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.
4JHBai_.
ELEMENTARY
FINITE ELEMENT
METHOD
CHANDRAKANT S. DESAI
Professor, Virginia Polytechnic Institute
and State
PRENTICEHALL,
INC., Englewood
Cliffs,
Desai,
Chandrakant
Elementary
finite
(Civil engineering
Bibliography;
in
Publication Data
S.
element method.
series)
p.
Includes index.
1.
TA347.F5D47
I.
Title.
7810389
624'. 171
ISBN 0132566362
Civil Engineering
N. M.
Newmark and W.
J.
Series
Hall, Editors
w.
may
No
part of this
book
in writing
10
987654321
PrenticeHall
PrenticeHall
PrenticeHall
PrenticeHall
PrenticeHall
PrenticeHall
Wellington,
New Zealand
2G3S8Q1
To
My
Parents,
Maya and
Sanjay
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
17
General Idea
References
34
17
METHOD
17
Contents
vi
3.
35
35
36
OneDimensional Problem
Interpolation Functions
38
41
Principle of
Minimum
Expansion of Terms
Integration
46
Potential Energy
47
52
53
Potential Energy
Direct Stiffness
Boundary Conditions 59
Strains and Stresses 64
Formulation by Galerkin's Method
Computer Implementation 76
Other Procedures for Formulation
76
76
References
55
67
4.
85
91
ONEDIMENSIONAL FLOW
Problems
93
102
Bibliography
102
42
43
132
[8, 9]
121
103
Contents
6.
Listing
134
144
BeamColumn
177
189
192
200
References
Bibliography
202
202
202
209
209
References
211
211
10.
172
172
Physical Models
9.
145
8.
133
134
Problems
7.
vii
213
222
224
Element Formulation
Problems 234
Finite
References
Bibliography
235
235
225
224
Contents
viii
11.
TORSION
237
Introduction
237
238
240
Form
Solutions
252
Approach 254
Review and Comments
Hybrid Approach 269
Mixed Approach 284
Stress
Static Condensation
Problems 293
References
12.
268
291
297
299
299
Flow
300
Electromagnetic Problems
327
332
Bibliography
13.
328
328
332
TWODIMENSIONAL STRESSDEFORMATION
ANALYSIS
Introduction
333
Plane Deformation
Finite
333
Element Formulation
Computer Code
Problems
References
368
371
354
338
333
Contents
14.
Computer Code
373
382
Transformation of Coordinates
Problems 392
15.
389
392
Bibliography
394
395
396
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.
and variational
calculus.
Some
It
may no
longer be
significant level
of
as special cases. It
is
arise
as a
This book
sufficiently
is
elementary so that
it
Its
approach
is
Preface
xii
essentially
enough so
broad
is
stressdeformation analysis, fluid and heat flow, overland flow, potential flow,
it.
of the method
and provides a distinct and
element method at an elementary
intrinsic nature
is
it
can
be used for the fresh graduate and the beginner with no prior exposure to
the finite element method.
The
will
The
text
written in such a
is
way
that
is
is
necessary.
differential calculus.
them
in the
The
first
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 35 cover onedimensional problems in stress
The fundamental
mon
method
fluids.
by showing the comcharacteristics of the formulation for these topics and by indicating the
generality of the
is
illustrated
computer code
in
is
It is
may
finite
element method
is
Preface
xiii
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 114 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 selfstudy.
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,
simultaneous equations.
Physical models can help significantly in the understanding of various
concepts of the method. Appendix 3 gives the descriptions of some physical
for
Preface
xiv
in the
could include a selected number of topics. For instance, a quarter course can
For a
16,
class interested in
is
influenced by a large
number of parameters.
and are
In order to understand
such a system,
possible
is
not
Chandrakant
S.
Desai
INTRODUCTION
BASIC CONCEPT
current form the finite element (FE)
its
engineers.
we
a.
we must make
words,
is
divided the matter of the universe into five interconnected basic essences
air,
by singing
^he number
[l].
fire,
and
Chapter 1
Introduction
<**JW
We
finite)
number of
composed of
its
own
stars,
planets,
on
is
and galaxies
composed of interconnected
When man
To compute
drew
and outside
size inside
to
a high degree of
made of
blocks or elements
accuracy.
In
(civil)
engineering
we
started buildings
building clue.
Chapter 1
Introduction
\_
IN!
_/
T_
.44
r
r
r
r
+
r
r
Block
or element
r
r
r
r~
When engineers surveyed tracts of land, the track was divided into
smaller
was surveyed individually [Fig. l4(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,
tract
For
stress analysis
l4(b).
civil
engineering clas
modern
scientific
thinking
Chapter 1
Introduction
recognizes,
all
The foregoing
many
activities
of
abstract
man
Chapter
Introduction
Closure error
Survey
of smaller
areas
(a)
^^5":
#
/>'
(b)
Figure 14 'Discretization' in surveying,
(a)
assemblage of a survey.
Figure 15 'Discretization' of engineering structure, (a) Actual structure, (b) 'Discretized' structure, (c) Idealized
(a)
(b)
onedimensional model.
(c)
PROCESS OF DISCRETIZATION
Discretization implies approximation of the real
We
use
potential,
compatibility,
minimum
residual,
and
error.
As we
although these
number of
figures
Subdivision
Zeno argued that space is finite and infinitely divisible and that for things
must have magnitudes. Figure l6(a) shows the concept of finite
to exist they
divisibility,
{From Ref.
[2],
Limited, London.)
(a)
(b) Infi
(b)
Chapter 1
Introduction
space If the earth were contained in space, what contained the space in turn
:
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
For
and continuous
(Fig. 17).
3
(b)
[2],
Limited, London.)
greater
circle.
Chapter 1
The
Introduction
idea
is
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
calculated.
an active
is
exhaustion
Figure 18 Convergence and bounds for approximate area of circle,
(a)
Polygons inside
circle, (b)
Polygons outside
circle, (c)
Conver
circle,
A(5) = 1.322
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)
is
18),
approach or converge to the area of the circle. Figure l8(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
:
11
Inside polygons
1.0
0.0
Number
10
11
12
13
of sides
(c)
Physical models:
board or
models to
plastic)
illustrate
Bounds
12
Chapter 1
Introduction
Parabola
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.
differ
error.
For
example, the areas (or perimeters) of the polygons inscribed in the circle (Fig.
18) are
always
less
circle,
(or perimeters) of the circumscribed polygons are always greater than the
number of
is
sides of the
The
polygons increases.
We
area as
A*
where A*
 A = e,
(11)
= error.
To
we need
to establish
states
F= ma
(l2a)
F=mp
d2b)
or
where
F=
force,
m=
mass, and a
body is
in
t.
