
Dr this

case

[C]{]
(1

v)(l

is

V
V

V
1

2v)
sy m.

-v

[el
1

(13-7)

-2v

Strain-Displacement Relations

From
we can

the theory of elasticity, assuming small strains and deformations,


define the following strain-displacement relations for the three

idealizations [2]:

Plane Stress and Strain.

du

dx
dv
dy

t>

du
y xy

(13-8a)

dv

\dy

dx

Axisymmetric:

du

dr

<

(13-8b)

<

dw
dz

du

Initial Stress
initial

dw

dz^

',

dr)

and Strain: As explained

in

Chapter

5, it is

possible to include

or residual strains or stresses existing in the structure before the load

is

applied.

The

initial strain (or stress) state

temperature, (fluid) pressure, creep


in the case

of temperature,

where dT

caused by factors such as known


and geostatic stresses. For instance,

may be

effects,

= T T

is

T
\

*dT,

(13-9)

the change in temperature

and a

is

the coefficient of

thermal expansion.

We

define total strain e as the

sum of

the effective elastic strain, e % and

the initial strain


{}

where

{e

} is

strain case

[C]"

= [q({e}-{e
which

[D], in

{},

(13-10)

and

the vector of initial strains

to
The matrix

=[C]" '{}

= [q{}

})

(13-11)

[D], the strain-stress matrix, for the

plane

is

v
1

[D]

-v

(13-12)

337

ELEMENT FORMULATION

FINITE

As shown

in Fig.

element discretization will involve two-

13-4, the finite

dimensional elements such as triangles and quadrilaterials (squares, rectangles, trapezoids) in the

x-y plane. The third dimension

by specifying unit thickness for plane


plane

(z) is

generally included

strain or a thickness h in the case of

stress.

y.v
P(x,y)
(L 1( L 2

,L 3

P(x, y)
(s, t)

Figure 13-4 Discretization with triangular and quadrilateral ele-

ments.

Detailed properties of the triangular and quadrilateral elements have

been covered

in

Chapters

the quadrilateral element

and 12. Here we shall discuss in detail the use of


and then state briefly the use of the triangular

11

element.

As

stated previously, there are

point P(x, y) [Fig. 13-4].


point in the element as
u ( x ,y)

v(x,y)

We

two unknown displacements

can write approximation models for

= i + &
= fi -F

fi 2

+
+ Prf +

-r

cc 3

cc 4

u,
u,

v at a
v at a

xy,
(13-13a)

fi A xy,

or

{u}

where
338

ju}

[u

v],

[a}

= [a

/1

(13-13b)

[<D]{<x},

cc 2

a 3 a4

f} 2

fi 3

fi 4 ],

and [O]

is

the

Chapter 13

~i::.\

::'

Di'j

.o-D:

coordinates. Evaluation

2;X

Pz x

P:

2:V.

Pis

:::

':Yr

o-ces gives

c:

a,
PS-

2.5.4.

or

vnere

=>

q-

ments and \V
Eo.

15- 14b

:<

:<

:*'..

A
Tne rrccuct
ere t~e

[^[[Aj-

"<X>

a^

>:

bo.ut.ion ::

c,

v*

--e

h i

that

is.

-,.

= * " 5X1 - 0,
= ifl -^Xl+0,
1

(15-16)

= K1 - *Xi + 0-

are local coordinates (Figs. 12-2

The geometr;.

15-15

ons hV.
"*-'"*

dehned n Eg

V;

tne vec

:s

,.

matrix of nocal

"A"

V;

5.

v =

Here

15-14:

for "a" gives

\v v

the square iS

:s

the

::.

;.

re expressed b> using :he sam;

and 15-4

coordinates

i:

any point

erpolation functions

r=VYx

ine

emeni. car

1=17-4
il3-l"ai

V-

"

'

::

il5-l-b>

(2x8)

(2x1)

wnereW = [x,
This leads

xa

:o :ne

jc3

xJandfoF

dehninon

o:

(8x1)

|>, y2 yl
:he elemen: is :

dement.
Requirements for Approximation Function

within the dement. This

polynomial form as

is

in Eq.

sutisrieu since
>

15-15

we have

z r zy\i: r ::

Two- Dimensional

340

Stress- Deformation Analysis

For plane problems, the approximation models must

Chapter 13

satisfy the inter-

element compatibility at least up to derivative of order zero; that

placement between adjoining elements must be compatible. This

is

is,

the dis-

tied in with

the highest order of derivative n in the energy function, Eq. (13-21) below.

Since n

bility is

equal to n

in Eq. (13-21), the

minimum

order for interelement compati-

=0.

The approximation of Eqs.


distributions of u

(13-13) and (13-15) yields continuous bilinear


and v within an element and linear distributions along the

element boundaries.

common

It

can be seen (Fig. 13-5) that the displacements along

two elements afe compatible since only the straight line can
pass through two nodal displacements common to both elements. For the
sides of

two-dimensional plane deformation problems, the approximation function


provides for rigid body displacements and constant states of strains: e x

However, the function does not include all terms in the polynomial
expansion represented by the Pascal's triangle, Chapter 11.
e y y xy
,

Equal

nes a 1 b, and a 2 b 2
coincide

Equal
Figure 13-5 Interelement compatibility.

Plane Stress Idealization


First
in

we

Chapter

By following the procedure


components can be evaluated as

consider the plane stress idealization.


12, the strain

_ du
dx

_
~

duds
ds

\J\

du dt

dx^Ttfa

&ML'\ds

dt

dt

dsF

(13-18a)

Similarly,

dv
dy

_
~~

dvd

dvdt^

ds dy

dt dy

(13-18b)

Two-Dimensional Stress-Deformation Analysis

Chapter 13

341

and
7xy

du ^ dv
+ dx
dy
.

_
~

dt\
(duds
(du d j_dz
dy \

(dv
$1
(1 ds
\ds dx

\ds dy

dt

'

i,

dy)

'

dv i\
dr
dx)

_
,

OH.

(13-18c)

dt

'

we have

Finally,

B 12

~B n

B 21

B 13

B 22

B 14
B lx
B xx

B 24

B 23

B 23
B l2

B 22
B 12

B 24

{q}

(13-19)

B 14

or
(13-20)

[B]{q},

where the
Step

We
is

are defined in Eq. (12-24c)

is

defined in Eq. (13-14b).

[3]

t If wmm* *
6

where {X} r

Ty

{q}

use the principle of stationary potential energy; the potential energy

JJ

[Tx

and

Derive Element Equations

4.

given by

n' =

tj

MT*kfc* - h \ {<m^,

3 - 21 )

[X Y]

components of body

the vector of

is

r
forces; {T}

the x and y

components of surface tractions in


and h is the (uniform) thickness of the element.
Substitution for {u} and {e} from Eqs. (13-15) and (13-20) in Eq. (13-21)
]

is

the vector of

direc-

tions, respectively;

yields

H,

=4

r
fa}

|1

(1x8)"

{qf

(1

8)

(1

By taking

8)

[C]

3) (3

[W
x

(8

J5l

(8

\\

h {qf

[B]

(8

3) (3

{q}
(8

8)

1)

{X} dxdy

2) (2

1)

{f } dS.

[Nf
x

dxdy

[B]

2) (2

(13-22)

1)

partial derivatives of 11^ with respect to

u t9 v lt

etc.,

and equating

to zero,

6U,

(13-23)

0.
d{<\]

which leads to the eight element equilibrium equations as

M
(8

(q)
8) (8

1)

(Q)
x 1)

(8

{Qil
x 1)

(8

{Q 2 }
(8

1)

(13-24)

Two- Dimensional Stress- Deformation Analysis

342

where

[k] is

Chapter 13

the element stiffness matrix,


[k]

= h ft [Bf[C][B]dxdy,

(13-25)

and {Q}

is

the element nodal load vector,

{Q}

{Qi}

Evaluation of

[k]

{Q,}

JJ

\NY{X}dxdy

js

\NY{T}dS

and [Q

The coefficients of [k] are functions of local coordinates s and


more convenient to perform numerical integration as follows

M - h S M*, tdriQMst,
where

{st ,

t t)

The

is

first

t,)]\J(s l9 tt)

and

t,

(N

2 or four-point

= 4)

and so on.
integration

used.

part

{QJ

of the load vector can be computed as follows:

{Qi}=AStN(*0] r {X}V
If

we assume a uniform body

13-6(a)],

it is

(13-27)

t.

denote the local coordinates of the integration point

Often, for quadrilateral elements, 2


(Fig. 12-4)

(13-26)

X = 0,

force intensity

the expanded form of {Q,}

(per unit volume) [Fig.

is

0~

~N,

(13-28a)

N,

N<

N
N
N_

)\W

J(st9

N,

tt

(13-28b)

t.

i;

t
}

_0
The

subscript (s (i

instance, the fifth

t t)

(Sl,tt)

denotes that the matrix

component of
1(5)

evaluated at points

ti

)\J(s h

ti

)\W Y,

/,).

For

(13-28c)

y direction at node point


The second part {Q 2 } arises due to surface

gives nodal force in the

(s h

{Q,},

hT N (s h
l

is

1,

and so on.

on the
boundary of an element. Often it is possible to evaluate this_part by using
closed form integration. For instance, consider fxl and fyl as applied
tractions applied

Two- Dimensional Stress- Deformation Analysis

Chapter 13

343

(a)

tttttttttt

(b)

Figure 13-6 Loading on quadrilateral element,

tractions

(b)

Surface tractions.

on

side 1-2 [Fig. 13-6(b)].

(a)

Body

force,

Then

0"

r^

N
N

{Q 2 }

Txi

N<
h

dS.

(_'

Nt

N
N
_0

N,_

(13-29a)

Two- Dimensional Stress- Deformation Analysis

344

Now, along

side

N = i(l
A^ = i(l
x

A^ 3
A^4

= i(l
= iO

to

and

- s)(l
+ s)(l
+ *X1
- *X1

= 1

-t)= i(l
-t)= j(l

Chapter 13

therefore,

- 5),
+ ^),
(13-29b)

+0=0,
+0=0,

substitution of which leads to a line integral as


(1
(1

{Q 2 }

We

hU

"

s)/2
s)j2

dS.

(1

-s)j2

(i

(13-29c)

y\

used here the transformation relation in Eq. (12-38).

integrations,

Upon

required

we have
IT
T.

{Q 2 }
1

where

lx

= length of side

1-2.

(13-29d)

y\

This implies that the applied load

equally at the two nodes pertaining to the side 1-2. This


the fact that the interpolation function [Eq. (13-15)]

of the quadrilateral. If
results

may

we

is

is

is

distributed

a consequence of

linear along the sides

use a different (higher-order) approximation, the

not be similar. Furthermore, in the case of the higher order

it may be easier to perform numerical integration, as was


done for {Q,}.
The assembly of element equations can be achieved by following the

approximation,

principle that the displacements at the

procedure

is

essentially the

same

common

nodes are compatible. The

as illustrated in Chapters 3, 4, 5, 7, 11,

and

12 and as illustrated subsequently in Examples 13-1 and 13-2.

The assemblage equations

are modified for the

boundary conditions

in

terms of prescribed values of u and v on part(s) of the boundary. Solution of

Two- Dimensional

Chapter 13

Stress- Deformation Analysis

Then

the resulting equations gives nodal displacements.

are

computed by using Eqs.

(13-20);

and

(13-3), (13-5),

345

strains

and

and

stresses

(13-7), respec-

tively.

Triangular Element

After the derivations in Chapter

1 1

for the torsion problem,

it is

relatively

straightforward to derive element equations for the plane stress idealization

using the triangular element (Fig. 11-2). Use of the linear function, Eq.
(ll-4a) for both u

and v

various requirements (for two-dimensional

satisfies

plane deformations).

Various terms for the triangular element are stated below:


[u

[N]

~N

(13-30a)

v],

_0

W=

"2

["l

b2

~b t

a2

#!

V2

(13-30b)
3

V 3 ],

(13-30c)

0"

b3

M=A
where the

N,

a3

fl t

fl 2

a3

b2

b3_

are defined in Eq. (11 -3b)

(13-30d)

and the b and a are given


t

in

Eq.

(ll-3c).

The general form of

the element equations

is

identical to that for the

quadrilateral element [Eq. (13-24)]; only the orders of various matrices are
different.

The

stiffness

matrix has the order of 6

[k]

and

is

given by

= h[BY[C][B] j\ dA
hA[BY[C][B].

The load vector has the order

6x1.

(13-31)

For a uniform body weight

Y (X

= 0)

[Fig. 13-7(a)],

[Q
The uniform

T
l

= ^[0

surface traction

is

divided equally

belonging to the side on which traction


acting

on

side

[Fig. 13-7(b)]

{Q 2 y

is

(13-32a)

7].

among

applied.

For

the two nodes

instance, for

Txl

we have

^ZkZi[0

110

0],

(13-32b)

Two- Dimensional

346

Chapter 13

Stress- Deformation Analysis

(b)

Figure 13-7 Loading on triangular element, (a)

Body

force, (b) Sur-

face traction.

where /, is the length of side 1. Contributions of tractions


and on other sides can be similarly evaluated.
Example

in the

directions

Details of Quadrilateral Element

13-1.

Figure 13-8 shows a square isoparametric element (see also Example 12-1, Fig. 12-4)
subjected to a surface traction
are

E=

10,000 kg/cm 2 and v

First

we show
we have

(12-20c),

fx

equal to

brief details for the

jc

13

*24

X34
12

*23

x 14

kg/cm on

side 2-4. Material properties

computation of matrix

=o - = -1,
= 1 -0-1,
= l -0-1,
= 0-1 - -1,
=
=

= 0.30.

= 0,
= 0,

[B].

= -1,
7,3=0-1 = -1,
- = 0,
y 12 =
y = - = 0,
7 14 =0-1 = -1,
^ 2 3=0-l = -1.
y 1Af

Referring to Eq.

'

Two- Dimensional Stress- Deformation Analysis

Chapter 13

347

y
i

(0.1)

"

kg/cm

(-0.577,

(0.577,

0.577)

0.577)

(-0.577,

(0.577,

-0.577)

-0.577)

(0,0)

>

(1,0)

Integration points

Figure 13-8 Integration over square element.

Then
|/|

= i[(-D(-D +

/[(0)(-i)

= id +
= 0.25,
which

is

s[\

(-1X0)]

(0)(-i)]

1)

one-fourth of the area of the square. Therefore, use of Eq. (12-24c) gives

T_! _

B
8

Bi 2

xs-{-\)t]=\{-\ +

x 0.25

= y[ + l +

B 13 =[\

+0XS +

B 14 = y[~l +1 X

B 22 \[-\
x
~2~L

(-1)/]

(1)/]

= y(l -

= y(l +
X

t)

= y(-l + S\

=
-lxs-0xt]
\{-\
~
2
x

x/

xn

,j

'}

B 2 * = y[l + (- -

/]

= y(l -

[1

"

/),

/),

*23

Now

(-D(-l) +

(_1) *

the product [B] r [q[B] in [k] [Eq. (13-27)]

is

(1

+ S)

>

s).

given by

s),

/),

**

**

Tt^t

cq

T*

'"t

cq

cq

co"

cq

cq

oq

cq

cq

cq

cq

cq

cq

on o
n
u+u u+u u+u u+u

<n

o +u

oq

cq

<J

<S

cq

c*

u+u

u+<o

cq

c*

oqoq

oq

cqoq

cqoq

o
o

r*

f>

U+O U+U

Oq

~
N
N

oqoq
N
<S

oqcq

on
M

cq

cq

oq

cq

^^h
Mr)

oqoq

oqoq

oqoq

p^

r<

r*
*

oqcq
o

N
~

cq

cq

<>

r*

~
N
n

N
N

^
N

cqoq
cqoq
o

oq

tori

u +

cq

cq

cq

cq

oq

cq

cq

oq

oq

cq

cq

cq

N:

.,

_1_ r

^_
oq

***

t)

""I

Cj

^Lj -}"^

u+

cq

"

cq

cq

oq
^<

cq

CQ

g
CO

cq

cq*

fin

tt\
<*>

^*
m

r>

cq

<s

CJ
-"

cq

oq^^

u +G

flq

cq

oq

cq

~*

**>

oq

fvrl

<n

oq

cq

on ofnV^'NJ
u+u

cq

nr
.

oq

<n

oqoq

cq

oq^^

oqoq

cqoq

U+<J ^+<j

oqcq
_

oqoq

cqoq

cq

oq

cq~

<s

rri

oqcq

oq

com

<s

oq

-*

oq*

J + <J O+U O+U O+O U+<J


cqoq

-i

oq

oqoq

cq

cqoq

cq

t>

<S-h

N
cq

oqoq

+U

r*

oq

348

**

*-*

cq

_L

Two- Dimensional Stress-Deformation Analysis

Chapter 13

Here we have arranged the vector

[qf

As

{q} as

u2

vl

[u l

u3

v2

we need

indicated in Eq. (13-27),

349

to

w4

v3

v 4 ].

compute the terms one by one

at

each

of the four integration points (Fig. 13-8) and add them together to obtain the matrix
[k].