= acceleration
The
principle
or second derivative of
is
in the
body
1(a)]
follows the
where v
= length
is
= ^>
of the column,
A =
(13)
AE
crosssectional area,
P = applied load, L
E = modulus of
and
elasticity.
13
14
Chapter
Introduction
= area of
cross section
4TP
tU^
(b)
(a)
Body
or structure.
failure.
When
effect
studying
finite
is
the cause
and
The foregoing
all
human
endeavors. Compre
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
some of the
::
Chapter
Introduction
will facilitate
Home
1.
Assignment
15
The following
are
many undergraduate
Define:
(a)
Stress at a point
(b) Strain
(c)
2.
Hooke's law
Define
(a) Principal stresses
and
(a)
strains
and
strains
sum
loads,
(b)
(civil)
Define
(a)
and
coefficient of thermal
expansion
5.
6.
wA EI ir*)= p  k w
where
w =
displacement,
spring constant,
coordinate along
E=
axis.
= support reaction, k =
= moment of inertia, and x =
modulus of
beam
>
elasticity,
1
Figure 112
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
2*i
4*!
and
3.
x and x 2
x
+
+
3* 2
5* 2
=
=
14,
10,
Define:
(a)
Total derivative
differentiation
REFERENCES
[1]
[2]
Books
Ltd.,
London, 1959.
Inc.,
Rathbone
THE FINITE
ELEMENT METHOD
STEPS IN
INTRODUCTION
Formulation and application of the
finite
consist of eight basic steps. These steps are stated in this chapter in a very
somewhat overwhelming.
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. 21) 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.
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)
head
(p.
(a) Discretization
We assume that
tional
it is
u,
temperature T, or fluid
finite
number of
of u by using conven
We
based on
body
is
elements are
is
divide the
now
easier to
distribution of u over
A
is
The subdivided
entire body and
it.
For stressdeformation
analysis of a
body
in equilibrium
under external
load relationship.
To
we make
principles governing the behavior of the body. Since our primary concern
to find the distribution of u,
terms of
u.
we
is
principles in
shape, or
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
qualitative statements.
Step
1.
This step involves subdividing the body into a suitable number of "small"
bodies, called finite elements.
The
and the
used.
it
may be
23).
On
many
other
if
is
approximate that
makes use of
this idea. It
may
Step
2.
Select Approximation
no great
difficulty.
Models or Functions
we choose
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
20
Steps
in the Finite
Element Method
Chapter 2
'
'
Onedimensional body
Line elements
(a)
unknowns
each node
at
Quadrilateral
Triangle
Twodimensional body
Quadrilateral and
triangular elements
(b)
Hexahedron element
Threedimensional body
(0
Figure 22 Different types of elements, (a) Onedimensional ele
ments,
(b)
elements.
Twodimensional
elements,
(c)
Threedimensional
Chapter 2
Steps
in the Finite
Element Method
21
body
Original
Discretized
body
finite
= N lUl + N u + N
Here u u u 2l u 3 ,
and u 2
N N
we
u3
+ Nm u m
Nm are the
(21)
, . . . ,
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. 22(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 twodimensional problem [Fig. 22(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
unknowns u
at
all
is,
at hand.
will
be
in
terms 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 24 Approximate solution as patchwork of solutions over
elements, (a) Assemblage, (b) Neighboring elements, (c) Section
along AA.
This
is
common boundaries.
is
Step
3.
is,
24).
Obviously,
computed solution
the error is a minimum.
is
we would
from
this
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
stressdefor
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
small,
is
given by
dv
y
where v
is
is
(22)
dy
direction.
in
one
(23)
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 25 Problems idealized as onedimensional, (a) Onedimensional stressdeformation, (b) Onedimensional flow.
we must
it
is
is
is
law
is
vital parts
The
stress
it is
defined to reflect precisely the behavior of the material or the system, the
results
illustration, consider
strain
ay
where a y
elasticity. If
= Eye
(24a)
y,
we
substitute e y
from Eq.
and
Ey =
Young's modulus of
we have the
~ dv
dy
(24b)
24
Steps
One of
in the Finite
Element Method
is
Chapter 2
= coefficient of permeability,
where k x
Step
= k x gx
4.
(24c)
vx
= velocity,
corresponding law
is
Ohm's
and gx
= gradient.
In
law.
By invoking
available laws
all
At
this stage
manner introduce
and
in a
somewhat
less
rigorous
ENERGY METHODS
These procedures are based on the idea of finding consistent
states
of
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
STATIONARY VALUE
maximum, minimum, or
Under certain conditions, the
function
may
simply assume a
find the
Chapter 2
Steps
in the Finite
Element Method
25
F(x)
Maximum
Neutral
Minimum
POTENTIAL ENERGY
In the case of stressdeformation analysis, the function
F is
often repre
Fto
and
minimum
we can
equilibrium,
it
To comply
potential energy.
is
np
and
assume
linear
will
potential energy.
The
potential energy
is
defined as the
sum of the
Therefore,
II P
When we
= U+
(26a)
minimum
potential energy,
U p and equate
it
to zero.
m
The symbol 3 denotes
= 8u
sw
we can
essentially
then
= o.
(26b)
we
We assume that
As
indicated
change as com
dW
Note
5W
(26c)
is
and
lost
Steps
26
The
Element Method
Chapter 2
minimum can be
of n^,
in the Finite
verified
is
S2
Up
is
is,
U P =3 US WP
2
>0.
(27)
For example,
np
where u ,u 2
1
3T1 P
unknowns
in terms of
which
it is
if
=np(u
19
u2 ,...,uj,
(28)
number of unknowns
then
implies
*L=o,
L=o,
du
2
(29)
Here n
total
number of unknowns.
we shall
In subsequent chapters
stationary energy
illustrate the
of various problems.
is
method of weighted
the
residuals
MWR,
and Galerkin's
MWR
problem. As a simple
equation
='<*>
<2
10a >
Chapter 2
where w*
unknown, x
the
is
is
Method
the coordinate,
27
the time,
t is
is
and f(x)
is
the
written as
Lu* =/,
(210b)
where
 dx
is
dt
We
approximate or
function u for u* as
trial
ft
cp
(211)
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. (211)
written as
is
u
in
which a
and
we
<p
are
left
(p Q
=*
If the
(212)
t <Pt
approximate solution u
is
R(x)=Luf
(213)
which
is
zero
In the
if
u*.
method of weighted
residuals, the
aim
to find an approximate
is
is
substituted
with a residual
is
made
as small as
R(x)W (x)dx
t
= 0,
1, 2,
n,
(214)
28
Method
Chapter 2
For
functions.
method,
W=
1.