As an

Buisi,

Bn(s2
Bu(s 3

t2)

t3 )

The

"49.45

[k]

are obtained by using

tj

final result for the [k]

at the four points are

found as

= i(-l 0.577) = -0.789,


= ^(1 - 0.577) = 0.212,
= i(l + 0.577) = 0.789,
= i(-l*+ 0.577) = -0.212,

Bui**, U)

and so on. The values of the

tt )

tt)

Bn(s h

of

illustration, the values

matrix

E and

v.

is

17.86

-30.22

-13.74

-24.73

-17.86

5.49

1.38"

49.45

1.38

5.49

-17.86

-24.73

-1.38

-30.2

49.45

-17.86

5.49

-1.38

-24.73

17.86

49.45

1.38

-30.22

17.86

-24.73

49.45

17.86

-30.22

-1.38

49.45

1.38

5.49

49.45

-17.86

102

(13.34a)

sym.

49.45.

Use of Eq. (13-29d)


(Q}r

leads to the load vector {Q} as

0.50

0.0

[o.O

0.0

0.50

0.0

0.0

(13-34b)

0.0].

Introduction of the boundary conditions


ui
leads to

v\

v2

v3

W4

v4

two equations as
4945w 2

4-

549w 3

549!

4945w 3

=
=

0.50,
0.50.

Solution by Gaussian elimination yields

u2

Use of Equation

{<y}

u3

=
=

1.000

kg/cm 2

0.300,
0.000,

stresses

Ox

are

x 10" 4 cm.

[C][B]{q] leads to element stresses as

ax =

and the principal

0.91

1.000 kg/cm 2

0.300

computed from

^ i V(<7* - O Y +
y

(13-34c)

Xly.

Thus we obtain the quantities displacements, stresses, and strains required


and design of a structure idealized as a two-dimensional plane problem.

for analysis

Example

13-2.

Evaluate element

Triangular Element
stiffness

two triangular elements


stresses

matrices for the problem in Example 3-1 subdivided into


(Fig.

13-9).

Assemble and solve for displacement and

with the following data.

The surface traction fy on


Boundary conditions

(0, 0)

t<l,0)

11(0, 1)

v(l,

side 2-4 of element 2

i)

=
=

v(0, 0)

kg/cm per

unit thickness.

0.0,

=0.0,
v(0, 1)

=
=

(13-35)
0.0,
o.O.

Figure 13-9 Discretization with triangular element.

Thickness h =

cm

Element 2
1

Local node

(T) Global node

Degree
350

of freedom

Two- Dimensional Stress-Deformation Analysis

Chapter 13

We

351

1
(Fig. 13-9). Here we have numbered local degrees
and 6, corresponding to local displacements u u 2 u 2
Vi, v 2 and 3 respectively. The global degrees of freedom are numbered by assigning consecutively two indices to each node corresponding to displacement components u, v. respectively. Thus in Fig. 13-9 there are a total of 8 degrees of freedom.
The terms required for finding the matrix [B] are evaluated as follows:

first

consider element

of freedom as 1.2,
i-

3, 4, 5,

az
a2

=
=
=

.r

Xj =

- x3 =
- x, =

x,

xz

bt

=j

0.

b2

1.

bi

=y -y

1.

-->-3

= "I
-y = 1.

v3

=0,

and

2A

a2 b2

az b 3

>

cm :

Eq. (13-30d>.

"-I

0"

B =

-1
-1

For plane
[k]

we use Eq.

stress conditions,

-1

13-36;

13-3(b)J

and

13-31) to yield

MBFicm

---l

ooo-

-1

= A^

1
1

J
!

10

-1

13-3

"a.)

and hence
f

Element

Global

1
I

Local

-1

-d -

V)

-<1

- - V)

-(

- - V)

V
1

-v-j

-;
1

2.1

W=

13-3-r.

-(1

" - VI

svm.

Local

Element 2

4431obal

Two- Dimensional Stress- Deformation Analysis

352

The

matrix for element

stiffness

2,

Chapter 13

which has the same area and dimensions as

can be deduced from that of element 1 simply by properly exchanging


the node numbers. For instance, in Fig. 13-9 we have marked local numbers for

element

element

2,

and the corresponding global numbers are shown

at the

bottom of

Eq. (13-37b).

With each node having two degrees of freedom, we assign global and local
numbers to them. Thus the global numbers are (1, 2), (3, 4), (5, 6), and (7, 8) for the
four nodes, 1, 2, 3, and 4, respectively. The local numbers for the nodes are (1, 4),
(2, 5), and (3, 6) for the three local nodes 1, 2, and 3, respectively.
Since there is no load on any of the sides of element 1, there is no contribution
to the load vector. In the case of element 2, the surface traction of 1 kg/cm on side
2-4 yields

Element 1
Global

Element 2

Local
1

|h
1

[Q]

X
2

Element

assign

E=

Global

Local

we

(13-37c)
'

.0.

If

Global

10,000 and v

0.3,

Eq. (13-37b) reduces to

1
i

M = 10,000
1.82

"

1.35

-1.00

-0.35

-1.00

1.00

-0.35

0.65

-0.35

_-0.30

-0.30"

0.65

-0.35

0.00

-0.30

0.00

0.00

0.35

-0.35

0.35

0.00

-0.30

-0.35

1.35

-0.35

-1.00

0.00

0.35

-0.35

0.35

0.00

0.300

0.00

-1.00

0.00

1.00_

0.30
(13-38)

Local]

Element 2
Globall

Assembly of the
follows

stiffness

matrices and load vectors yield global relations as

oooooooo

~|
<*
:=

as

(S

sT

&

7
n

o m o
en
o en

o O o
7
+
o
en
d
in
m
en
en
en
m 8 o

o O o
+
+
1

en
o o

in
en

in
en

o O
in

>n
en

in
en
-"

^
3

5*

in
m
en
en
en

n
en
~

7
in
in
in
m
m

en
en
en
en O

en

in
in
in
n

en
en
rn
en

in
m
m

en
en
<n
7
d

n
n
o
8 8 en
O en en

o O O

in

as

n
en
"^

'o

,-H

in
n

en
m in en

en
en

m

en
en


m
+
en
in
en

in
en

in
en

o
o

in
en

in
8 en
7
in
o
en
o o

o
o o
+ 7

o o

en

in
in
in
m

en
en
8 en
8 en en
o
+

o
o

m
o
en
o

7
1

(A

in
en

n
en

DO
rN
"

en

in
>n
n

en
en
en

8

1

353

Two- Dimensional

354

Stress- Deformation Analysis

Chapter 13

Introduction of the four boundary conditions in Eq. (13-35)

is achieved by
rows and columns corresponding to u u v it v2 , u 3 , t> 3 , and v 4 that is,
global degrees of freedom 1, 2, 4, 5, 6, and 8, respectively. This leaves only two

deleting the

equations
1.35 2

-0.35 4

2^o =

0.000091

and

-0.35m 2

1.35w 4

0.000091.

Solution by Gaussian elimination gives

u2

These

u4

results agree with the solution

x 10" 4 cm.

0.91

by using the quadrilateral element.

Comment on Convergence
In the case of the displacement formulation, algebraic value of the
potential energy in the system

As

is

higher than the exact energy at equilibrium.

is improved by using higher-order approximation and/or by using refined mesh, the energy converges to the exact value.
Hence, an element which has lesser value of approximated stiffness can be

the quality of the formulation

considered to be superior (see Fig. 3-15).


Often, the trace of the stiffness matrix

The

trace

is

defined as

2K

ih

that

trace for the quadrilateral element

is

used as a measure of the

stiffness.

sum of the diagonal elements. The


4945 x 8 = 39,560, whereas the trace
the

is,

is

= 59,344. This shows that use of the


which contains an extra term (xy) in the approximation
model can yield better solutions. Such an inprovement may not be evident
with the use of only one or two elements. However, for larger problems the
improvement can be significant.
For general mathematical analysis we need to use the concept of the eigen
system of the matrix. This topic is discussed in advanced texts.

for the triangular element

is

7418 x 8

quadrilateral element

COMPUTER CODE
The computer code PLANE-2DFE (Appendix

4)

permits linear elastic

analysis of bodies idealized as plane stress, plane strain, or axisymmetric.

The

code incorporates the four-node isoparametric element [Fig. 13-4]. The


quadrilateral can be degenerated to be a triangle by repeating the last node.
In the following are presented typical problems solved by using this code.
Example

13-3.

Analysis of Shear Wall

Figure 13-10(a) shows a shear wall, which

The

wall has an opening 2

m wide

may

constitute part of a building frame.


deep in the lower floor level. The prop-

Chapter 13

erties

Two- Dimensional Stress- Deformation Analysis

355

of the wall are

= 2.1 x 10 9 kg/m 2
v (columns, beams, and wall) = 0.3,
Thickness of columns and beams = 0.3 m,
Thickness of wall = 0.15 m,
(columns,

Loading

is

beams, and wall)

shown

as

in Fig. 13-10(a).

By using the code PLANE-2DFE or other available codes, analyze the loaddeformation behavior of the wall assuming plane stress conditions. Obtain two sets
of results with two values of E for the shear wall with lateral loads only.
1.

E^m =0.21 x

10 9

2.

wall =

x 10 9 kg/m 2

0.0021

kg/m 2

Plot the deflected shape of the wall


the

first

value of "wall

and the distribution of vertical stress a y for


x 10 9 kg/m 2 at selected sections. Compare the results,
expected from the conventional beam bending theory

0.21

qualitatively, with the results

Beam
10,000 kg
1

I
-44

!-!
1

-4

4--I-I

JBeam'

m-

20,000 kg

h
Column
Wall

Opening

Wall

n
Column

'///////////////////A

y////////////////////y

8m
(a)

Figure 13-10 Analysis of shear wall, (a) Elevation of shear wall.

Two- Dimensional Stress- Deformation Analysis

356

Chapter 13

y
,

10,000 kg

20,000 kg

A
A

Origin-^ W/a

A
w/AW/A

wA

w//,W/A

(b)

Figure 13-10 (Contd.) (b) Finite element mesh.

and

offer

comments concerning

the influence of the stiffness of the shear wall

on the

load-deformation behavior.
Figure 13-10(b) shows a

finite element mesh for the shear wall; relatively finer


used near the two horizontal beams. Figure 13-1 1(a) shows the computed
deflected shape of the left end vertical side of the structure for the two values of

mesh

is

stiffnesses for the shear wall.

Figure 13-1 1(b) shows the deflected shapes of the

entire shear wall for the

two conditions.

Figure 13-12 shows


A-A, B-B, and C-C [Fig.

13-10(a)].

vertical (bending) stress

along three horizontal sections

We have assigned the stresses at the center of the

element for sketching the variation shown in Fig. 13-12.

Comments. Conventionally, shear walls are often designed by assuming them


beams idealized as one-dimensional. Here the usual assumptions of beam
bending are considered to be valid (see Chapter 7). In view of the irregular geometry
and the large area of the structure, such an assumption may not be valid and can
give results that are different from reality.
to be thick

Two- Dimensional Stress-Deformation Analysis

Chapter 13

357

A two-dimensional finite element analysis permits inclusion of irregular geometry

and the multidimensional aspects of the problem.

displacements, stresses, and bending

moments

It

allows computations of

in all elements; thus

one can compute

concentrations of stresses at critical locations and perform improved analysis and


design as compared with the conventional procedures. For instance, Fig. 13-12
shows that the distributions of bending stress o y at various sections are not linear
as assumed in the conventional beam bending theory. The bending stresses show
irregular and nonlinear behavior depending on the geometry, material properties,
and loading conditions. Moreover, it is possible to perform parametric studies to
include the influence of the stiffness of the wall on the behavior of the frame consisting of beams and columns. With a finer mesh, it is possible to identify the zones of
stress concentrations, particularly

Example

13-4.

Analysis of

Figure 13-13 shows a concrete


the bottom layer

is

near the corners.

Dam

on Layered Foundation

dam resting on

a foundation consisting of two layers;

underlain by a rock mass extending to considerable depth. There

exists a crack at the crest of the gallery.

By using PLANE-2DFE and assuming plane

-n

Scale:

Horizontal:
'

'

cm-

3
1.5 x 10"

Vertical:

'

cm

= 0.788

/
/

E^-O^xlOHg/m 2

E^n-0.21 x10 7 kg/m

(a)

Figure 13-11

shapes for

Computed

deflections for shear wall, (a) Deflected

left vertical side.

Two- Dimensional

358

Scale:

Horizontal:

Vertical:

Stress- Deformation Analysis

cm = 7.5 x 10
cm = 0.788 m

E^ = 0.21
E vwaii =

Chapter 13

x10 9 kg/m 2

- 21 *

10? k g/ m2

r
I

i
I

-T

///

//
'/

/
'

i/i/

//

i
(b)

Figure 13-11 (Contd.) (b) Deflected shapes of shear wall.

strain conditions, solve for displacements

dam and

foundation.

Concrete

in

The

stresses

under the self-weight of the

dam:

E=
v =
y =
Layer

and

properties are given as follows:

432 x 10 6 psf

(=

2.1

x 10 9 kg/m 2 ),

0.3,

150

pcf(= 2400 kg/m 3 ).

1:

E = 144 x 10 4 psf (= 7.0 x


v = 0.4,
y = 100pcf(= 1600 kg/m
3

).

10 6

kg/m 2 ),

T*o- Dimensional Stress- Deformation Analysis

Chapter 13

Scale

cm

for a

= 1.57

104 kg

359

+ Tension
- Compression

A 0'
B 0-

C 0-

Figure 13-12

Computed

[Fig. 13-l(Xa)] for

Layer

distributions

w ,u =

0.21

x 10

of c >

kgm 2

at

A- A, B-B,

C-C

2.

E=

144 x 10 5 psf

(=

7.00 x 10 8 kg

=0.3,

= HOpcf ('= 1760kgm

2
),

j.

Partial Results

Figure 13-14 shows a finite element discretization for the dam. In this figure, the
nodes and elements are numbered in the x direction. It may be computationally
economical to number them in the y direction, because according to Eq. ( 13-39)
below, this numbering will reduce the bandwidth B. At the same time, it will need
a greater number of data cards. The mode of numbering will depend on the given
problem and capabilities of the code and the computer. In general, however, selection of a

numbering system

that minimizes

B may

be a better strategy.