The
expression in Eq. (214) implies that the weighted value of R(x) over the
domain of a
R(x) over
u*
w,
The residual
and the sum or the
minimized.
is
R(x)= Luf
(a)
As a simple
u over D.
secondorder
differ
tion in a
in
Chapters 3 and
d 2v
dy
where v*
the
is
unknown
4, respectively
(21 5a)
f,
(deformation), y
is
Assume
E is
that
the
= EA, /is
/=
EA =
specializes to
d 2 v*
dy
An
(215b)
10.
Eq. (211) 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
Method
29
given by
R(y)
According to the
MWR,
f*
W~
1
I0
(2 " 17a)
'
Riy)W {y)dy
)
= 0,
(217b)
1, 2, 3,
Jo
or
^(^10)w,(y)dy =
or
j^d^y w
(y)dy
0,
(217C)
(2 . 17d)
Here L
au
cc 2
and a 3
as
[[^^^]w
[[^^y
(y)dy
[ ^10]w
d
<x 2
At
this stage,
to proceed
in the
it is
from Eq.
and a 3
we
y)dy
i(
0,
= 0,
(y)dy
When
(218)
0,
these values of
a a 2 and a 3
,
are
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
(219)
where
and {Q}
=
=
= vector of nodal forces. Details of the matrices in Eq. (219) will be
Step
5.
Conditions
Our
final
aim
is
body
that define
fact, as will
body
it is
we view
be
is
the procedure in
Once
element,
we
1). It
is
we add them
is,
the neighboring points should remain in the neighborhood of each other after
the load
is
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
of the dis
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
we
first
deriva
tive
Finally,
we
matrix notation as
[Ktfr}
= {R},
(220)
BOUNDARY CONDITIONS
Until now we have considered
Equation (220)
[Fig. 29(a)].
tells
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. 29(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
perform
It is just like
displacements
is
conditions.
At
moment
zero
that
is,
is
is
flows.
is
boundary condition
proportional to the
first
is
specified as fluid or
derivative of fluid
head or tem
[K]{r}= {R}
(a)
[K](7}=
{R]
Constraints
(b)
Body with
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
Step
6.
Equation (221)
is
set
= {R}.
(221)
Unknowns
K
K
ll
rl
2l r t
+ K 12 r +
+ K22 r +
+K
+K
ln r n
2n rn
= Ru
= Ri>
(222)
K s + Km2 r +
n
+K
nn rn
=R
solution
problem
is
be the
fluid or velocity
Step
7.
will
be
head or potential.
strains, stresses,
and
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
Steps
34
in the Finite
Element Method
and
less
we can
Chapter 2
is
REFERENCES
[1]
Turner, M.
J.,
Deflection Analysis of
Sept. 1956.
[2]
[3]
Crandall,
[4]
Argyris,
S.
J.
New
in
York, 1956.
London, 1960.
[5]
[6]
Element Method,
Vah
Mechanic^"
York, 1972.
[7]
to the Finite
Finlayson,
Academic
in
Advances
Press,
in
12,
Academic
Press,
New
New
ONEDIMENSIONAL
STRESS DEFORMATION
203SQ01
INTRODUCTION
From
we shall consider engineering problems idealThe main motive for treating these simple problems
ized as onedimensional.
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
increasingly difficult.
many
As our
first
many
practical problems.
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
stiffness
EA lumped
finite
at the centerline
element method in
2.
35
36
Chapter 3
ic
s
Nodes
Finite
elements
S EA
zMx^
777Z
(0
(b)
(a)
Figure 31 Axially loaded column, (a) Actual column, (b) One
dimensional idealization,
Step
1.
(c)
Discretization.
We
We now
discretize the
column
into
The
At
it is
number of advantages
it is
its
derivatives
and integra
possible to obtain
tions
to using a
use, particularly
(as
all
we
deriva
shall see
it
far
Chapter 3
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
^^
77777,
L =
(yy 3 )/(//2)
(b)
(c)
Figure 32 Global and local coordinates, (a) Concept of global and
local coordinate systems, (b) Local coordinate
point
ever. Point
1. (c)
B is
accessible
from
all
3.
We
can
first
define the
B is x B
then
*a
its
= x* +
*ab
A is
(3la)
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,
38
much
is
in the vicinity,
Chapter 3
make
the definition
is
more
measurements.
there are a
number of ways
in
first
case,
we measure
31).
We
consider
(c).
is
Note
that
analogous to point
node
is
in Fig. 32(a).
y=y+y
We
in the element
coordinate
is
(3lb)
hence
=y yi
y
Often
it is
0 lc)
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\ _
yiyx
(3ld)
is
written as
L
The values of L range from
^
at point
to
(32)
at point 3 to
at point 2.
Chapter 3
An
39
is
nondimensionalized form, and their values are expressed as numbers and often
lie
It is this
subsequent derivations.
Step
2.
Select Approximation
for the
Unknown
Model
or Function
(Displacement)
method
it is
difficult to find
we
Generalized Coordinates
The
we can use
is
=a +
oc 2
(33a)
y,
or in matrix notation.
l
[1
y]\
<;)
(33b)
or
W=
where o^ and a 2
in the element,
y and y2 To show
x
 3c
(3
[+]{}.
this,
we
first
fory,
Vl
=,
+a y
v2
<%!
l9
a 2j 2
(34a)
(34b)
or in matrix notation,
N=P "IN
(340
= [A]{}.
(34d)
or
{q}
40
y3
Yi
Y2
'1
Chapter 3
(a)
V =
OLy
OL
(b)
N, =/ 2 //=1sorN 1 = 1(1
L)
Equal
A A
(e)
Here
{q}
[v x
v2]
is
we
solve
for {a} as
{a}
= [A]i{q},
where
(35a)
[A]
yi~
yi
_i
l/l
~
1
yi
(35b)
yi
y\ _i
yC
l
(35c)
Chapter 3
Here
= Jacobian
/
element.
=y y =
determinant
41
Thus
\[y
fai]
(2
y
(35d)
1
(2
1)
2)
(2
v2 \
x 1)
Therefore
(35e)
that a's are related to but are not explicit functions of nodal displacements.
This
is
Now we
nodal displacements:
v
y 2v
y,v 2
+*>
(
(36a)
,,
(Z ^)*.(ZL f 2 )^
=
fv,
tf,*,
+ fv =
l
(l
y.) Vl
(3
jv 2
+Nv
= s, N = y/l = s,
6b >
(36c)
(36d)
2i
"
v=[N,
N ]H
(37a)
= [N]{q},
where [N]
is
(37b)
N +N =
t
is
that their
sum
INTERPOLATION FUNCTIONS
Since our aim in the finite element analysis
is
42
tion functions
makes
it
Chapter 3
and
integra
it
is
pertains
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. 32(c)] described in Step 2 and in Eq. (32), and
t
express the v as
v=i(lL) Vl +i(l+L)v
= N v, + N v
2
(38a)
(38b)
= [tfi N ]l Vl \
(38c)
(38d)
Note
that
(i
1,
[N]{q].