360

Www

iZ

994

<D

a 'o

361

Two- Dimensional

362

Stress- Deformation Analysis

Chapter 13

Discretization of "infinite" masses such as the foundation of the


special treatment. Since only a "finite"

zone can be included

dam

requires

element
mesh, we must include adequate extents in the mesh in the vertical and lateral directions so that approximate boundary conditions can be defined. For instance, if in
in the finite

mesh we include sufficiently large distances in the two directions, it can be


assumed that the displacements at such large distances are negligible. Thus in Fig.
13-14 we have included a lateral distance of about twice the width of the dam from
the edge of the dam and have assumed that the horizontal displacement u at that
distance is approximately zero. For the problem in Fig. 13-13, a rigid rock boundary
is available at a distance of 50 ft (15.25 m) below the base of the dam; hence, we have
assumed both displacements to be zero. If such a rigid boundary is not available,
we can go to a sufficiently large depth, say about three to five times the width of the
dam, and assume either v = and u free or both u = v = at such a discretized
boundary. For new problems, the analyst may need to perform a parametric study
the

to decide the extents of discretized boundaries.

The crack at the crest of the gallery is introduced by providing node numbers on
both sides of the crack. In Fig. 13-15 is shown the distribution of vertical displacements along the ground

level for gravity loads

and for both gravity and hydrostatic

loads.

Detailed analyses of stress concentrations around the crack and of such other
aspects as the effect of hydrostatic forces are

Comments.

on

On

left

to the reader.

the basis of the results in Figs. 13-14 and 13-15 and the results

stress distributions

which can be plotted from the same computer

results,

possible to perform (preliminary) design analysis for the behavior of

it is

dams on

multilayered foundation systems.

The computed concentrations of stresses near


(shear) stress in the concrete

whether the compressive


crete. Similar
soils.

check

show us if the tensile


we can check

is

within allowable compressive strength of the con-

conclusions can be drawn regarding the stresses in the foundation

Moreover,
if

stress

the crack can

within allowable limits. Similarly,

is

if

the design limits allowable deformations,

it

may

be possible to

the loading causes deformations within allowable limits.

In this example, we have considered essentially only loading caused by gravity


and have assumed linear material behavior. Hence, the results can be treated as
preliminary. For a final design one needs to consider other loadings such as hydrostatic load due to water in the reservior, earthquake loading, uplift due to seepage
through foundation, and loading caused by variations in temperature difference
between the dam and the surrounding atmosphere. Nonlinear behavior of the soil
and rock foundations will influence the behavior of the dam. Finite element procedures that can allow inclusion of these factors are available and will be stated in
advanced study of the method and its applications, Chapter 15.
For the present, we note that it is possible to perform a detailed stress-deformation analysis of dams on nonlinear foundations; the latter can include both soils
and jointed rock masses. In addition to computations of stresses and deformations
for various loading conditions, the

metric studies. For instance,


tion

it is

method can

possible to find

also be used for performing para-

optimum locations for instrumenta-

and for (underground) structures such as caverns and

galleries.

3
Gfl

-a

c
a

363

Example

As discussed
results that

Beam Bending

13-5.

in the earlier chapters, a numerical

procedure should generally yield

converge or approach the exact solution as the mesh

is

refined con-

To study convergence for the two-dimensional problems, let us consider


a (deep) beam (Fig. 13-16) subjected to uniformly distributed load of 1000 kg/m.
The beam is divided into three progressively refined meshes as shown in Fig. 13-16.
sistently [3].

Note

mesh

that a refined

includes the previous coarse mesh; this

is

required for

mathematical convergence. This and other requirements for mesh refinement are
discussed in Ref.

[3].

Table 13-1 shows computed values of vertical and horizontal displacements at

The number of nodes N,

points A, B, and C(Fig. 13-16).

TABLE

the

maximum

difference

Results for the Uniform Loading

13-1

Displacement x 10" 3 (m)


Point

N D B

==

(/>

10

25

14

81

10

22

1)/

NB 2

Point

Point

//

C
V

900 0.0478 -0.1948 -0.0448 -0.1924 -0.00045 -0.0766


-0.0645 -0.2796 -0.00073 -0.1119
0.00079 -0.1278
39,204 0.0761 -0.3198 -0.0726 -0.3173
4,900 0.0678 -0.2821

between any two node numbers D, the semibandwidth B, and the quantity NB 2
shown in the table. The semibandwidth B is computed from the formula

are also

# = (/> +

l)/

(13-39)

where /is the number of degrees of freedom at a node; here/= 2.


Figure 13-1 7(a) shows convergence behavior of computed displacements u, v
at points A, B, and C with the number of nodes, and Fig. 13-17(b) shows their relationship with NB 2 In both cases the computed displacements tend to converge as
N and NB 2 increase.
.

The quantity

NB 2 is proportional

to the time required for solution of the global

equations, which constitutes a major portion of the time for the finite element solu-

Figure 13-1 7(b) shows that refinement of mesh results in a faster increase

tion.

NB 2

in

computational effort or cost. In other words, for a given desired accuracy, if the number of nodes are increased, the quantity NB 2 increases at a higher
rate than the gain in accuracy. Thus there exists a trade-off between accuracy and
,

that

computer

is,

effort or cost of a finite

Figures 13-1 8(a),


three meshes.

(b),

and

(c)

element solution

show

[4].

distributions of the bending stress

We can see that the distribution of bending stress

o x for

the

improves with mesh

refinement.

Comments. The

results of this

example

a finite element solution by progressive


refined, the

number of elements

for data preparation

As noted
364

illustrate that, in general,

mesh

we can improve

refinement. Obviously, as the rnesh

is

increase with corresponding increases in the effort

and computer

time.

in previous chapters, a finite

element solution can also be improved by

1000 kg/m

II!..,
A
i

E =
'c

2x10 9 kg/m 2

t>=0.3
Thickness = 0.2

-*

(a)

C
i

(b)

64A

'c

(c)

Figure 13-16

mesh,

(c)

Beam

bending,

(a)

Four-element mesh,

(b) 16-element

81 -element mesh.

365

CO

O
I

v,

A denotes

displacement at point

and so on

0.40
v,

0.30

0.20

0.10

30

40

Number

50

60

of nodes,

80

90

00

(a)

0.30

0.20
v,

u,

/.
0.10

jL
\u, B

/U, C

/
10,000

20,000

30,000

40,000

NB 2
(b)

Figure 13-17 Behavior of numerical solutions for displacements,


(a)

Displacement

vs.

number of nodes:

13-16). (b) Displacements vs.

366

NB 2

points

points A, B,

A,B, and C (Fig.


and C(Fig. 13-16).

'

15,810 kg cnv

*^H

2,670 kg cm?
(a)

45,000 kg cm 2

cm

1,660 kg

a
:

j-

23,220 kg cm?
(b)

^^

/24,000

70 0O0

*^-<

ss *
R

^r

'

>

'<,
I

tf
>

J
1

j|
423
[c

Figure 13-18 Distributions of bending

stress,

ax

at typical sections.

367

Two- Dimensional Stress- Deformation Analysis

368

Chapter 13

using higher-order elements. For instance, for the plane problems,

we can use

quadratic or second-order approximation for displacement within the element. This

nodes

will require additional

and

for the triangle there will be six nodes,

for the

quadrilateral there will be eight nodes. There will be a corresponding increase in the

of the element matrices and the number of equations to be solved.


Which of the two approaches, refinement of mesh with a lower-order approximation or higher-order approximation, should be used will depend on the type of
size

problem. There are trade-offs

wide

The

in

one approach over the other. This subject

in using

is

scope and usually needs parametric studies for a specific problem on hand.

analyst

may need

to use both approaches

and then derive

criteria for their

use

for the given problem.

PROBLEMS
13-1.

Compute

the

13-2.

Compute

[A]

load vector

initial

triangular element
-1

in

number

{Q

for a given initial strain {e

for the

in Fig. 13-9.

Eq. (13- 14c) and then find the interpolation functions

in matrix [N] [Eq. (13-15)].

the details of the derivation of x y and y xy in Eq. (13-18).

13-3.

Fill in

13-4.

Obtain the load vector {Q 2 } for a traction load

fy

applied on side 4-1

[Fig. 13-6(b)].

13-5.

Evaluate the

stiffness

matrix for the triangular element in Example 13-2 by

rearranging the load vector as


[q}

13-6.

Find the element

[u l

u2

vt

v2

u3

v 3 ].

matrix for the triangular element in Fig. 13-9 for

stiffness

plane strain conditions. Hint: Use [C] from Eq. (13-5).


13-7.

Evaluate the element

stiffness

matrix for the triangular element in Fig. 13-9

for axisymmetric idealization.


13-8.

By using PLANE-2DFE or another


problem

(Fig. 13-10) with

available program, solve the shear wall

an additional

and compare with those without the


13-9.

kg at the
and stresses

vertical load of 10,000

centerline of the wall. Plot the results in terms of displacements


vertical load.

In Example 13-4, consider hydrostatic force caused by a water level of

m) above

ground surface,

due to the
and obtain distributions of displacement along the ground surface and plot a x a y and r xy within the dam.
Compare results with those from the analyses without the hydrostatic force.
Figure 13-15 shows the deflected shape of the ground surface with hydro-

40

ft

(12

the

weight of the dam. Use

in addition to the load

PLANE-2DFE

static loading.

13-10.

Vary the

relative values of

E in the two foundation

layers in Fig. 13-13

and

study their influence on displacements and stresses.


13-11.

For the beam in Fig. 13-16, consider a concentrated vertical load of 5000 kg
at point A, and study convergence of displacements vs. TV and NB 2

13-12.

Figure 13-19 show a

finite

element mesh for the

beam

in

Example

3-5,

with

<3

<8

E
cr>

cn

II

II

MOO
II

LU

-C

41

<$

^3
369

Two- Dimensional

370

Stress- Deformation Analysis

Chapter 13

a hole around the centerline. For the concentrated load shown, obtain the
finite

element solution and study the concentration of stresses around (the

corners of) the hole.


13-13.

Figure 13-20 shows a


is

medium

subjected to external loading.

The medium

of "infinite" extent. Assuming discretized boundaries (u

and

distances of four times the width of the loading in the lateral

0)

at

vertical

on a suitable mesh. Choose your own value of B, material


medium, and the loading. By using PLANE-2DFE or
another code, evaluate stresses and deformations in the medium assuming
(a) concentrated load at the center point with plane strain and axisymmetric
approximations, (b) strip load with plane strain approximation, and (c) cir-

directions, decide

properties of the

cular load with axisymmetric approximation.

Change

the boundaries to

6B, 8B, and 10B in both directions and examine the displacements near the
discretized

end boundaries and near the loading.


^Concentrated load

Q_

=4,6,8,10,

Strip or circular load.

_:

T3.

Discretize

X 2 =4,6,8, 10,

Discretized end boundaries

Figure 13-20 Load-deformation analysis for "infinite" medium.

13-14.

Figure 13-21 shows an underground tunnel (3-m diameter) with a structural


lining of thickness equal to 30

cm. The tunnel

is

excavated in a (homo-

geneous) rock mass of large lateral and vertical extent.

It is

required to

compute the changes in the states of stress and deformations due to the
tunnel excavation. The material properties are shown in the figure. By using
the mesh shown in the figure (or any other suitable mesh) and a computer
code, compute stresses and deformations (a) before the tunnel is excavated,
that is, under the gravity loading. Here use the entire mesh including that
in the zone of the tunnel and lining, (b) Consider the elements in the tunnel
zone to be removed, and include the elements for the lining. It may be
necessary to renumber the nodes and elements. Compare the two analyses
and draw conclusions regarding the influence of tunnel excavation on the
states of stresses and deformations, particularly in the vicinity of the tunnel.

Chapter 13

Two- Dimensional Stress- Deformation Analysis

20

///&/>

371

//W/AZ

Rock
E =

6x10 8

kg/m 2

^=0.3

u =

3
7 = 2400 kg/m

v free

Lining

E = 2x 10 9 kg/m 2
v = 0.25

u =
v free

V77777777777777777777777777777777777777777777777,

U = O, V = o

Figure 13-21 Analysis of underground tunnel.

REFERENCES
[1]

Popov, E.

P., Introduction to

Cliffs, N.J.,

Mechanics of

and Goodier,

Timoshenko,
York, 1951.

[3]

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Abel,

J.

F.,

S.,

and Desai, C.

Bending," Proc.
2148.

Englewood

N., Theory of Elasticity, McGraw-Hill,

[2]

[4]

Solids, Prentice-Hall,

1968.

ASCE,

S.,

J.

to the Finite

New

Element Method, Van

"Comparison of Finite Elements for Plate


Vol. 98, No. ST9, Sept. 1972, pp. 2143-

J. Struct. Div.,

MULTICOMPONENT SYSTEMS
BUILDING FRAME
AND FOUNDATION

INTRODUCTION
Very often, in practice, configuration of a system or structure is such that its
approximate simulation may require the use of or it may be beneficial to use
more than one type of idealization. For instance, if we need to idealize a
three-dimensional building frame and its foundation (Fig. 14-1), it is convenient and economical to treat the building frame as composed of onedimensional beams and columns and two-dimensional slabs and plates. As
a rather crude approximation, the foundation can be included by assuming
the structure resting on a bed of (individual) springs representing the foundation (Fig. 14-1); this approach is often referred to as a Winkler foundation.
Thus, the system contains three components beam-columns, slabs or plates,
and foundation.
Many other situations in stress-deformation analyses and field problems
require such multicomponent idealizations. We shall illustrate here the
problem of a building frame (Fig. 14-1). For simplicity we shall consider an
orthogonal frame with only horizontal and vertical members. Extension to
the general case of inclined members can be achieved by appropriate trans:

formations

372

[1-3].

Chapter 14

Building

Frame and Foundation

373

zation of supports

Idealized

%>

column

Idealized fou ndation

Figure 14-1 Building frame: multicomponent system.

VARIOUS COMPONENTS
As

stated earlier, we need to consider structural components that act as


beam-columns, slabs and plates, and foundation (springs).

Beam-Column
In Chapter 7 we have already derived the element equations for a beamcolumn. However, now we need to consider the possibility of loads causing

bending in both x and y directions. Figure 14-2 shows the beam-column


idealized as one-dimensional.

The degrees of freedom for the element consist of axial displacement w,


displacements u and v, and rotations 9 X yi and 6 Z about the three

lateral

Here the subscript denotes derivative; for instance, 6 X


= dw/dx. Hence, the beam-column element has 6 degrees of freedom at each
node and a total of 12 degrees of freedom. For the time being, if we do not
consider rotation or twist around the z axis, we have a total of 10 degrees of

axes, respectively.

374

Building

Frame and Foundation

Chapter 14

w2

A,E,I X ,T,I 2

/I

Bending

Axial load

Bending

in

x-direction

in

y-direction

Figure 14-2 General beam-column element.

freedom. The nodal displacement vector

W=
where 0.^

As

in

[i

0*1

= (dw/dx)
Chapter

7,

lf

"2
yl

0*2

0,i

^2

0,2

w,
i

w 2 ],

(14-1)

= (dw/dy) u and so on.

we assume

the following approximation functions

= a, + a z + a
v(z) = a + a z + a
w(z) =a + a
u(z)

then given by

is

10 z.

z2

7z

+a
+a

4z

(14-2a)

(14-2b)
(14-2c)

Chapter 14

Here

u(z)

Frame and Foundation

Building

and

v(z)

375

correspond to bending displacements in the x and y


and w(z) denotes shortening or extension due to axial

directions, respectively,

(end) loads.