2) in
(i
1,
2) in Eq.
An
where
T
{y]
[y
y2 ]
is
(39a)
(39b)
(39c)
is
can be given at this stage. A comparison of Eq. (38) and (39) 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,
This
is
is
we
shall subsequently
[1].
look at other
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
(310)
(311)
2]
nodes
2]
and
2, respectively.
simple words.
An
and (38)]
is
it
v,
up
to a degree required
lem involving
axial deformations,
it
is
is,
the
The displacement
tied in with the highest order of derivative in the energy function such as the
potential energy.
derivative dv/dy
For example,
= ,
is
in
44
up
to
(zero), that
is,
Chapter 3
displacement
v.
In general,
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
is
For
instance, in
relate
it
(motions) and constant states of strains (gradients), then the function can be
considered to be complete.
mode
For
rigid
it.
=a +
2
cc 2
a 3y 2
a 4 j> 3
we have chosen
n
cc n+
(33d)
iy
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
rigid
may
the
first
derivative or gradient
v.
exist in other
is,
Chapter 3
is
from Eq.
first
45
all
terms up to n
6,
to
It
may happen
not include
satisfy the
all
we
all
it
terms
automatically
are chosen
Then
up
is
includes
all
it
terms
3.
may
still
satisfying rigid
strain.
Step
3.
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 stressdeformation
problem,
4),
the relation
in later chapters
we
shall use
= ^>
ay
strain
(312a)
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
(32)
dv
,~
101 v
(3 " 12b)
= d_(l^M)=\
dy\
dy
dL
dL dv
ayTL'
we have
d_L
dv
= Ty =
>
(312c)
1/2
(38)
= ^[(1 
L)v t
+ i(l +
L)v 2 ]
1]N.
[1
(312d)
= }[
e,
1]H
(313a)
or
{,}
(1
where
[B]
(1//)
x [ 1
1)
{q}
2) (2
1)
is
{e y}
(313b)
1]
[B]
(1
is
We
shall
the displacement.
is
because
we have chosen
of
STRESSSTRAIN RELATION
For
simplicity,
we assume
column element
is
linearly
law,
ay
Ey y
(3 14a)
or in matrix notation,
{,}
(1
where
[C]
is
1)
[C]
(1
{,}
1)(1
(314b)
1)
Chapter 3
47
i\
A^
= U cl
u
III
now
allows us to express {a y}
in terms of {q} as
{y }
Step
4.
(315)
number of procedures
Among
= [C][B]{qJ.
potential
PRINCIPLE OF
In simple words,
is
in equilibrium
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 35 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.
(316)
48
Chapter 3
IP
w^%
Figure 35 Idealized linear spring.
done
SW
6Wr
(317)
we can consider
it
into
is
work by
these loads,
The negative
of external loads
is
lost
expressed as
sn B = su + 8wD =
susw =
(318)
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. 35), assuming undeformed state of the spring as the datum
for potentials, as
Pv
Ikv  Pv,
2
(319a)
where kv = force in the spring and %(kv)v denotes strain energy as the area
under the loaddisplacement curve (Fig. 35). 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
a positive v
is
5v 2
minimum by examining
lOv
assumed
v.
The
(3 19b)
10v.
results are
shown
in the following
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.
1.
deform by
On
Hence, under
1
known
is
when
minimum
in equilibrium,
derivative
v. It
unit.
on the function
P=
zero.
Applying
minimum
\kvbv
Pdv
well
SU,
fro"
Eq.
(3 19a),
we
its
obtain
=0
or
(kv
 P)Sv = 0.
(320a)
40
Is**
30
20
\
\ 10
\
3
1
^^" '2
Minimum
point
10
20
30
50
Since Sv
is
Chapter 3
kvP =
or
=P,
kv
(320b)
which is the equation of equilibrium for the spring. STl p = in Eq. (320) is
analogous to equating d\\ p \dv = 0, which will result in the same equilibrium
equation (320b). Substitution of the numerical values gives
\0v
Therefore, v
We
method
10.
same answer
as before.
Hp
is
a function of a large
ments. Consequently,
it
is
number of parameters
finite
element
or nodal displace
mathe
cumbersome and
often
impossible.
we
n> =
most of the
37)
and write FL
as
[5]
c e dv
iff ^ ' >
iff
fvdv
if
T'vds
S Pv
jp
U\
%
\
(32 la)
Chapter 3
where \o y e y
Ty =
area,
= part
M number
maximum
value of
vt
level,
M=
Y = body
Pa =
applied
displacement corresponding to
P a is now appropriate.
Pu can be treated as
Pih
Since a
is
force
2.
we
volume,
comment concerning
joint force
and
volume,
51
Pig
applied at point
a local
it is
The terms
in
now we
except that
assign
volume
(3 19a),
it
as a
spring.
For the present, we assume that the crosssectional area A of the element
is
 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
(321b)
(Fig. 37).
now
be expressed
in
dy
(322)
LdL.
Therefore,
1
11,
Next we
ay y
dL^C
YvdLlj TyvdLtPVi'
(lc)
J"'
substitute for v, y9
(38), (313),
and
(314),
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)
(321d)
v,
l)(l
1)
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
(32 le)
(1
1)
dL
fy dL
{q}
dL
{q}
(1
1)
1)(1
1)
a y e y = Eel m Eq. (32 lc). The need for transposing will become
when we expand the terms in Eq. (32 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
and expand
as follows:
First term
"*!> <r;j4<.
AE
4/
{v\
2v v 2
l
..{:)
v\)dL.
(323a)
Second term:
AIY
W
Here
is
assumed
fy
for uniform
[y(l
to be
L)v,
+ i(l +
L)v 2 ~\^dL.
(323b)
pi
Finally, the
'
sum of
J',
[t (1 "
L)Vl
(1
+ L >*\ dL
Chapter 3
l',[i
Notice that
2) to find its
minimum
+ L>
</L
C P,^.
if J
minimum
\\
i(l
Tyl
gig
L) Vl
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)
the
minimum of lip,
differential calculus
for
SU P =
where 6
is
We
(326)
0,
book
will
be
INTEGRATION
For the onedimensional problem, the integrations in Eq. (325) are simple;
in fact, the first term in these equations is independent of the coordinate L
aeV
41
(2,
2v 1 )dL
= ^(2t>, 
1
= <f(v, A
4\
2v 2 )L
4)
AIY
2
54
Chapter 3
^0>, ^fiVi
v2)
+v
MfTA P u =
)^^ P
2l
0,
(327a)
= 0.