As

discussed in Chapter

Eq. (14-2) can be transformed to

7,

express w(z), v(z), and w(z) in terms of interpolation functions and their nodal
values as [Eq. (7-37)]

= Nxl u + Nx2
= N v + Ny2

u(z)
v(z)

yl

+N w
2

yl

w(z)=N w

+ Nx3 u + Nx4 x2
+ N v + Ny4 6 y2

xl

y3

(14-3a)

(14-3b)

(14-3c)

where

Nxl

+ 2s
ls(\ -2s + s
5 (3 - 25),
Is (s - 1),
1

>!

yl

Nx3
Nx = N,

3* 2

3
,

2
),

(14-3d)

>'3

and
AT,

N =
9

where

is

By

(14-3e)
s,

the local coordinate given by


S

where f

-s,

=z

and z

=-r

(14-4)

the coordinate of node

is

1.

following the procedure outlined in Chapter

7,

we can

derive the

following element equations

mii)

'{Q]

(14-5a)

{Q}
{Q.3

where
a,[kj

[0]

[0]

[0]

ajkj

[0]

M
.

[0]

12

[kj

[kj

[0]

6/

4/

(14-5b)

fcJJ

-12

6/"

-6/

2/ 2

12

-6/

(14-5c)

sym.

4/2

(14-5d)

[kw]
/

-1

(14-5e)

376

Building

Frame and Foundation

{QF = A |W{X}</z
Jo

and the subscripts a and b denote

Chapter 14

fW{f}</z,

(14-6)

Jo

axial

and bending modes,

respectively.

Plate or Slab

The general loading conditions on the frame can cause plane deformaand twisting; for convenience we consider only the first two.
Figure 14-3 shows the two effects, which can be superimposed if we assume

tions, bending,

small strains and deformations.

Membrane

Effects

For the in-plane or membrane loading we can assume the plane stress
and use the element equations derived in Chapter 13, Eqs.
(13-24) and (13-31) for the quadrilateral and triangular elements, respectively.
idealization

Figure 14-3 Plate

Membrane
patibility

elements

and interelement compatibility,

or in-plane behavior, (b) Bending behavior,

between plate and beam-column.

L
(b)

(a)

e-

(c)

(a)

Com-

Bending

For isotropic

flat plates,

Dn /(9

vv*

{-d^

D=

where
v

is

Eh

/I2(l

the Poisson's ratio,

the governing differential equation


4

w*

2d w*
d w*\
d^W + -d^) =p

+
,

),

is

written as:
, tA

E is

is

(14_7)

the elastic modulus, h

the plate thickness,

is

the transverse displacement

and p

is

the applied

surface traction.

Our aim

herein

is

to

show

essentially the treatment of a

multi-component

system for a building frame in which plates or slabs occur as one of the main

components.
problem.

bending

It is

not our intention to go into

It will suffice

is

essentially similar to the other

considered. Here

we

much detail

of the plate bending

to say that the finite element procedure for plate

problems that we have previously

shall describe only the salient features of the

problem.

For convenience, we consider only orthogonal frames; hence, we have


only orthogonal (horizontal or vertical) plates. For a general formulation,
plates with other configurations can be handled without

much

difficulty.

For the rectangular element in Fig. 14-3(b) we consider three degrees of


freedom at each node, namely the transverse displacement w; the rotation
about y axis, 9 X = dw/dx, and the rotation about x axis, 6 y = dw/dy. Together with the two in-plane displacements u and v [Fig. 14-3(a)], there are a
total of five degrees of freedom at each node.
As shown in Fig. 14-3(c), we have an equal number of degrees of freedom
for the beam-column and the plate element. Consequently, we can observe
the requirement of interelement compatibility between adjacent plate and
beam-column elements.
The transverse displacement w can be approximated by using hermite
interpolation functions as in the case of the beam bending problem [Chapter
7 and Eq. (14-3d)]. In fact, the interpolation functions for the two-dimensional problem can be generated by proper multiplication of the hermitian
functions defined for the x and y directions. For example, we can adopt an
approximation for

as

[3, 5]

= N^N^w, + Nx2 N 6 xl + Nxl Ny2


+ Nx3 Nyt w + Nx ,N x2 + Nx3 Ny2 6 y2
+ Nx3 Ny3 w + Nx4 Ny3 x3 + Nx3 NyA 6 y3
+ Nxl Ny3 w + Nx2 Ny3 9 xA + Nxl N ,6
= [N, N N Nt .-. N l2 ]{q
(14-8a)
= [NJfo},
the matrix of interpolation functions, N = Nxl NyU N =

w(x,y)

yl

yl

yl

where [NJ

Nx3 N

yl

is

etc.,

y,

b]

where
377

378

Building

Frame and Foundation

Nxl =l- 3s + ls\.


Nx2 = as(s - l)
Nx3 = s (3 - 25),
tf, = as (s -

+ 2V
l)
btit
Ny3 = (3 20,
W = bt\t - 1).

{a

and

1,

1),

{q^F
is

= j>/6, < <


r

[Wl

0*1

1,

y4

[Fig. 14-3(b)], s

= x/a,

<s

and

^2

^yl

(14-8b)

of the element

6) denotes the size


t

3t

<

Here

N.

Chapter 14

0*2

H> 3

0,2

0*3

0,3

0*4

yi]

the vector of nodal unknowns.

The

N N

interpolation functions

N = Nxl N = (1 -

the definition of inter-

etc., satisfy

3s 2

yl

l9

we have

polation functions. For instance,

25 3 )(1

2t

=
= 0,
= 0,
= 0,
=h
=h
- 3t +

2/

3r

(14-9a)

for

5=0, '=0,
5

=
=

=0,1

_
^, _

1,1

1,1

#,

f=0,
f

AT,

1,

#,

l,

tf,

,>

AT,

s=i,i'=0,

Now N = Nx2 Nyl =

and so on.
tive

of

with respect to x

^2 =
ox

Then

a (3S

tf,

1) (1

45

1)(1

-3/ 2

for

N2

the case of

2 ldx

1,/

=0,

0.0

0.0

l,f

1,

0.0

0.0

=0,

1,

0.0

0.0

= i, = 0,
t

9a/64

16

= , = 0,
/

Sa/64

= |, r = o,

3a/64

2r

Plots of #,,

= 0,

s == o,

= o, = ,
^, %,
and

0.0

1.0

0.0

4a/64

(?A^2

).

The

first

deriva-

gr /en by

is

05(5

1,

/^

are

shown

(14-9b)

).

(rad)

5_

16

i
-iin Figs. 14-4(a), (b),

and

(c).

In

the slope correspc nding to the degree of freedom 6 xl at node

Chapter 14

Building

Frame and Foundation

379

(a)

Figure 14-4 Plots

of

bending, (a) Ny. (b)

is

unity. It

interpolation

typical

N2

(c)

functions

for

plate

dN2 jdx.

can thus be shown that

other functions in Eq. (14-8a) satisfy

all

the definition of interpolation functions.

The

strain (gradient)-displacement relation for plate


2

d w]

z<

d2w
dy 2

d2 w

w yy

<

z[Bb ]{ qi

The strain-displacement transformation matrix


required derivatives of w from Eq. (14-8).
The constitutive law is expressed through
moments and second derivatives or curvatures

Mxx

"1

is

the relation between the

"

Eh"
~ 12(1

-v

w
w

-v
2

[C]{]
is

obtained by finding the

yy

Here h

(14-10)

2w xy

dxdv>

[6]

w xx

dx 2
(]

bending is given by

the plate thickness.

w
J
(14-11)

380

Frame and Foundation

Building

The

Up

potential energy for the plate bending

=
"T

If

problem

Chapter 14

is

& T[C^ dxdy - h JJ {^YM dxdy - j

expressed as

(tfMd&

(14-12)

The components of

can be found by taking appropriate


and by using the transformations

{e}

derivatives of w [Eq. (14-8)]


-=-

dx

= ->-

and

= [BJ{q

Substitution of {e}

= --=-.

-=-

dy

a ds

(14-13)
v
7

b dt

and w = [NJ{q 6 } in IT^ and taking variacomponents of {q b } lead to its stationary

A}

tion of Il p with respect to the


(minimum) value as

= 0=>|&L =

(?n,

partial

(14-14a)

0,

which leads to the element equations as

KK*} =

(14-I4b)

{Q b },

where
\kb ]

{Q,}

\\

[B b Y[C][Bb ]dxdy

\\

[Wmdxdy

and

+
The

coefficients

[N] T {t}ds.

of [k b ] can be found in closed form by integration. Their

values are tabulated in Ref.

[4].

The load vector {Q b } can

also be

found

in

closed form. For instance, for uniform transverse surface load/? on the plate,

fO
ivJ

[ P ab
|_

P a2 b

pab 1

pab

pa 2 b

24

24

24

pa b

pab

"24

24

= P^[6

pab

pa b

"24

4
a

-b

pab 2

pab

"24
pab

24 ]

-a

-b

-a

b].

(14-15)

Here we have considered only one of the many possible and available
fact, the function in Eq. (14-8) can be
improved by adding the additional degree of freedom Q xy = d 2 w/d x d y [4, 5].
approximations for plate bending. In

Assembly

The assembly of the element equations for beam-columns and plates can
be achieved by using the direct stiffness approach, which assures the interelement compatibility of nodal displacements and rotations.

Chapter 14

As

Building

381

numbers of degrees of freedom


and beam-columns, the assembly procedure at

stated earlier, since there are equal

for the junctions of plates

such junctions gives no

difficulty.

The element equations


the

Frame and Foundation

membrane behavior

for the

beam-column

are given in Eq. (14-5a). for

of the plate in Eq. (13-24). and for the bending

behavior of the plate in Eq. (14-14b).

The

assemblage equations can be expressed as

final

[K]{r}={R}.

(14-16)

These are modified for the boundary conditions and then solved for the nodal

unknowns. Then the

strains,

stresses,

bending moments, and shear forces

can be evaluated as secondary quantities.


Representation of Foundation

Figure 14-5 shows an approximate representation for a foundation by


using a series of (independent) springs. Structural supports provided by
adjacent structures (Fig. 14-1) can also be simulated by using this concept.
In the case of foundations, the constants for the springs. k f can be
tests on soil samples and or from field
,

evaluated on the basis of laboratory

experiments. They can be evaluated also from the concept of subgrade


reaction

[7].

Figure 14-5 Idealization

column,

of

foundation

by

sp:;-zs.

[a]

Beam-

(b) Plate or slab.

Since the springs are assumed to be independent, their stiffness coefficients

can be added directly to the diagonal coefficients of the global matrix [K]
[Eq. (14-16)]. For instance., if kft denotes the stiffness coefficient of spring
support at node i in the x direction, then it is added to the diagonal element.
Kjj, where j denotes the corresponding global degree of freedom. We note
that the spring supports can be specified in the direction of translations
(x, y, z)

and rotations

(9 X

6 y ).

COMPUTER CODE

A computer program STFN-FE that


foundations

is

allows analysis of building frames and

described in Appendix

Some examples

4.

solved by using this

code are described below.

Example

Plate with Fixed Edges and Central

14-1.

Load

Figure 14-6 shows a square plate 25.4 cm x 25.4 cm divided into four equal elements [3]. The thickness of the plate is 1.27 cm, with E = 2.1 x 10 5 kg/cm 2 and
v =0.3. The concentrated load P = 181.2 kg. According to the closed form
approach [6], the maximum central deflection is given by
0.0056 x

P x
D

(25.4) 2

(14-17)

where

D=

Eh3
12(1

~~
- v2 =

21 X

1Q5(L27)3

12(1

0.09)

3 94 x
3jm

10*
1U

Therefore
0.0056 x 181.2 x 25.4 2
3>94

The value of

central deflection

x 1Q4

,,
= nm
0.0166 cm.

computed from the

finite

element analysis

is

0.01568 cm.

25.40

cm

Figure 14-6 Square plate with central load.

Example

14-2.

Plate (Beam) with

Two

Two

Fixed Edges,

Free Edges, and Central Load

Figure 14-7(a) shows a plate fixed at two ends and free at the other two
plate can also be approximately treated as a beam.

0.453 kg

is

applied at the center point.

The value of

[3, 8].

This

concentrated load equal to

E=

7.0

x 10 5 kg/cm 2 was

assumed.
finite element mesh for the plate. In Fig. 14-8 are shown
computed values of transverse displacements at center section C-C of the plate

Figure 14-7(b) shows a


the

[Fig. 14-7(b)].

We can compute the displacement by using the results from strength of materials
[9] if

the plate

is

and (b) show the


moments and conjugate beam, respectively. Displacements
approach are shown in Fig. 14-8 in comparison with the finite

treated approximately as a beam. Figures 14-9(a)

distribution of bending

computed from

this

element solutions. For instance, the

382

maximum

deflection at the central section

is

Chapter 14

Building

Frame and Foundation

383

0.453 kg

cm

1.905

0.3175

jL

26.67

cm

cm

(a)

&

*A

(b)

Figure 14-7 Analysis for (thin)

beam bending

[8].

(a)

Beam

with

fixed ends loaded centrally, (b) Finite element mesh.

Finite element solution

Strength of materials solution

Figure 14-8 Comparisons for displacements

from

computed and closed

solutions.

PL
\92EI
~
3

W max
m ax

~~

453 x 26.67 3
192 x 7.0 x 10 5 x 0.00508

(14-18)

0.0126 cm,

where
l

Example

14-3.

1.905

12

x 0.3175 3

0.00508.

12

Frame with Lateral Wind Loads

one-dimensional idealization of a part of the building frame [Fig. 14- 10(a)] is


shown in Fig. 14-10(b) [10]. The loading and other properties are shown in Fig.
14-10(b).

384

Building

Frame and Foundation

Chapter 14

1.510 kg-cm

1.510

1.510

(a)

1.510

1.510
EI

(b)

Figure 14-9 Procedure for closed form solution for displacements,


(a)

Bending moment diagram,

(b)

Conjugate beam.

732 cm

732 cm

(a)

Figure 14-10 Details of building frame

[10]. (a)

Plane layout.

Figure 14-11 shows the computed deflections and the deflected shape of the
frame. Table 14-1 shows the computed values of bending
bers in comparison with those

moments

in various

mem-

from the conventional portal method of frame analy-

E = 2.1
I

Area

106 kg/cm 2

= 27346 cm 4
= 1544 cm 4

= 94.84

cm 2

1268

A
2536

kg.

kg-

732 cm

732 cm
77777

-777777,

777777,

(b)

Figure 14-10 (Contd.) (b) Single idealized frame.

Scale: Horizontal (displacement):

cm

= 0.0394

cm

(0.4298, 0.0022 cm)

~~7
/

/
/

/
/

-J

_J

/
/

/
/

/
/
/
1

/
/
/

Figure 14-11 Deflected shape of frame.

385

386

Building

Frame and Foundation

Chapter 14

based on the assumption of a point of inflection at the midsection of


can be seen that the values from the two methods differ widely at
various locations. The finite element approach shows the redistribution of moments
sis;

the latter

a member.

is

It

in the frame.

TABLE

14-1

Member
1

2
3

Comparison of Bending Moments


for the Frame [Fig. 14- 10(b)]

End

6
7
8

9
10

Example

14-4.

Method

x (kg-cm)

Finite Element

203,103

327,224

203,103

191,750

67,701

21,474

67,701

95,712

406,205

355,139

406,205

258,680
120,860

4
5

Portal

135,402

135,402

17,260

203,103

312,616

203,103

179,523

67,701

30,914

67,701

99,495

270,803

213,223

270,803

190,466

270,803

189,069

270,803

210,436

67,701

95,714

67,701

85,694

67,701

87,565

67,701

99,495

Building

Frame with Floor Slabs

Figure 14- 12(a) shows the layout of a building frame idealized by using a one-

dimensional beam-column, two-dimensional plate membrane, and bending elements; Fig. 14-1 2(b) shows the loading at the two floor levels. Two analyses were
performed one without floor plates and the other with floor plates. The properties
;

of the frame material are

E = 2.1 x 10* kg/cm 2


Ix = 27,346 cm 4
ly = 27,346 cm 4
A = 94.84 cm 2
v = 0.3.