(327b)
In matrix notation,
AE
1
or
= {Q}.
M{q}
Here
it is
[k]
stiffness
(328b)
tural analysis;
{Q}
it is composed of body
and the joint loads. It is
approximation
AIY
surface load.
as
lumped loading; we
shall consider
In the foregoing,
we
all
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. (321e),
.
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},
(328b)
Chapter 3
55
where
ai f
h=%
[B]
E[B]dL
and
(Q}
The terms
in
y J'
[NTF^L
+ JJ"
[Nff/L 
{P,}.
Eq. (328) are the same as in Eq. (325), except that here they
The transpose in [N] r arises because after
?.
on n^, the
a set of (linear) simultaneous equations [Eq. (328b)].
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
5.
to
As an example, we
38).
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. (324) 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
56
Global
Local
Nodes
Displacement
Nodes
Properties
Displacement
Chapter 3
" y
1
J
v2
An to
,
v?
v4
^fj"'
W  2v v
3
+ vl)dL
^f
[y
i>.
yO
+ L ^] rfL
y^ J'
[yd 
L) Vl
1(1
+ L)v^dL
y^
fir
J'
%^ J'
/',
 P!,, 
[t
~~
L)v
'
+ T (I + LKl rfL
[yd " iK + yd +
(PL
+ P*u)v  (P, +
2
),]<.
P\,)v 3
(329)
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
(330)
0,
5n p
(331a)
which denotes
cttPp
dn<p
dv 2
(33 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
(332a)
Ph + Pl
+
One Dimensional Stress Deformation
58
Chapter 3
or
[K]{r]
(332b)
{R},
METHOD
DIRECT STIFFNESS
tial
added together;
this
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
larly,
loads at the
are also
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/
(333a)
Local Global
\v\
i_ [vl
>
Local
>v 2
>v
_A
Global
r
\
2 l2
Y2
^}
(333b)
Local Global
h\
3
i_ [v 2
>v
vj
3)
>
_A
3 l3
Y n\
3
,
,
Ty3 l
K;:l
(333c)
Chapter 3
59
v\
is
the displacement
at
To assemble
order of
we note
and hence the assemblage matrix and load vector are of the
Consider Eq. (334) 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
==
(334)
<
blocks of 4
x 4
A 3 hY 3
Ty3h
A3I3Y3
+n
+ p32i
fy 3 l 3
and 4 X
elements [Eq. (333)] into the proper locations in Eq. (334). For instance,
the coefficient in Eq. (333b) for element 2 corresponding to the local indices
(2, 2) is
(3, 3)
loads corresponding to the local index 2 are added to the global location 3
and so on.
We
direct stiffness
approach
is,
is
is
the
same
as Eq. (332).
We
common
common
nodes.
BOUNDARY CONDITIONS
we
now we have
Until
when
the
column
straints.
As
the
name
displacement or
implies, a
its
gradient(s)
60
body.
boundary condition
we
space. After
how
us
tells
body (column)
the
Chapter 3
we have
is
supported in
a structure that
is
ready
Eq. (332)
forces
infinite
singular, that
is
number of
is,
its
determinant van
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
it
first
is
condition;
if
the base
is
and
if
the base
is
mixed condition.
in
called geometric or forced, while those in terms of gradients are often called
on the
boundary.
is,
at
illustration, let us
nodal 4
R.
\Vt
v2
>
^32
We
^34
K43
^44 _
row and
(335)
<
v3
"33
JV
iJ
23
33_
>r
<
as
'':
^'3,
Jtl
>
= R
>
<
,*3,
(336a)
Chapter 3
61
or
= {R}.
[K]{r]
(336b)
stiffness
is
sparsely populated.
that
diagonals, whereas other coefficients are zero. These properties, which occur
in
many
make
engineering problems,
economical. This
is
we
somewhat
different,
ku
fc,2
k 21
k 21
"
k 23
R>
v2
<
k 32
k 33
now be
v3
34
_
= *
(337a)
v4
3
c5
.
when
dealing
symmetry
of a matrix for reducing storage requirements, it becomes advisable and
necessary to restore the symmetry of the matrix in Eq. (337a) 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
(337b)
v3
Ij
0
<
k 34r x 6
v*.
side.
For instance,
0xt),+
k 32 v 2
k,*
x S
=R
or
+ k 32 v +
2
k 33 v 3
=R 3
k 34 S.
(337c)
62
Chapter 3
conditions usually do not need the special consideration required for the
the
an integrated sense
sum of
the
in the finite
unknown,
is
if specified
satisfied
automati
derivative at the
boundary vanishes
unknowns and
different categories
approximately.
Different problems involve different
we
6.
Equation (332)
set for the
is
shall discuss.
column problem
is
linear
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. (332) will be nonlinear. Although we shall
briefly discuss nonlinear
topic
is
The unknowns
assemblage vector
this
text.
{r}
Appendix
of the
2.
common
substitution
To
Example
We
[5, 7].
procedure,
we now
31
(Fig. 38):
Crosssectional 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 crosssection and material properties.
Substitution of these values into Eq. (328) leads to the following element and
assemblage equations
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.
(332)]
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
>
100
'
10
5
"
100v 2
100^ +
200v 2
100v 2
lOOvj
first
+0
+
=
=
100v 3
200^3
7.5,
(338a)
15.0,
(338b)
15.0.
(338c)
we
7.5
(338a) gives
100^!
Substitution for
100^
in
100v 2
7.5
100v 2
200^2
100^3
15
or
100v 2
Now, Eq.
22.5
100v 3
(338c) gives
22.5
I00v 3
200v 3
15
or
37.5
v3
100
cm.
100
cm
and
67.5
V\
100
cm.
Example 32
Instead of uniform
and
fyy
in this
is,
at
64
joint load
P lg
Chapter 3
lOOvj
100^!
100v 2
+

200v 2
100v 2
=10,
f
 100^3 =
+ 200v =
3
(339)
0,
0.
Note that the equations on the lefthand 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
;
Step
7.
vi
v3
=^cm
cm
3$q
v4
(specified)
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. (332). 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.
substituting the
Example
For element
31 (continued)
ey (D
iQ
[l
7.5
1000"
and
^=
3 are
22.5
1000
and
37
tyK
(67.5)
l]^o(
1000
60
Chapter 3
The negative
strains
OyV)
(3)
100Q
now
x 1000
X 1000
mi x
100
7.5 kg/cm 2
22.5,
37.5.
Example 32
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.
8.
Figure 39 shows plots of displacements and stresses for the two previous
examples. For Example 31, the distribution of displacements within each
element
is
linear,
need not be
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
oy
_4A =
^=
as
10 kg/cm 2
and
Vl
_ PQl) _
~ AE ~
10 x 30
1000 x 1
30
100
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
v2
1
1
1
i
v 4 = 0.00
Disf )lacements
Stresses
(b)
Example
Example
Ty The computed
.