Area of cross

section,

Table 14-2 shows a comparison between moments (Mx ) at typical nodes for the
It can be seen that without floor slabs,

analyses with and without floor slabs.

the middle section of the frame will have higher loads (moments) that those at the

Chapter 14

Building

TABLE

14-2

Frame and Foundation

Comparison of Moments

(Mx )

387
at Typical

Nodes
Plates

Node

Not Included
(kg-cm)

Plates Included

(kg-cm)

32,597

31,289

93,747

104,875

12

49,077

22,741

15

158,825

116,723

21

32,576

32,386

24

93,712

105,206

First

floor level

Figure 14-12 Analysis of "three-dimensional" building frame


frame.

[10]. (a)

Details of

Building

388

Frame and Foundation

5817 kg

634

Chapter 14

5817 kg

11,633 kg

23,266

11,633

11,633

1268

5817
634-

5817

11,633

j.

i
Second floor

level

10,274

20,548

1268-*-1First floor level


(b)

Figure 14-12 (Contd.)(b) Loadings at

first

and second

floor levels.

edges. Moreover, the displacements at the central section are significantly reduced
if

floor plates are included.

is

0.24393

cm and

Example

To examine

14-5.

For

instance, the

that without plates

is

y displacement

at

node 12 with

plates

0.40225 cm.

Effect of Springs at Supports

on the behavior of the frame shown in Fig.


k 6x (kg-cm/rad) is introduced at the three supports
(Fig. 14-13). The magnitudes of k 0x were varied as shown in Table 14-3. The results
indicate that a very low value of k 6x essentially models a simple support, whereas a
the effect of restraint

14-10(b), a rotational spring

Chapter 14

Building

Frame and Foundation

389

A
A
A

732 cm

732 cm

At, k.

Figure 14-13

Frame

with

rotational

restraint

represented

by

spring.

very high value produces results close to those with total restraint (see Table 14-1).

Near the value of k 6x between 10 8 and 10 7 the point of inflection can be near the
midsection, as assumed in many conventional analyses for frames.

TABLE
kex
(kg-cm/rad)
lOis

lOio
10 9
108

107

10 5
103

14-3

Moments

x (kg-cm)

End

Member

for Typical Members

Member

327,011

312,412

191,627

179,406
312,151

326,690

191,927

179,727

323,835

309,812

194,620

182,581

297,794

287,448

219,196

208,307

164,984

162,128

344,492

335,943

3,297

3,260

497,105

489,437

500,217

48,422

Member

-188,945
-210,299
-189,132
-210,524
-190,816
-212,537
-206,185
-230,761
-284,898
-322,438
-380,988
-433,436
-382,943
-435,691

TRANSFORMATION OF COORDINATES
As we

discussed earlier, it is convenient and useful to adopt a local coordinate


system for an element different from the global coordinate system that is

390

Building

used to define the entire body.

Frame and Foundation

We

Chapter 14

note, however, that our final

develop element and assemblage relations in the global or

aim

common

is

to

coordi-

nate system.

Although the

were used for applications in

local coordinate systems

Chapters 3-13, the formulation process involved transformations that yielded

For

the element equations in the global system.

instance, in the case of the

one-dimensional line element (Chapters 3-10), the direction of the local


system was the same as that of the global system, and the transformations
occurred essentially in the derivatives and integrations. The isoparametric

formulation involved use of global displacements in the development of

element equations. Hence, both these instances did not involve transformation
of coordinates.

For certain

beam-columns, nonorthogonal
and curved structures (e.g., shells), it is often
systems whose directions are different from the global

situations such as inclined

slabs in the building frame,

necessary to use local


systems.

Then

it is

required to transform the element relations evaluated in

the local system to those in the global system. Such transformation

is

achieved

by using a transformation matrix consisting of direction cosines of angles


between the local and the global coordinates.
Figure 14- 1 4(a) shows a beam-column element in an (orthogonal) global
reference system, x, y,

z,

with a local (orthogonal) system,

x',

y\

z'

attached

with the element. For this case, the transformation matrix can be expressed as

[t]

where l xi

mx

etc.

ly

my

ni

I,

m,

n,

(14-19a)

are the direction cosines of angles between the local

and

global axes; for instance l x represents the direction cosine of the angle

between

As

x'

and x axes and so on.

let us consider the beam-column element in the


two-dimensional x-y plane, Fig. 14- 14(b); the local coordinate x' is inclined

at
i

a simple illustration,

an angle of a with the global axes

x.

Here the transformation matrix

[t]

given by

m,

cos a

sm a

sin

cos a

(14-19b)

[t]

J,
The transformation matrix

[t] is

orthogonal, that

is, its

inverse equals

its

transpose
[t]"

We
{Q ( }

can

now

=[t] r

(14-19c)

and
and load vector evaluated

write transformations for various quantities. If [kj

respectively denote element stiffness matrix

Chapter 14

Frame and Foundation

Building

x, y, z

391

Global
Local

x', y', z

(a)

XcL

lb)

Figure 14-14 Transformation


sional, (b)

of coordinates,

(a)

Three-dimen-

Two-dimensional.

with respect to the local system x\ /,

W=

z',

then

T
[T] [ki][T]

(14-20a)

and

{<W

= imft)

(14-20b)

where [kj and {Q g } are the corresponding matrix and vector referred to the
global system, x, y,

z.

In the foregoing, the transformation matrix [T]


(14-19a); for instance, [T] in Eq. (14-20)

is

is

composed of [t]

in Eq.

given by

[/]

(14-20c)

[71
[t]

[/].

The transformation between displacements can be

m = mw

written as

(14-20

392

Building

where

[u,]

and

[u g ]

Frame and Foundation

Chapter 14

are the vectors of displacements at a point for local

and

global systems respectively.

The element equations

at the global level are


lk.]{q,}

where

{q g } is the vector

now

expressed as
(14-22)

{Q,}

of element nodal unknowns obtained by using Eq.

(14-21).

Once the global

relations are obtained, they can be assembled

by using
com-

the direct stiffness assembly procedure by fulfilling the inter-element

unknowns.
The foregoing concept can be used and extended for other one-dimensional, and two- and three-dimensional elements. The process is usually

patibility of the

straightforward. For further details, the reader can consult various references

including those in Chapter 15.

PROBLEMS
14-1.

Derive the load vector {Q} in Eq. (14-6) for uniform body force
form traction f causing bending. Hint See Chapter 7.

X and uni-

14-2.

Evaluate the matrix [B 6 ] in Eq. (14-10). Hint: Find second derivatives of

in Eq. (14-8a) as indicated in Eq. (14-10).


14-3.

Derive the part of the

stiffness

matrix for the beam-column element cor-

responding to the twist about the z

axis.

Solution:

[k,]

where

G is

the shear

= GJ

-1

!]

modulus and / is the polar moment of inertia. The corzX and


z2 at nodes 1 and 2, respectively.

responding degrees of freedom are

REFERENCES
[1]

[2]

Martin, H. C, Introduction to Matrix Methods of Structural


McGraw-Hill, New York, 1966.
Tezcan,

S. S.,

Struct. Div.

"Computer Analysis of Plane and Space


Vol. 92, No. ST2, April 1966.

Analysis,

Structures," /.

ASCE,

S., and Patil, U. K., "Finite Element Analysis of Building Frames


and Foundations," report, Department of Civil Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Va., 1976-1977.

[3]

Desai, C.

[4]

Bogner, F. K., Fox, R.


element-Compatible

L.,

and Schmidt, L. A., "The Generation of Interand Mass Matrices by the Use of Interpolation

Stiffness

Chapter 14

Building

Formulas,"

in Proc.

Frame and Foundation

Second Conf. on Matrix Methods

393

in Struct.

Mech., Wright

Patterson Air Force Base, Ohio, Oct. 1965.


[5]

[6]

[7]

[8]

S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Desai, C.

Timoshenko, S., and Kieger,


Hill, New York, 1959.

S.

to the Finite

Element Method, Van

W., Theory of Plates and

Terzaghi, K., and Peck, R.


New York, 1967.

B., Soil

Durant, D., "Analysis of a

Plate Bending

Mechanics

in

Shells,

McGraw-

Engineering Practice, Wiley,

Problem Using Code STFN-FE,"

course project report, Department of Civil Engineering, Virginia Polytechnic


Institute
[9]

and State University, Blacksburg, Va., 1977.

Timoshenko,

S.,

Strength of Materials,

Van Nostrand Reinhold, New York,

1930.
[10]

Malek-Karam,
project report,
tute

A., "Design and Analysis of a Two-Story Building," M.E.


Department of Civil Engineering, Virginia Polytechnic Insti-

and State University, Blacksburg, Va., 1976.

PRELUDE
TO ADVANCED STUDY

AND APPLICATIONS

In compliance with our aim to present an elementary treatment,

covered relatively simple problems in Chapters 1-14.

It

is

emphasize, however, that one of the basic advantages of the

method

lies in its

we have

necessary to
finite

element

capacity to permit solutions of complex problems for which

conventional solutions are not available or are

difficult.

The reader interested in the detailed knowledge of the method and its wideranging applications would now need to pursue advanced aspects of the
method.

The complexity of problems rendered tractable by the finite element


method arises due to many factors. These can include irregular and arbitrary
shapes of structures and continua, complex boundary and initial conditions
and loading, nonhomogeneous materials, nonlinear constitutive or stressstrain

behavior, existence of liquids in the media, interaction between

dissimilar media,

and time dependence.

Higher-order approximations and mesh refinement, improved element


formulations, relative merits and trade-off analyses, mathematical properties

convergence,

stability,

and consistency

number of other aspects


method are important.
Following are some of

of various

related to the theory

formulations, and a
and mathematics of the

these topics and their significance. Many of the


advanced topics can be quite involved and can constitute subjects of significant depth by themselves. It will indeed be very difficult to describe them in
detail. Hence, we have essentially listed topics and subtopics relevant to

394

Prelude to Advanced Study and Applications

Chapter 15

engineering.

The bibliography

at the

395

end of the chapter contains publications

that can be consulted for the study of these topics.

THEORETICAL ASPECTS
1

Formulation Procedures
a.

Variational potential, complementary, mixed, and hybrid approaches


:

Ritz and Raleigh-Ritz methods.


b.

Weighted residuals:

collocation,

least

squares,

subdomain,

and

Galerkin.
2.

Transformations of Coordinate Systems

3.

Isoparametric Formulations

4.

Higher-Order Approximations

5.

a.

Polynomial, Lagrangian, Hermitian, and spline interpolation.

b.

Requirements for approximation functions.

c.

Mesh

d.

Higher-order approximation versus mesh refinement.

refinement.

Mathematical Aspects
a.

Basis of formulation

b.

Accuracy, convergence,

c.

Initial

linear

and nonlinear operators, normed

stability, consistency, error

spaces.

bounds.

and boundary value problems.

d. Integration

6.

of equations.

e.

Time

f.

Singularities.

g.

Eigenvalue problems,

h.

Solution of equations.

integration

and

properties.

Factors
a.

Arbitrary geometries.

b.

Nonhomogeneities.

c.

Composite materials.
Boundary conditions, nonlinear boundary conditions,

d.

free-surface

problems.
e.

Loading.

dynamic, and repetitive,


Monotonic and path dependent,
iii. Time dependent.
Interaction effects: between two dissimilar media such
and soil and structure and water.
i.

Static,

ii.

f.

7.

Constitutive (or Stress-Strain) Behavior


a.

Linear

elastic.

b. Piecewise linear elastic.


c.

Higher-order

elastic.

as structure

Prelude to Advanced Study and Applications

396

Chapter 15

d. Hypoelastic.
e.

Elastic-plastic.
i.

ii.

Perfectly plastic: nonfrictional.


Perfectly plastic: frictional.

iii.

Strain softening,

iv.

Cap models and

f.

Viscoelastic.

g.

Creep.

critical state

concepts.

h. Thermoviscoelastic.
i.

8.

Endochronic.

Nonlinear Analysis
a.

Material and geometric nonlinearities.

b. Incremental, iterative,
c.

9.

Initial stress, strain

and mixed procedures.

procedures.

Coupled Problems
a.

Thermoelastic.

b.

Consolidation.

c.

Liquefaction.

10. Fluid

Flow

Multidimensional seepage and consolidation.


11.

Environmental Problems:

Mass

Transport, Diffusion, and Convection

Mechanics and Hydrodynamics


Thermodynamics
Eigenvalue Problems
Computer Implementation
a. Bandwidth and wavefront techniques.

12. Fluid
13.

14.
15.

16.

Evaluation and Comparisons of Various Schemes

BIBLIOGRAPHY
Brebbia, C. A., and Connor,

J. J.,

Fundamentals of Finite Element Techniques,

Butterworth's, London, 1973.

Cook, R. D., Concepts and Applications of

Finite

Element Analysis, Wiley,

New

York, 1974.
S., and Abel, J. F., Introduction
Nostrand Reinhold, New York, 1972.

Desai, C.

Desai, C.

S.,

neering,

to the Finite

Element Method, Van

and Christian, J. T. (eds.), Numerical Methods


McGraw-Hill, New York, 1977.

in

Geotechnical Engi-

Finlayson, B. A., The Method of Weighted Residuals and Variational Principles,


Academic Press, New York, 1972.

Gallagher, R.
1975.

H,

Finite

Element Analysis, Prentice-Hall, Englewood

Cliffs, N.J.,

Prelude to Advanced Study and Applications

Chapter 15

Huebner, H., The

Finite

Element Method for Engineers, Wiley,

Martin, H. C, and Carey, G.


Hill, New York, 1973.

F., Introduction to Finite

Oden,

J.

T., Finite

J. T.,

New

Fundamentals and

Elements of Nonlinear Continua, McGraw-Hill,

and Reddy,

J.

N.,

Variational

Methods

York, 1975.

Element Analysis, McGraw-

Norrie, D. H., and de Vries, G., The Finite Element Method


Applications, Academic Press, New York, 1973.

Oden,

397

in

New York,

1972.

Theoretical Mechanics,

Springer, Berlin, 1976.

Pinder, G.

demic

F.,

and Gray, W. G., Finite Elements


New York, 1977.

in

Subsurface Hydrology, Aca-

Press,

Prenter, P. M., Splines and Variational Methods, Wiley,

New

York, 1975.

Strang, G., and Fix, G. J., An Analysis of the Finite Element Method, Prentice-Hall,
Englewood Cliffs, N.J., 1973.

C, The Finite Element Method


York, 1971.

Zienkiewicz, O.
Hill,

New

in

Engineering Science,

McGraw-

APPENDIX

VARIOUS

NUMERICAL PROCEDURES
SOLUTION OF
BEAM BENDING PROBLEM

INTRODUCTION
To

illustrate the

use of energy procedures and of the methods of weighted

an example of beam bending by


and finite
difference methods the former is based on the minimization concept and is
often considered a forerunner of the finite element method. The coverage of
different methods herein is intended only as an introduction to various
procedures and to give the reader an idea of the available schemes detailed
study of these procedures is beyond the scope of this elementary treatment.
Before the problem of beam bending is considered, we shall give further
details of the methods of weighted residuals (MWRs) introduced in Chapter
2. A good introduction and applications of
are given by Crandall [1].
As stated in Eq. (2-12), the trial or approximation function in the
residuals, in this

appendix we

shall solve

using these procedures. Moreover,

we

also consider the Ritz

MWR

MWR

is

expressed as

(2-12)

,?,.

The undetermined parameters a, are chosen such that the


the domain D vanishes. This is usually done in an average
R(x) (Fig. 2-7) with respect to weighting functions

Jd

398

R(x)W (x)dx
i

= 0,

1, 2,

W
,

(x).

residual R(x) over

sense by weighting

Thus,
(Al-1)

Appendix

Solution of Beam Bending Problem

D is

For a one-dimensional problem, the domain

399

simply the linear extent of

the body.

VARIOUS RESIDUAL PROCEDURES

Wu and depending

There are a number of ways to choose

we

ti

on the choice of

obtain different procedures.

In the case of the collocation method,

W = 3(x - x

(Al-2a)

t).

Then
RixdSix

= 0,

x)
t

1, 2,

1,

=x

[0,

^x

(Al-2b)

n,

where

3
is

the Dirac delta function. This

selected

number of

A 1-1 (a),

the residual

The

means

if

t9

that the residual

is

equated to zero at a

points in the domain. For instance, as


is

equated to zero at n

shown

in Fig.

5 points.

domain can be divided into a number of subdomains [Fig.


and the residual is integrated and equated to zero over each subdomain. This yields the subdomain method. Here the weighting functions are

total

-1(b)],

,+1

x 'J
< x and x

0,

'

>

(Al-3a)

xt+l

Then

R(x)dx

= 0,

In the case of the least-squares

1, 2,

...,-

method

1.

[Fig. Al-l(c)], the

(Al-3b)
weighting

functions are chosen to be

*-g.
and

(Al-4)

minimization of the integrated square residual as

this leads to

f
Jd

R 2 (x)a dx =
t
{

(Al-5)

0.

In Galerkin's method [Fig. Al-l(d)], the weighting functions are chosen as


the coordinate functions from Eq. (2-12), and hence
f

Jd

*(*)?,(*)=

0,

1,2, ...,.

This expression implies that the functions

q> t

are

made orthogonal

(Al-6)
to the

residual R{x). In finite element applications, the approximation or trial

Solution of Beam Bending Problem

400

Domain D

Appendix 1

R(3)

12

I
4

(x=)

= 1,2

(a)

R(x)dx =

-1

0,

= 1,2,3, 4

f:

(b)

JR

(x)a, = 0,

= 1,2

Ry?,dx = 0,

Figure Al-1 Methods of weighted residuals,

Subdomain.

functions are
functions

t ;

(c)

1, 2, 3,

(a) Collocation, (b)

Least squares, (d) Galerkin.

commonly
then the

expressed in terms of shape, interpolation, or basis

are usually chosen as the weighting functions.

BEAM BENDING BY VARIOUS PROCEDURES


The beam and

details are

shown

in Fig.

A 1-2. Assuming flexural rigidity to be

uniform, the governing differential equation


residual

is

given by Eq. (7- lb), and the

is

R(x)

F^ -

P(x),

(Al-7)

Appendix

Solution of Beam Bending Problem

401

Collocation
Finite difference

Subdomain
3L/4

L/2

Galerkin,
Least squares,

(c)

Ritz

0-L

(d)

Finite element

Figure Al-2

Beam bending with

finite difference, (b)

methods,

different

Subdomain.

(c)

(a) Collocation,

Galerkin, least squares, Ritz.

(d) Finite element.

where F = 7 is the
p(x)

is

w is the assumed transverse displacement,

flexural rigidity,

the forcing function,

and x

the coordinate.

is

The boundary conditions

associated with Eq. (7- lb) can be expressed as

W (X

= 0) =

d w,
dx 2

The

first set

tions

w(x

=L)=0,

Ax

0)7

dx

L)

(Al-8a)

(Al-8b)

0.

[Eq. (Al-8a)] represents the essential or forced

and the second

set [Eq.

boundary condi-

(Al-8b)] represents the natural boundary

conditions.

We now choose
unknown w*

the following trial or approximation function for the

a, sin

^+a

sin

^+

a3

sin

^+

a4

sin

= t *0M,
where the a are the undetermined parameters and the
t

(Al-9)
(p t

are the

known

Solution of Beam Bending Problem

402

Here we have chosen w* in terms of the trigonometric functions,


Chapter 7 the approximation function was chosen in terms of

functions.

whereas

Appendix

in

interpolation functions

t.

Note that the function w in Eq. (A 1-9) satisfies the boundary conditions in
Eq. (A 1-8) at the two ends of the beam. In Chapter 7, when we used
Galerkin's method for each element, then only the geometric boundary
conditions [Eq. (Al-8a)] at the ends were used to modify the assemblage

equations; as explained in Chapter

boundary conditions

the natural

3,

[Eq. (Al-8b)] are satisfied automatically in an integrated sense.

To

express R(x) in terms of

w and

its

derivative,

we

differentiate the

expression in Eq. (A 1-9) four times as

-T-;

1*cl x sin -j-

+
where X

256A 4 a 4

16/ 4 a 2

sin

-j-

+ 8U

a3

sin

-j

sin^,

(Al-10)

= n/L.

Now we shall consider the solution of the beam bending problem by using
a number of different procedures. For this illustration, the following prop-

assumed

erties are

E=

10

10 6 psi,

= 10 in.,
A = in. x in. = in.
PA = 500 and PB = 1000 lb/in.
L

Collocation

As shown
4L/5,

A 1-2, we chose four points at x = L/5,

in Fig.

1, 2, 3, 4,

at the supports.

from the end A. Note

instance, for Xj

X4a

sin

4.77

103<xi

7.72

7.72

x 10 3 ai

4.77

103<xi

103<xi

2L/5, 3L/5, and


identically zero

j-

= x )=0,

= 1,2,3,4.

(Al-lla)

= L/5, we have

~ + 16A a
4

256A 4 a 4

and so on. The

is

Then we have
R(x

For

that the residual

sin

sin

^^ -

^ y + 8U a

Pb

Pa

sin

^~

--Pa =

(A-l lb)

resulting equations are

+
+
-

1.23

x 10^2
10 4 a

7.63

7.63

x 10 4 a 2

1.23

x 10*a 2

+ 6.25
- 3.86
- 3.86
+ 6.25

x 10 4 a 3
x 105a 3
x 10 5 a 3

105<x 3

+
+
-

=
10^4 =
10 6 a 4 =
10 6 a 4 =

x 10a 4

600,

1.98

700,

1.98

1.22

1.22

(Al-12)
800,
900.

Appendix

Solution of Beam Bending Problem

403

Solution of these equations gives values of a, as


a,

a3

= 0.11374399,
= 0.00033150,

a2
a3

=
=

-0.00105974,
(Al-13)

-0.00001564.

Hence, the approximate solution according to the collocation method

0.11374 sin

^ - 0.00106

-0.00001564

sin

^+

0.0003315

sin

is

^*

4nx

(Al-14)

sin

Subdomain Method
Here,

we have

[Fig.

-2(b)]
rL/4

R(x)dx

o,

R(x)dx

o,

R(x)dx

0,

R(x)dx

0.

Jo
r 2L/4

J L/4

(Al-15)
r

LI 4

J 1L 4

L
[

3Z.

After integrations, the resulting four equations are


7.57

x lCPai

1.83

10 4

1.83

x 10 4 ai

7.57

x 10 3 ai

ai

+
+
-

1-19

x 10*a 3

4.93

x 105a 3

x 10 5 a 2

+
-

4.93

x 10 5 a 3

1.19

x 10 6 a 3

2.07

2.07

x 105a 2

2.07
2.07

105<x 2

10 5 a 2

+
-

3.31

x 10^4

3.31

x 10*a 4

+
-

3.31

x 10 6 a 4

3.31

x 10 6 a 4

=
=
=
=

1410,

1720,

(Al-16)
2030,
2340.

Solution of these equations leads to the following approximation

0.12388 sin

^-

0.00004724

0.00151 18 sin

^+

0.00078719

4nx

sin

^
(Al-17)

sin

Least-Squares Method

In this procedure, the weighting functions are

&*

(Al-18)

Therefore,

W sin-p,
x

3nx
3

sin

W,

2nx
sin

W = sin Anx
fjjr

>

Solution of Beam Bending Problem

404

According to the least-squares method

[Fig.

Appendix

-2(c)],

*(*)*& =

1
or

0,

sin

j- dx =

0,

sin

j- dx =

0,

!*(*) sin

^<fr =

0.

R(x)

R(x)

sin -j-

dx

(Al-19)

R(x)

The

final

4.06

four equations and the resulting approximate solution are

x 10 4 ai
x

ai

ai

Oxai
w

0.11760 sin

+ 0xa
+ 0xa
+ 3.29 x 10 6 a
+ 0xa

+ 0xa 2
+ 6.49 x 10 a 2
x a2
+
+ 0xa 2

^ - 0.001226

0.00003830

sin

sin

+
+
+
+

0xa 4
0xa 4
x
1.04

^+

= 4774.65
= -795.77,
a4
= 1591.55,
10 7 a 4 = -397.89,

0.0004837

sin

(Al-20)

^
(Al-21)

Galerkins Method
It

is

method

incidental that the weighting functions

are the

same

in this specific case,

same

as the functions

(p t

in the least-squares

used for the Galerkin method. Hence,

both the Galerkin and least-squares methods yield the

solutions.

Ritz

Method

In the Ritz method [1,2] the potential energy in the body (beam)
trial functions, and the resulting expression

expressed in terms of the

minimized with respect to


a For instance,
f

a,.

is
is

This leads to a set of simultaneous equations in

^ = \[ F &) dx -\j wdx

<

A1 - 22 >

with

p{x)=Pa

+ ^(Pb-Pa)-

(Al-23)

Solution of Beam Bending Problem

Appendix 1

405

Minimization of IT, with respect to a gives


t

dll p

= 0,
~

da 2

'

(Al-24)

dUp

0,

da 3
dll p

= 0.

(?a 4

For

problem, the equations are the same as in the Galerkin and

this specific

least-squares methods,

and the approximate solution

is

the

same

as in

Eq. (Al-21).

is

Comment: It may be noted that in the Ritz procedure the potential energy
minimized for the entire beam in other words, the limit of the integral is
;

from

to L.

The concept is thus

similar to the finite element

method (Chapters

method, the minimization of


achieved for the domain composed of a patchwork of elements.

3-5), except that in the case of the finite element

11,

is

Finite

Element Method

method can be achieved in a manner


Chapter 7, with the subdivision consisting of two
elements [Fig. A 1 -2(d)]. We can use Eq. (7-15) for generating the twoelement equations and then performing the assembly with boundary conditions, and the solutions of the resulting assemblage equations are [3]

The

solution by the finite element

identical to that covered in

"

12

30

30

100

-12
-30

F -12 -30

24

125

30

30

50

50

-12

30

200

-12
-30

-30

12

-30

50-30
Solution after introduction of w = w =
w = 0.1171875,
w = 0.000000,
6 = 0.0007295,
0, = 0.036665,
x

rv 2

Finite Difference

'

w2
>

50

rw

100_ [e 3l

1437.5'

1250.0

e,

30

W;

3750.0

(Al-25)

<

417.0

2312.5

-1875.0,

gives

= 0.0000000,

(A 1-26)
-0.0383335.

Method

Before the era of the

finite

element method, the

was the commonly used technique

method
and mathe-

finite difference

for problems in engineering

Solution of Beam Bending Problem

406

matical physics. Details of this

However, we

method, the reader can

The

beyond the scope of

are

this

book.

beam problem using this method mainly to


of commonly used numerical methods. For study of

shall solve the

complete the discussion


this

method

Appendix 1

refer to various textbooks

method

finite difference

[1].

based on the concept of replacing the

is

continuous derivative in the governing differential equation by approximate

For

finite differences.

Fig.

instance, various derivatives are

approximated as

(see

Al-3)

Second

w i+1 w
Ax

dw

First derivative:

dx ~

derivative:

d2 w
dx 2

w,_,

"

Third derivative:

Fourth derivative:

d4 w
dx 4

w i-2
-i

Substitution of

2Ax

w i+1

2w,_

2w i+l

(Al-27)

w i+2

2(Ax) 3
yv t

4w

_,

6vv,-

4w /+1

Ax 4

to replace the fourth-order derivative in Eq. (7- lb) leads to

4w -_
t

6w

4w i+l

w i+2

= 0,

1, 2, 3, 4,

5 for six points in the

-2(a)] gives six

(A 1-28)

Pf

Ax'

into five segments [Fig.

tion of the

Ax

2w

w,-_ +

p Wj-2

Ax 2

d3 w _
dx 3

Use of Eq. (Al-27)

^ w w _i w
Mw
~~

beam domain

boundary conditions [Eq. (Al-8)]

finally gives

four simultaneous

Figure Al-3 Finite difference approximation.

i-1

divided

simultaneous equations. Introduc-

+3

Appendix

Solution of Beam Bending Problem

wu w2

equations in

vv 3

5w 4h>! +
w
+
i

and

u> 4

+ w
4w
+ 6w
- 4w

4w 2
6h> 2

4w 2

w2

407

+
+

=0.0115,

w4
4w 4
5w 4

= 0.0134,
= 0.0154,
= 0.0173.

(A 1-29)

Solution.

with

==

w5

= 0.070656,
= 0.115968,

w 2 =0.114432,
w4

(A 1-30)

= 0.073344,

0.

COMPARISONS OF THE METHODS


To compare

from various methods, we first state the closed form


on the strength of materials theory [4]

results

solution for the displacement based

^(L - 2Lx + x
2

) -|-

)X

(/7

f80ir^l

(3^

10L 2 * 2

Results for displacements at typical locations on the

7L 4 ).

(Al-31)

beam by

using

various methods and the closed form solution [Eq. Al-31] are compared in

Table

AM.
TABLE

Al-1

Comparisons for Displacements


Deflection,

Location from

Method

in.

in.

End A

5 in.

in.

8 in.

Collocation

0.067

0.110

0.113

0.109

0.068

Subdomain

0.072

0.117

0.123

0.118

0.075

Galerkin

0.068

0.111

0.117

0.112

0.071

Least squares

0.068

0.111

0.117

0.112

0.071
0.071

Ritz

0.068

0.111

0.117

0.112

Finite element

0.067

0.110

0.117

0.069

Finite difference

0.071

0.114

0.116

0.116

0.073

Strength of

0.068

0.111

0.117

0.112

0.071

materials

The

results for the first five

methods and the

last

procedure are obtained

and
by substituting various values
(Al-31). Since we did not have a point at x = 5, there is no direct result at
that point for the finite difference method. In the case of the finite element

of*

into Eqs. (Al-14), (Al-17), (Al-21),

Solution of Beam Bending Problem

408

Appendix

method, once the nodal displacements and slopes are obtained, values at
other points can be obtained by substitution of (local) coordinates into
Eq. (7-2). We can also compute moments and shear forces by using the
results with second and third derivatives, as we did in Chapter 7.
For examples of solutions of problems similar to the beam bending and
other problems solved by using different procedures presented herein, the
reader can refer to various publications such as Crandall [1].
For the foregoing beam bending problem, the results for displacements
from various procedures are close to each other and to the result from the
closed form solution. These comparisons are presented only for the sake of
introducing the reader to some of the available numerical procedures. The
merits of the finite element

method become evident as we solve problems with


and geometric properties and

greater complexities in factors such as material

loading characteristics.

REFERENCES
[1]

Crandall,

[2]

Abel,

[3]

Desai, C.

J.

S.

H., Engineering Analysis, McGraw-Hill,

F., (private
S.,

[4]

Timoshenko,
1956.

S.,

York, 1956.

communication).

and Abel,

Nostrand Reinhold,

New

J. F.,

New

Introduction to the Finite Element Method,

Van

York, 1972.