32.
stresses are
an average
element computations
finite
is
consid
ered satisfactory.
For Example
32, the
numerical solution
is
tion,
because there
is
factors.
Chapter 3
67
puter cost.
it
for v as
= \L{L = [Ntfq},
X)v x
+ \UL +
\)v 2
(1
L
(340)
where
[N]
1)
{qf
[L(L
L(L+1) (1L 2 )]
and
[,
It is
In
fact,
we can choose
interpolation functions
node
in the
TV,
will
l = +1
model.
The
is left
to the reader as
38.
we applied
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
was only
to the entire
domain, and
to analyze element
by element.
was applied
we chose
To
first
68
As
dure
is
1,
Chapter 3
Another important aspect of this procedure is that the approximation function and the functions q> [Eq. (212)] 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. (214)], 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.
{
=V
(341)
JVjwJ,
k=\ j=\
Sum for
Sum
elements, k
=2
;=i
1,2,3:
N)v)

;=i
Njvj
N)v).
(342a)
j=\
1, 2,
m;
For the
N\v\
4
f
N\v\
N\v\
N\v\
(342b)
N\vl
N\v\.
(342c)
=v u
Chapter 3
K
N?
2 2 N k vk
v =
N?
69
N^
N?
N3
Vt
Vt
;<
/^
(b)
NJT 1
+ Ni
i1
(0
Figure 311 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.
(c)
i.
= v\
v 3 and v\
,
=v
we have
N\v A
(343)
A,
1;
for instance, in
70
y
yt
N (y)
 ytx =
NV,
(344a)
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.
\
/ii
Si,
where s
the
1,
and
/,
element
the
<s<
is
(344b)
y
?*+!
i.
adjoining elements
and
= crosssectional area
<*Z&
"Y
X*#
Figure 312 Equilibrium of column segment.
to use Galerkin's
column problem.
derive the governing differential equation for the column, Fig. 312.
sidering the equilibrium of forces for an elemental
method
We
first
By con
we have
Acy
or
dGy
~dy'~
f(y).
(345a)
Substitution
gives (after
AE,d
v*
dy
f(y).
(345b)
Chapter 3
Here v*
is
Pa
traction
PuW = Ptiyj)
where Puiyj)
71
x S(y
It is
 yj
(346a)
),
is
the Dirac
delta function,
&(yyj)
The
'
yj '
(3
(36))
is
R(v)
EA^ f.
(347)
46b )
residual for the differential equation (345b) (and the linear approxi
mation function
to
"
R with
respect
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. 38).
Integration by parts for the first term gives
,
(348)
[ % f *.
(349a)
is
<%>=l>^*4>
We
inserted
the
/,,
(349b)
9y
= cos =
column and
1.
/=
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.
(350)
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
dy
'
dy
dy
dN,
dN,
dN,
dy
dy
dy
dN,
dN,
dN,
dN,
dy
dy
dy
dy
dy
"
dy
dN,
.dy
dy
'
dy
dN,'
dy
<*>N,
dy
fN,
fN2
\dy
<Jy\
i\
dy.
+ EA\
^N
dy
(35 la)
fN,
fN,
or
ffi
It is
total length
f
in
(351b)
fR,
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
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
dN,
dy
dN
3
'
dN
dy
dN
'
dy
or
[K]{rj
'
dy
dN4
dy
dy
dy
dN<
dN4
dNt
dy
dy
dN,
dy _
(352a)
&
*> 5>
*>
<5>j
^
>j
j!
I> >!
5
73
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,
=
=
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
(353b)
By noting
that
dN = dN
dy
ds
>
dy
ds 2
we have
= 7T
#22
and
f(lXl)<fc+ f'ciXi)^
Jo
(353c)
Jo
finally
1
EA
[K]
1
2
1
1
(354a)
1
which is the same as [K] in Eq. (332) for constant element length and
form A and E.
The first term on the righthand side of Eq. (35 la) denotes
uni
>
ffNXdy
ryi
y
1
'fN{dy
{R;
fN\dy
n
\
Jy
2
fNldy
n
\
(354b)
fN\dy\
\"fN\dy\
which
is
the
same
by
T, in Eq. (332).
points
and
4.
For
instance,
at the
end
Chapter 3
dv
Ndy
since
N
N\
at
dy
nodes
\,x
and
dy
Hence
4.
15
it
reduces to
(
dv
dy
AEi
(354c)
m>
dv
at
nodes
and
4,
we have
(<))
R
(354d)
K)
dy/*
= AEe y = Ao y
{R.}
(354e)
F4
will
(351)] are
is
final results
when
natural boundary condition at a point will not have influence on the final
results if that point has a prescribed geometric
boundary condition.
we may
it
to be applied element
the interpolation functions for the element to have nonzero values within
76
Chapter 3
themselves and zero elsewhere. For instance, in Fig. 311, 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
a rather elementary
manner
problem
is
[8].
In
nonselfadjoint, the
two
results
can be
different.
COMPUTER IMPLEMENTATION
Although we could perform hand calculations for the onedimensional 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
DFT/C1DFE
in
we plan
Chapter
6.
to intro
This code
in
Examples
31
we
methods of formulation by
will
not be covered
1 1
to
a graduate
course.
U = U + Wpc
C
(355a)
Chapter 3
Graphical interpretation of Uc
shown
is
in Fig. 34.
11
we have
8U = 8U + 5Wpc =
C
or
= 5U  SW = 0,
<5n c
where
is
the complementary
is
lost into
(355b)
work of external
loads.
sign occurs
and displacements,
n =
<
W MW dv ~ If PFWdS,
T
* JIf
V
where [D]
is
S2
is
[C]"
1
,
boundary on which
of prescribed
A p o A Cydy  p T vdy.
n =
express
them
(355d)
unknowns and
(355c)
st
displacements, and
as
The minus
in
terms of nodal
we assume the
stresses, or joint
stresses
or nodal
to
where [NJ
is
[N.]{Q n },
The
(356a)
statically
is
the vector of
13].
where [NJ
is
= [NJ{Q
(356b)
n },
U =
c
T
i J]J {Q n } [KY[D]\N a ]{Q n }dV
V
J
{QJTNJW&
(357)
St
[f]{QJ
(q)
58a )
78
Chapter 3
where
[f ]
JJJ
[N.r[D][NJ<fo
(358b)
and
{q}
[f] is
r
JJtNJ {u}.
and
(358c)
of prescribed
displacements.
flexibility
stiffness matrix.
Example 33
Consider the column problem (Fig.
and
stress
37).
The
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)
approach.
For
F acting at node
1,
v2
0.
Under
these
= ^[lW
[S a ]{Q n }.
(359)
Chapter 3
79
P.J_ dy
[f]
A E A
J_
(360)
AE
The equation
is
given by
(361)
From
flexibility relation
[k]
[f]
[G][fr
AE
which
is
same matrix
the
[fFtGF
[owr
AEl
AE
AE
m.
(362)
Eq. (328).
[k] as in
junction with the concept of stress functions. This approach can be easier to
it is
we have
somewhat
in
difficult to
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
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,}.
(363)
St
S2
{q,}
and prescribed
dis
80
Chapter 3
The stationary value of IT* yields a set of equilibrium and stressdisplacement equations. In these equations both the displacements and the stresses
simultaneously appear as primary unknowns.
Residual Methods
We illustrate
column problem
assumed to be linear:
is
= #1  L)v +
=
Here [NJ
The
is
the
same
axial stress
where
[a
polation functions.
Variational
As in Eq.
+ L)v 2
(38d)
(38e)
[NJ{q}.
ay
r
{<T}
i(l
(Fig. 37).
a2
=
=
i(l
 L)a, +
i(l
L)o 2
[NJ{aJ,
is
Note
(364)
that here
and [NJ
is
[NJ
[N].
Method
Substitution of
v,
a y and
,
*[
{<fn}
T [NY~[N){<f}dL
(TV,
+ P2 v 2
(365)
).
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.
respect to
a u a2
({tf})
and v u v 2
({q})
and equate
^ = => y
[W\B)dL
{q}
[NFrNtfLto}
0,
(366a)
Chapter 3
0=>y['
[BRN]d!L{<Fj
{0}
81
(366b)
[P,,},
or
+ [k TU ]{q] = 0,
IKuYiQn) + {0} = {Q},
[k TT ]{<T}
(366c)
or
{0}
(367a)
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
(367b)
/I
^4
(367c)
2J
and
in
(367d)
13
For further illustration, consider the following properties of the column divided
two elements (Fig. 314):
A =
Pi
/
=
=
E=
Figure 314
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 =
82
Chapter 3
Substitution of these properties into Eqs. (367) gives the element equations as
Element
10
10
3000
6000
10
10
6000
____
3000
____
1"
<*2
>
<
or

(368)
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!
10
10
3000
6000
_J_
2
10
 6000
a,
__1_
10
10
6000
3000
2
1
(369)
v2
10
Vx
20
3000
10
6000
Ox
<7
[V 3 )
0/
= 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 32. 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. (369) contains zeros on the main diagonal.
For the standard Gaussian elimination, this can cause computational difficulties since the
Then
it
is
2).
[7],
which involves exchanges of rows and columns during the elimination procedure. Mathematically, the matrix [K] in Eq. (369)
is
positive semidefinite in
placement approach.
Galerkin's
With
Method
we used
[Eq. (345b)]. It
We can
* = * 
fy
(370a)
and
**
% + *'
we
for v
(3  ?0b)
and a y
as in Eqs. (38)
and
^
4'( dy
T^N dy =
(371a)
and
g + )#,</, =
<).
(371 b)
After proper integration by parts, Eqs. (37 la) and (37 lb) will lead to
the
same
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
As
procedures.
[9]
described in Appendix
1,
possible.
it
will
be
difficult to
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
84
tial
Chapter 3
respectively, lower
The approximate
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
it stiffer.
of complementary energy
flexibility
is
minimum
value.
The
that
is,
exact stiffness.
The
Physically, this
Bounds
in
finite
element
approach.
[K]
Number
Number
of nodes
of nodes
(a)
u => displacements
[f]
Number
Number
of nodes
(b)
of nodes
Chapter 3
85
Thus the two approaches bound the exact solution from below and from
[Figs. 31 5(a) and (b)]. This property is useful, and often we can bound
above
yield satis
useful to
list
finite ele
advantage
element.
is
no
is
that
we can
medium hence
geometries cause no
difficulty.
Any
included. There
arbitrary
and
irregular
accommodated
We
it is
codes.
PROBLEMS
31.
Figure 316
*~
1
r*7/4H
86
Solution:
32.
1, L(2)
if
Chapter 3
L(l)
3 )/(3//4),
modulus
N =
and
t
$, L(2)
3,
vary linearly as
E=
[{{\
L)
JO +L)}<
and
A=[\{\ L)
where
A and A 2 and
j and
^(1
+^)]{^}'
and
2, respectively.
Answer:
[k]
_l
"!]
given by
is
~0
6/
Mi
A,
E2
i
33.
of each element
in Fig. 317
J" 'A,
10^0
0^01
1
0_
E,
{
A2
E2
Ty
Length
strains,
Figure 317
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
87
Partial solution:
soor
1
Assemblage equations:
0~
7
15
8
8
16
(V
0.20
0.55
<
8
8
i'i
34.
0.29732 cm, v 2
8_
0.26875 cm, v 3
0.65
*>3
.V 4
0.40,
0.
kg/cm 2
at the base,
Hint: Consider
fy =
Then the load vector
L)
H(l
(1
+L)]{J")
is
af + fy
*K."+ *3
/
35.
For Example
and stresses.
31,
36.
in
37.
For Example
31,
stresses as the
number of nodes
Consider
yi
settles
by
0.1
4, 8, 16,
31
results,
6.
in accuracy, if any.
Partial solution:
See Fig. 318. Note that in view of the linear model and plane
is not significantly influenced by the
refinement of mesh.
38.
TV,
in
16
14
16
16
16
32.
14
[k]
39.
Derive element
= AE
stiffness for
"ST
varies linearly as
'A,
A=[#l L)
i(l
+^){ //}
if
the area
88
310.
By
Chapter 3
using the data of Example 31, solve for the displacements and stresses in
Prob. 38.
311.
312.
TV,, i
is
1, 2, 3,
or righthand node.
the
left
(a)
and quadratic v
in
i(l
 L)f +
yi
i(l
fy
given as
+ L)fyt
Eq. (340).
(a)
Convergence of
(c)
Convergence of displacements,
stress at midsection.
from
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)
90
Chapter 3
Solution:
iL{ {Q}
J "'
i(l
+ L)T]dL
Y=
Assume
L)T
+ 2Ty J
127V,
Compute
l)ki(l
(1*)J
(b)
Dl
= j\r UUL +
>
i(l
Answer:
= Ali2Y + Y2]
l
[Q]
(c)
if
= iL(L 
!)?
^L(L
1)7V 2
(1
 L*)f
Answer:
Tv + 2TV
4TV
313.
27V,
+ 2fw + 16fJ
and
stresses
column
in Fig. 319;
by using DFT/C1DFE.