Strength of Materials,

Van Nostrand Reinhold, New York,

APPENDIX

SOLUTION OF
SIMULTANEOUS EQUATIONS

INTRODUCTION
Most problems

solved by using numerical methods result in a set of algebraic

simultaneous equations of the form [Eqs. (2-21) and (2-22)]


[K]{r}

= {R}.

(A2-1)

These equations can be linear or nonlinear;

we have

in this text

essentially

dealt with linear problems.

In general, Eq. (A2-1)

is

often expressed in matrix notation as


[A]{x}

= {b)

(A2-2a)

or simply as

Ax
where

[A]

is

= b,

(A2-2b)

the matrix of (known) coefficients such as

the vector of

parameters such as {R}.

*n*i
021*1

a Hl x

(J

[Eq. (2-22)], {x}

is

and {b} is vector of (known) forcing


In expanded form Eq. (A2-2) can be written as

unknowns such

as

+ 012*2 +
+ 022*2 +
+

an2 x 2

{r},

+
+

a ln xn

a nn xn

a lH xn

=b
=b

l9

29

(A2-3)

=b

n,

409

410

Appendix 2

Solution of Simultaneous Equations

where n

number of unknowns and denotes

the

is

number of

the total

equations.

METHODS OF SOLUTION
The two common methods

for solution of Eq. (A2-3) are the direct and


Gaussian elimination and a number of its modifications

iterative procedures.

are examples of direct methods, and Jacobi, Gauss-Seidel, successive over-

and symmetric successive overrelaxation (SSOR) are


We shall briefly illustrate some of

relaxation (SOR),

examples of the

iterative techniques [1-3].

these techniques.

Gaussian Elimination

This

is

equations.

perhaps the simplest and a


It is

common method

for solving linear

based on the idea of creating a sequence of equivalent systems

of equations by using a number of steps of elimination, and then solutions for


the

unknowns x

lent systems

are obtained by a process of back substitution.

The equiva-

have the same solutions, but each successive sequence

than the previous one. Let us consider the system of equations

and denote the

sequence by superscript

initial

+
+

iV*i
fllV*i

tfi

tfixt
ffl*a

*i+.ff*a

+
+
+

+
+

.-.

flfite

is

simpler

Eq. (A2-3)

(1) as

= M
= W\

a[xn

in

+ ax =b
n

(A2-4a)

>\

or

=b m

A (1) x

(A2-4b)

step of elimination, we use the first equation in Eq. (A2-4a) and


by appropriate multipliers so as to annihilate the first terms of all
the subsequent equations. For instance, if the first equation is multiplied by
k 2l (tfiV/fliV) an d then added to the second equation, we have

In the

multiply

first

it

(a +Xi*iV)*i

(flffl

*2ia 2 l)x 2

(affi

Aaiaffitei

+ hiW)

(A2-5)

or
fl>X
In

= b?\

where a 222 == dil + A 21 2 and so on and the superscript (2) denotes the
second sequence. Similarly, we can multiply the first equation by
( 1 )

A1 31

fl 3

-TIT'

flu

A1 41

"4 1
(i)>

flu

*nl

u n\
(1)

flu

Appendix 2

Solution of Simultaneous Equations

and add to the second,


the end of

first

third, etc., equations to obtain a modified

I'M

+ a^ixi +
+ a%x +
+ ax +
2

+ a$x +
2

In the next step,

we

+ aiite
+ a%xn
+ flff*. - o

/>< 2 >

/)

'

aff x.

(2 >
3

(A2-6a)

.(2)
b

multiply the second equation in Eq. (A2-6a) by

/32

42

wr

~wr

x *

- sgf

results of the third, fourth, etc., equations,

iV*i

sequence at

step of elimination as
(l)

and add the

41

!)

+
+
+
+

iy*a
(2) Y

"22-M

+ fl,Vxj
+ a&Xt
+ JV*
+ ax
S

+
+
+
+

which leads to

+ affjc. = 6
+ t&x. = 6i2
+ a>* = W,
+ a%x = Ai,
(1)

'i

>

>,

(A2-6b)

+ fl& *3 +
Finally, at the end of n
elimination steps, we shall have
a.'M" + ai*2 + <*,V*s + JV*4 +
+ ai'.'x, = M
2
+ Sx2 + ai 3>x + a$JC +
+ tfi*. = b{ \
+
+ a$x + aflx, +
+ a%x = 6<
+
+
+ a^xt +
+ ai?x = b\
}

.',

>,

(A2-6c)

fl

We

(-l)

fl

(-l)

A(n-l)

note that Eqs. (A2-6a), (A2-6b), and (A2-6c) are equivalent to the

original Eq. (A2-4) in the sense that they

(A2-6c)

is

x n can be obtained
Back

The

all

have the same solution. Equation

a triangular system of equations, and the solution for the


directly

from the

unknown

last equation.

Substitution

first

(A2-6c) as

step here

is

the solution for

x n from

the last equation of Eq.

412

Solution of Simultaneous Equations

Appendix 2

= fcu

(A2-7a)

n.n

Now

the solution for

xn _ can be found from the


x

*.->

(n

l)th equation as

(A2-7b)

^fe^*-

Since x is known from Eq. (A2-7a), xa _, can be found easily. The process can
be repeated until the solution for x, is obtained.
The foregoing is the basic Gaussian elimination technique. A number of

modifications and alternatives such as Crout, Jordan, Aitken, and GaussDoolittle can be used depending

on the

characteristics of the system equa-

tions [1-3].

Banded and Symmetric Systems


In finite element applications, very often

we encounter systems of equa-

banded and symmetric. In banded matrices, nonzero coeffioccur only on the main and adjacent diagonals, and other locations

tions that are


cients

have zero
a tj

=a

jt .

coefficients.

For

Moreover, the system

is

very often symmetric; that

instance, the following represents a

banded

is,

(tridiagonal)

symmetric system of equations

flii*i+tfi2*2+
#12*1

+
+
+
+

=b u

-f-

=b
+
+ #23*3 +
=b
23 *2 + #33*3 + #34*4 +
+ #34*3 + #44*4 + #45*5 =
+#35X4 + 055X5=65.
+

#22*2
tf

(A2-8)

2i

^>4>

Bandedness and symmetry of the equations allow

significant simplifica-

tions in the foregoing general elimination procedure. It

is

necessary to store

only the nonzero elements in the computer and only the coefficients on the

main diagonal and the upper or lower diagonals

[4].

SOLUTION PROCEDURE
Almost

all

systems of equations resulting from the

are relatively large

portion of computational effort in a


solution of these equations. It

is

not

the subsequent iterative procedures

For sake of introduction, we


elimination process.

finite

element analysis

and require the use of the computer. In


finite

element solution

fact,

is

a major

spent in the

difficult to program the elimination and


on the computer.

shall

now

illustrate the use

of the foregoing

Example A2-1.

Solution by Gaussian Elimination

Consider the system of three equations, Eq. (3-38). In Chapter

we

equations by a form of Gaussian elimination. Here

3,

we

solved these

follow the foregoing general

procedure
100 (1) jc!

+
+

100 (1) * 2

-lOO^*! + 200 (1) * 2


(1)

^!

100 (1) x 2

(1)

*3

100 (1) *3

200 (1) *3

=
=
=

7.5 (1)

15.0 (1 \
15.0 (1)

(A2-9)

Details:

+ 10Q(i) _
Therefore

we have

the second equation as

(200

100) (2) x 2

100 (1) * 3

15.0 (1)

100 (2) * 2

100 (2) * 3

22.5 (2)

7.5 (1)

or

Since the third equation already has zero in the

lOO^*!

- 100
+ 100
- 100

(1)

(2)

x2

(2)

Now

A 32

+100/100

*2

+
- 100
+ 200

term, the modified equations are


7.5 (1 \

(2)

*3

(1)

*3

=
=

22.5 (2) ,
15.0 (2)

Therefore, the third equation becomes

1.

x2

first

(200

100) (3) X3

15.0 (2)

37.5 f3)

22.5 (2)

or

+
Hence, the

final

100 (3) x 3

modified equations are

IOO^jcj

- 100
+ 100
+

(1)

(2)

The back

jc 2

*2

+
- 100
+ 100

7.5 (1)
(2)

(3)

;t3

jc 3

=37.5 (3)

22.5

(2)
,

substitution gives

*3

_37.5

100
22.5

100

100
100

37.5

(A2-10)

100

60
100
7.5

100

100
100

60
100

67.5
100'

413

Iterative Procedures

In the iterative procedure, an estimate

made

is

for the

successively corrected in a series of iterations or trials.


is

continued until convergence, which

is

The

unknowns x and
iterative

is

procedure

often defined by selecting a small

number by which the final solution differs from the solution in the
previous iteration. The simplest iterative procedure is the Jacobi scheme [1-3].
acceptable

To
it

illustrate this

procedure, consider the set of equations (A2-3) and express

as

a lt xP
(

021*

O)

o)
3

i*i

o)

0*i*i

+
+
+

a 12 x?>

* 30)

0)

*23*3

a 23 x 20)

+
+
+

a 33 x 3 0)

a n2 x 20)

aB2 x 3"

- - -

0)

tf22*2

tf

13

where the superscript (0) denotes


compute the values of x as
Y (l)
Xi

0n

Y 2(l) _

Y 3(l)

\2 xY

21 XY
fl

a 22

_ 031
Y

Now we

X
0_nl

032
Y

XY

~7,

0_n2.

a nn

compare
\x\

' ' '

= bu
2nX = b
=b
a 3n x

+
+
+

a ln x

a nn x^

<l

2t

x\
l)

x)

\3

0u

xy (0)
3

032 Y (0)

X3

0)

=b

\n

a 22

_ 03
Y

~7,

033

(O)

0_n3_

a nn
0)

(0)

(0)

"

b
|

we

-\

j_
-f-

'

#2
a 22

b3

(A2-llb)

'T,

033

j_ ^3

...

Xv (0)
3

first iteration,

_ a 2n XY

(0)
n

(A2-lla)

m9

In the

_ a xY

...

"

(0)

with x\
{

0)

033

Y (0)

a nn

a 22

(0)

7,

initial estimates.

(0)

033

Y n(l)
x

(0)

-f-

033

by finding the difference

\<e,

1, 2,

(A2-12a)

,n,

where e is a small number. If this condition is satisfied, we accept x} l) as the


approximate solution; otherwise, we proceed to the next iteration. In general
at the rath iteration we have
xv (m)
l

X2

x (m>

ai\

an

= _q

fl33

x nm)
(

414

an

y (m-l)
2

'

Cl\n

# 23 v (m* 3

1)

ain v (m-l)

32 **2
v (m-l)
033

Qnl -*2
v (mQnn

1)

Q/3

a nn

v (m-l)

an

an

x (m-l)

Ojjym-i)

an

x (m-l)

Qn\ v (m-l)
x

a nn

fll2

x (m-l)
3

a22

an

a 3n v (m-l)
x"
7.
#33

as3

an

(A2-llc)

a nn

Appendix 2

Solution of Simultaneous Equations

and the convergence condition


\xi

m ~ l)
x\

m)

415

is

<

1, 2,

(A2-12b)

n.

The foregoing procedure can be slow to converge; that is,


number of iterations before an acceptable solution is

a large

it

may

require

obtained.

The

improved by using the Gauss-Siedel procedure in


which the solution for an unknown during an iteration is used in the comrate of convergence can be

unknowns

putation of subsequent

v (m)

*2

x3

A21 Y (m)

A3 1 urn)
X
A33

flln

1)

fl/2

xv (m)
2

a nn

Example A2-2.

first

A3

iteration

*<>

^(0.6125)

Second

(i)

( o.575)

2)

_ A^^fm-l)
an n

_|_

bi_

aii

_ 1^-1) +

^_,

(A2-13)

A33

**.
fl/m

x <s
x

We

x 2>

guess a solution set as

x 3>

0.50,

1^0(0.5)

^=

+ i|

a575

(o)
|

=
=

0.50.

0.1125

0.04375

>
>
>

'

0.6125,

(A2-14a)

0.38.25.

0.005, then the convergence check

is

0.005

0.005>

not acceptable.

(A2-15a)

0.005J

iteration:

^.=^(0.6125)

+T55

100

_
_

we have

x a)

x (m)
3

a nn

x (i) -jc<>| =0.075

b\

All

A22

0.50,

4o

we choose

..

xv (m)
2

*i"=Jo>- 5>

Tf

Solution by Gauss-Siedel Procedure

x<>

for the

A2 3 ^(wi-l)
3
A22

Consider the equations in Eq. (A2-9).

Then

v (m-\)

All

A3 2 (m)
-*2
A33

a n

flu

A22

flwl

y(m)

Al2 (mx2
flu

x? =

+1

^(0.6875)

^(0.38125)

^(0.609375)

45s

=0.6875,

^=
=

0.609375,

0-3796875.

(A2-14b)

416

Solution of Simultaneous Equations

Appendix 2

Convergence check.
jc(2)

*< 2 >

=
=
x[
x^\ =
x \n\
l)

>
<
<

0.1125
0.003 1 25

0.0015625

0.005^

not acceptable.

0.005 }

(A2-15b)

0.005,

Third iteration.

(3)

100
77^(0.609375)
100

+H^
100

100
0.684375)
200

^(0.3796875)

^(0.6070312)

x?

7.5

15

200

0.684375,

0.6070312,

0.3785156.

(A2-14c)

Convergence check:
\

X U) - x (2)| =

0.003125

xy =
_x =
|

0.002344

0.001172

.12)

Hence the

x (3)

j2)

final solution at the

< 0.005]
< 0.005 >
< 0.005J

end of three
Xl

x2
x2

=
=
=

(A2-15c)

acceptable.

iterations

is

0.6844,

0.6070,
0.3785.

We can improve on this solution by setting a more severe condition on convergence,


say = 0.001. Then additional iterations are needed for an acceptable solution.
For

large systems of equations generated in the finite element applica-

becomes necessary to use improved iterative schemes. One such


improved scheme is successive overrelaxation (SOR) [2]. To show the
motivation of this method, we first write the Jacobi and Gauss-Siedel schemes

tions,

it

in matrix notation as follows [2]

Jacobi:

x (m

>

=(L+

U)x lm

>

(A2-16a)

fi.

Gauss-Siedel.

Lx (m)

L and U are

the lower

Ux (m

>

and upper triangular parts of the matrix


For instance,

the constant term at a given stage of iteration.

A2-2, at

m=

1,

100

Too
ioo
200

(A2-16b)

fi.

[A]
in

and fi is
Example

Appendix 2

Solution of Simultaneous Equations

100
100

417

100
200

U
7.5

fi-m

15
15

We

where

can

now

x(m>

= X (m-D

write
_j_

^[(^(m)

ux (m

l)

P)

x {m

X)

(A2-16c)

the overrelaxation factor and the second term

co is

on the right-hand
Rearrangement of terms in Eq. (A2-16c)

side denotes the correction term.

gives

x (m)
which

=(/-

is

coL)-

[{\

a statement of

For

significantly.

(o)I

SOR

coU]x (m

SOR

[2].

details the reader

may

l)

co(I

- coL)-

/},

(A2-17)

accelerates the solution procedure

consult Refs.

[1, 2].

COMMENTS
In the foregoing, we have presented only a very elementary introduction to
some of the solution procedures. There are also available a number of other
schemes and subschemes. Moreover, there are a number of aspects related to
the numerical characteristics of the set of equations that can influence the

accuracy and

reliability

of a procedure. For instance, the magnitudes of the

diagonal elements ati which become "pivots" in the elimination procedure

can influence the computational characteristics. The initial guess or estimate


of the unknowns and the value of o become important in the iterative
procedures.

of the available procedures and the


beyond the scope of this book. The reader

Detailed descriptions

characteristics of the equations are

can refer to many publications [1-4] for further study.

REFERENCES
[1]

[2]

L.,

New

York, 1965.

2,

Numerical Linear Algebra, Oxford University Press,

T., A Survey of Numerical Mathematics,


Addison-Wesley, Reading, Mass., 1972.

Bathe, K.

J.,

Prentice-Hall,
[4]

to

Young, D. M., and Gregory, R.


Vol.

[3]

An Introduction

Fox,

and Wilson, E. L., Numerical Methods in


Englewood Cliffs, N.J., 1976.

Desai, C. S., and Abel, J. F., Introduction


Nostrand Reinhold, New York, 1972.

Finite

to the Finite

Element Analysis,

Element Method, Van

APPENDIX

PHYSICAL MODELS

INTRODUCTION
may

often be instructive and useful to construct physical models to illussome of the principles of the finite element theory. There can be a number
of possibilities for constructing such models. Here we consider details of
some of the simple models that the author has tried.

It

trate

MODELS FOR CONVERGENCE


As

stated in Chapter

plastic.

As shown

1,

one can construct models made of cardboard or


can illustrate the idea of convergence as

in Fig. 1-8, they

polygons of an increasing number of sides are drawn inside or outside of a


circle.

MODELS FOR BEAM BENDING


One of

the

cretization

(Chapter 7)

main ideas here can be the illustration of the principle of disand interelement compatibility with respect to the order of

approximation functions.

The models can be made of a


described subsequently were

RTV

Products Department, General Electric, Watford,


418

The models
Rubber (Silicone

suitable flexible material.

made from

Silicon

New

York). This material

Physical Models

Appendix 3

419

can be cast into molds and is available in different colors. Its properties are
suitable for illustrating deformations with small loads and hand pressure.
Figure A3-1 shows a plastic mold used to cast the models. Figure A3-2(a)
shows a model of a continuous beam. The model in Fig. A3-2(b) was cast

Figure A3-1

Mold made

Figure A3-2 Physical model for

beam.

beam

of plastic.

bending, (a) Continuous

Physical Models

Mesh with "rigid" elements: linear apMesh with deformable elements: higher-order

Figure A3-2 (Contd.) (b)

proximation,

(c)

approximation.

A flexible string was passed through holes in each


The holes are provided by (copper) pipe, which makes each element
considerably stiff. Under a load, the composite will deform such that gaps
into four pieces (elements).

element.

may

develop at the junctions, thus violating interelement compatibility.

With

sufficient tightness in the string, the

model may not show gaps and can

indicate displacement compatibility at the junctions, but slopes at the junc-

Appendix 3

Physical Models

421

may be considered to illustrate a linear


approximation function for the transverse displacement w.
Figure A3-2(c) shows a model made in four elements. The elements are

tions can be seen to be different. This

connected at the junctions by using hooks at the top and at the bottom of the
beam. This arrangement permits deformation of each element and improved
connection at the junction as compared to the arrangement in Fig. A3-2(b).

Under

load,

the

model deforms without gaps and

indicates

improved

interelement compatibility. This can be considered to illustrate higher-order

approximation for w.

MODEL FOR PLATE BENDING (Chapter 14)


Figure A3-3 shows a model for plate bending
sheet.

The four elements

[Fig. A3-3(a)].

made out of

(thin) plastic

are riveted (or glued) at the edges of the

A screw arrangement [Fig.

A3-3(b)]

is

model

provided at the middle

of the base plate such that the top of the screw touches the bottom (at the
center) of the four elements.

Figure A3-3 Physical model for plate bending, (a) Four-element


mesh, (b) Arrangement for applying "load."

Physical Models

422

By moving
sively)

the screw

up and down, the

Appendix 3

plate assembly can be (progres-

deformed. The interelement compatibility in terms of displacement

and d 2 w/dxdy) can be qualitatively illustrated by observing the movements at the junctions of the plates.
The effect can be accentuated by attaching colored tapes on either side of a

(w) and

its

three derivatives (dw/dx, dw/dy,

junction.

COMMENTS
The foregoing

are examples only of simple models. Other similar

and

sophisticated models can be constructed to facilitate introduction of various

concepts in

finite

element analysis.

APPENDIX

COMPUTER CODES

INTRODUCTION
Descriptions of a

number of computer codes

Most of the codes

described herein have been prepared such that the beginner

are included in this appendix.

can understand and use them with relative ease. They are relevant to many
topics covered in this book, and hence can be used by the teacher and the
student for solving specific problems. Brief statements of other codes relevant
to a given topic are also given

problems for

these codes

class illustrations or for

may be used

for solving

advanced

term projects.

The codes described here can be made available to the reader and the
and teaching purposes at the cost of reproduction, mailing and other required costs. For details such as descriptions of
teacher essentially for personal

background theory,

guide,

user's

release of codes, the reader

may

sample problems, conditions for the

contact the author. In addition to the codes

described here, other advanced codes can also be

CHAPTER

made

available to the user.

CONS-1DFE: One-Dimensional

Consolidation

with Nonlinear Properties

This

is

dependent

a modification of
settlement

DFT/C-1DFE

analysis

described in Chapter 6 for time

of foundations

with

nonlinear

material

properties.

423

Other Codes on Consolidation

Codes for time dependent settlement analysis of foundations

idealized as

plane strain and axisymmetric consolidation of problems such as layered

foundations and earth banks are available. Both linear and nonlinear (plasticity)

models are included for material behavior. Capabilities for time

dependent loading and simulation of construction sequences such as embank-

ment and excavation

CHAPTER

also exist.

BMCOL-1DFE:

Analysis of Axially and Laterally Loaded

Beam-Columns such as

Piles

and Retaining Walls;

Linear and Nonlinear Analysis

This code can solve problems such as beams, beam-columns (Chapter

beams on deformable foundations,

axially

and

laterally

retaining walls idealized as one-dimensional. Linear


strain behavior for

deformable supports such as

loaded

piles,

and nonlinear

soil

7),

and

stress-

foundations can be

included.

The output

in the

is

form of nodal displacements and

rotations,

and

bending moments.

CHAPTER

MAST-1DFE:
This code

is

One-Dimensional Diffusion-Convection

based on the formulation described

in

Chapter

8.

Other Codes on Diffusion-Convection

This

is

a two-dimensional code for solution of situations such as

salt

water intrusion and other concentration problems.

CHAPTER

10

WAVE-1DFE:

One-Dimensional

Wave

Propagation

This code is based on the formulation described in Chapter 10. Wave


propagation for problems idealized as one-dimensional such as in bars and
soil

424

media can be

solved.

CHAPTERS

AND

11

FIELD-2DFE:
Field Problems

12

Analysis of Two-Dimensional Steady State


:

Torsion, Potential Flow, Seepage, Heat Flow

This is a common code for solution of a number of steady state field


problems governed by similar differential equations (Chapters 11 and 12).

The following

table gives the details of each

problem and

its

corresponding

option code.

Option

Code

NTYPE

Problem
Torsion

Relevant Problems

(Chapter 11)
Potential

Output Quantities

Torsion of bars of regular or

Nodal

irregular cross sections

stresses, twisting

Flow through

Nodal

stress functions, shear

moment

Flow

Velocity

pipes and open

channels

Potential

Stream
Function
Seepage (flow
through
porous
media)
Heat Flow

Flow through

velocity potentials,

velocities, quantity of flow

pipes and open

Nodal stream

functions,

channels

quantity of flow

Steady state confined seepage


through pipes, foundations

Nodal heads,

of

dams and

sheet piles,

velocities,*

quantity of flow

and

earth banks and wells

Steady state heat flow in bars,

plates, slabs,

and other plane

Nodal temperatures, quantity


of heat

bodies

The formulation

is

based on four-node isoparametric quadrilateral

element with linear constitutive laws; for instance, for seepage, Darcy's law

assumed to be

is

valid.

Other Codes on Flow Problems

Codes

for two-dimensional transient free (phreatic) surface seepage

through plane bodies such as earth banks, dams, and wells.


Code for three-dimensional steady and transient free surface flow
through arbitrary three-dimensional porous bodies such as earth
banks, dams, and junctions between various structures, e.g., abutment

and dam.

Code
plates

for two-dimensional time dependent heat flow in bodies such as

and

bars.

425

CHAPTER

13

PLANE-2DFE: Two-Dimensional

Plane Stress-Deformation:

Linear Analysis

This code

is

capable of analysis for problems idealized as plane

plane strain, and axisymmetric (Chapter

13). It

can handle linear

stress,

elastic analy-

of engineering problems such as (approximate) bending analysis of beams


and inplane or membrane analysis of flat plates or slabs, shear walls, earth and
concrete dams and slopes, foundations, underground pipes and tunnels, and
sis

retaining walls.

Surface and nodal point loads as well as self weight load can be applied.

version of the code with graphic option which can plot zones of equal

and displacement intensities is also available.


The finite element formulation is based on a four-node isoparametric
quadrilateral element (see Chapter 13) and on linear elastic stress-strain law.
The output is in the form of nodal displacements and element stresses.
stress

Other Codes for Plane Bodies

code for two-dimensional nonlinear analysis of problems idealized


and axisymmetric is available. The stress-strain behavior
can be nonlinear elastic, elastic-plastic (critical state model), with linear
elastic behavior as a special case. The code has an additional feature of
handling soil-structure interaction problems this is done by providing
a special nonlinear interface or joint element between the structure and
soil. The formulation is based on a four-node isoparametric quadrilateral element (see Chapter 13). Some examples of engineering problems are: retaining structures, axially loaded structures such as pile
foundations, cylindrical tanks, dams and slopes, and underground
as plane strain

structures.

In the case of pile foundations, the code also has the facility of
outputting design quantities such as bearing capacity, wall shear

and point loads,

friction

in addition to

nodal displacements and element

stresses.
2.

This

is

a code for three-dimensional analysis with linear and nonlinear


and plasticity) constitutive models. It has a provision for

(elasticity

interface elements.

426

CHAPTER
STFN-FE:

14
Analysis of

Frame

Structures and Foundations:

Linear Analysis

In a general building frame type of problem the beams and columns are
approximated as one-dimensional beam-column elements, the slabs as plates
subjected to inplane and bending (Chapter 14), and the foundation is replaced
by equivalent spring elements. The spring elements can also be used to simulate supports provided by members such as adjoining structures.
Surface and point loads can be applied in all the three coordinate
directions.
is in the form of nodal displacements and
and bending moment and shear forces.

The output
stresses,

rotations, element

modified version of this code includes nonlinear characteristics for

foundation media.

427

INDEX

Advantages of finite element method, 85


Approximation functions, 39, 43, 44, 95 176
(see also Approximation models)
complete, 44
conformable, 44
requirements for selection, 43

Approximation models, 19, 43, 109, 173 338


(see also Approximation functions)
338
beam bending, 173
requirements, 173, 174
for time-dependent problems, 109
generalized coordinate, 39, 175
linear, 39-43, 95, 109
quadratic, 67
requirements for, 43
beam bending, 176
field problems, 95,304
column, 43
plane deformaions, 43, 339
bilinear,

Assemblage property matrix, 31


Assemblage vector of nodal forcing parameters,
31

Assembly, 55, 98, 116, 181, 182, 184


direct stiffness method, 58
of element equations, 55
(potential) energy approach, 55
Axisymmetric idealization, 335, 370
for footing, 370

for

torsion, 242

Area coordinate (see Local coordinates)


Assemblage equations, 30-33
building-foundation, 381
column, 58-62
field problems, 312
flow, 98
mass transport, 205
modified, 33,60, 61
overland flow, 216
plane stress, 349, 353
stiffness matrix, 58
torsion, 248, 258, 264, 279, 288
wave propagation, 227
Assemblage nodal vector, of unknowns, 31

Banded matrices, 412


Band width methods,

61-63, 396
Basis functions (see Interpolation functions)
Beam bending, 15, 172, 173, 364, 369, 398,

400,418
approximation models, 173
physical models for, 418-421
Beam-column, 172, 173, 189, 373

Boundary conditions,

27, 28, 31, 32, 59, 60-62,


99, 108, 117, 123, 207, 217, 231, 249-

279,301,317,325,395
categories of, 27

concept, 59
Dirichlet, 60
essential, 27, 31

explanation, 59-62
first,

60

forced, 27, 31, 60


geometric, 27, 31, 60
heat flow, 117
homogeneous, 27, 60
in consolidation, 123
in Galerkin's method, 74, 75, 180
in mass transport, 207, 208
in overland flow, 217-219

429

430

Index

Boundary conditions

(cont.):

in potential flow, 301, 302,

317

in seepage, 325

stream function approach, 317


in temperature problem, 1 17
in

258,264,279
wave propagation, 230-232

in torsion, 249,
in

mixed, 60
modification for, 32, 33, 60-62
natural, 31,60

Neumann, 60
nonzero, 61
types of, 60
zero-valued, 60

Bounds,

7,

11,83-85,261,320,321

lower, 11,84
symbolic representation, 84
upper, 1 1 84
Building-foundation systems, 372, 392
spring supports, 372, 388
,

Calculus of variations, 24
Collocation, method of, 26, 27, 47, 398-408
Comparisons: numerical predictions, and
closed form solutions, 65-67, 183-185,
252, 261, 267, 282, 283, 316-325
in column problem, 65-67
in beam bending, 183-185
in potential flow, 316-321
in seepage, 325
Compatibility, 7, 30, 43
interelement, 30, 43
for approximation functions, 43
Complementary energy, 24, 47, 76-79, 192,

271,295,296
for beam bending, 192
for one-dimensional problem, 76-79
in torsion, 269
modified expression for hybrid approach,

271
principle of, 47
stationary, 24

Completeness of approximation functions,


44,45,243
Pascal's triangle, 243
applications:

Computer codes,

Constant-strain-triangle (CST), 259


Constant-stress-triangle, {see Constant strain
triangle)

Constitutive laws, 23, 95, 110, 395 {see also


Stress-strain laws)
Constraints {see Boundary conditions)
Continuity, 4, 8, 30
conditions, 30
for approximation functions, 43

Convection, 202
Convergence, 7,

8, 10, 44, 87, 131, 186,

354,

364,418

comments on, 354


in beam bending, 364
in consolidation, 131
in heat flow, 131

mesh refinement, 87
monotonic, 44
one-dimensional column, 87, 88
physical models, 419
Coordinate systems:
element, 36
global, 36
local,

36

Coupled problems,

103, 121, 396

Damping, 234
Darcy'slaw, 15,24,45,95
Degree of consolidation, 124, 125
Degree of freedom, definition of, 21
Derived quantities, {see Secondary quantities)
Diffusion-convection, 202-210
Direct methods, for solution of equations,
410,413
Direct stiffness method, 58
Discretization,

1, 5, 13, 18,

22, 36, 173, 203,

213,225,246
Discretizations of "infinite" boundaries, 329
Displacement approach, 39, 240, 281, 282,

293,295,338,375-381
for torsion, 240
Displacement vector:

assemblage, 58
element, 54
nodal, 54
Dynamic problems, 224-236

beam-column, 424
building-foundation, 382, 427
field problems, 266, 328, 330, 425
flow problems, 102, 124,425
heat flow, 158-164

mass transport, 424


plane deformations, 354-368, 426
stress-deformation, 76
torsion, 266, 293-296
wave propagation, 424
philosophy of, 134
stages in, 134
Condensation {see Static condensation)

Conformable functions, 44
Consistent mass matrix {see Mass matrix)
Consolidation, 108, 113, 121-132
one-dimensional, 108, 121-130
layered media, 112, 113, 129
Constant-strain-line element, 46

Electromagnetic problems, 327, 332


Element coordinate system, 36, {see also Lo