L
E
=2000 kg/cm 2
Ty
= 2.0 kg/cm
= 10 kg
.0
kg/cm 3
y//////////////y
Figure 319
314.
Consider the coordinate system in Fig. 32(c) and derive the quadratic
inter
polation function model [Eq. (340)] from the following generalized coordi
a2x
a3 v2
[<J>]{a}.
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
[][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(Ll) iUL +
[N]
1)
*].
REFERENCES
[1]
Zienkiewicz, O.
F.
C,
C,
Dimensional Analysis,"
I,
[2]
[3]
and
J.,
in Finite
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.,
Desai, C.
S.,
and Abel,
Nostrand Reinhold,
[6]
[7]
[9]
Element Analy
J. F.,
New
Van
York, 1972.
[8]
in Finite
PrenticeHall,
L.,
An
New
York, 1965.
New
York, 1975.
PrenticeHall,
92
[10]
in Elasticity
and
Chapter 3
Plasticity,
Pergamon
Mechanics,"
in
Advances
in
Continuum
New
York, 1972.
[12]
Przemieniecki,
New
[13]
J. S.,
York, 1968.
Beaufait, F. W.,
N.J., 1960.
ONEDIMENSIONAL FLOW
Flow of heat or
fluid
through solids
is
a problem that
is
frequently
line.
The
be expressed as
/(*)
dx
#x),
(41)
two problems.
important and useful to note that Eq. (41) is similar to Eq. (345b)
that governs onedimensional deformation in a column. This indicates that
It is
the
phenomenon of deformations
in a
We
Chapter 4
94
(a)
(b)
(c)
Figure 41 Idealization for flow in pipe, (a) Flow through pipe,
(b)
Onedimensional idealization,
(c)
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.
we
all
we
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
finite
element formulation.
Step
As
Chapter
in
Step
We
1.
we
3,
2.
(p
fltj
cc 2
fluid potential as
(42a)
or
<p
where
=N
N = ^(1 L), N =
^(1
+N
l(Pl
2 <p 2
+ L),
and
q>
(42b)
[N]{<p],
is
{^}
[<p
p2
is
the vector of
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
Eq.
only up to order
The completeness requirement is
allows for rigid body motion and
equal to
= 0, that
is,
Eq.
Eq. (43a).
Step
3.
and Constitutive
The
gx
(42),
relation
Law
(p
and
(p
(3 12a);
hence
.
where gx
The
is
the gradient of
(a)
with respect to x.
(p
we can
use
is
42),
given by
vx
where v x
= velocity
= k x d = k xgx
in the
direction
(43b)
and gx
= hydraulic
gradient.
The
negative sign occurs because the velocity decreases with gradient (Fig. 42).
The value of
q>
in
Eq. (42)
with respect to x as
3?
= fx
? ({
" L)<Pl + ^ (1 +
L)<Pl]
'
(4_4a)
95
96
9x
Chapter 4
3y?
3^
= 411
dx
I
As
i]
*
\ fl
[B]{<p}
is
(44b)
[RPKfJ,
where [R]
case
it is
= matrix
of coefficients of permeability
It is
(45)
between
Eqs. (44b) and (313) and between Eqs. (45) and (315); the gradient and
velocity terms here are, respectively, analogous to the strain
in the stressdeformation
and
stress
terms
problem.
Step
4.
The
finite
it
may
finite
method.
we
of Galerkin's
Chapter 4
The
r^k
Q=A
where
97
is
(^fdx  pfydx,
(46)
and q
is
the
In Eq. (46)
we used
the symbol
to denote a
is
(46).
H=A
q>
Differentiation of
with respect to
f
<5n
= o=
<p
I""
and
^=
<p t
m]{<p n }dx.
(47)
yields
0,
(48)
1^ = 0,
or
P q[Wdx,
(49a)
or
M{<P}
where
[k]
{Q}.
(49b)
=A
("
>
[WkJWx
XI
i'J'
and {Q}
{WkAWL,
(49c)
h;
q[N] T dL.
(49d)
Evaluation of
[k]
and (Q
Substitution for [B] and [N] in Eqs. (49c) and (49d) and integrations
lead to
Ak.
[BY[B]dL
[k]
Ak
r[
wD
qi
{Q}
2 J_,
is
(410)
dL
li(l+L)J~
ql
(411)
[i
has replaced E. The lumping of half the applied flux at each node
is
kx
the result
of the linear approximation model, and for higherorder models, the forcing
vector can be different.
Step
As
as
ql/2
(412)
1
ql/2
<Pi
Assemble
5.
explained in Chapter
3,
we can
ment
is
for all
essentially
approach and is based on the physical requirecommon node between two elements are equal.
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]
(413b)
= assemblage
{r}
= assemblage
parameters.
Step
6.
98
To
assume
Example
41
=
=
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. (412) to find the element equations. For instance, for element
number
we have
1
icm:
10
1
1
1
9i
cp 2
>
1
The boundary conditions
set
1 1
k 44
=
=
<
<?3
<Pt
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
"^
10" =
"
10
<P*
or
<Pi
but
(p
+2<Pi
 "?3
2; therefore
2<p 2
From
(p A
(p 3
2.
(i)
tp 2
but
 = 0,
2(p 2  <?4
Therefore,
<Pi
2(Pz
1.
(ii)
99
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,
in v xl denotes element
Step
8.
We can
see that
the finite element computations give exact solutions both for fluid heads and
velocities.
This
is
Chapter 4
101
cm
1 0.0333 cm/sec
(b)
finite
element computations,
(a)
Fluid heads,
(b) Velocity.
is
linear.
k x %
As
indicated in Chapter
3,
is
 q.
(414)
The details of derivation are almost identical to those for the column problem
in Chapter 3. The final results for the threeelement system (Fig. 41) will be
>
1
1
dx).
\<Pl
1
<Pl
(
Lid
2
2
>
1
1
\
[K]{r]
= {R r +
}
{RJ
(41 5a)
<
= {R}.
(*%1
(415b)
PROBLEMS
41.
42.
43.
cm
Example
41.
= 1.0,
= 0,
kx
<7
/=
and temperatures
10 cm,
at the ends,
T(0)
7(30 cm)
=
=
100 degrees,
degrees.
BIBLIOGRAPHY
Desai, C.
S.,
Drain Div.,
Desai, C.
S.,
ments
wicz, O.
Desai, C.
"Finite Element
in Fluids,
S.,
C,
Vol.
eds.),
"Seepage
in
for
Flow
in
Wiley,
Engineering (Desai, C.
Methods
1973.
New
Porous Media,"
J.
T., Taylor,
in Finite Ele
C, and
Zienkie
8.
Pinder, G.
F.,
Finite
Remson,
I.,
102
Element Simulation
York, 1977.
in
New
New
in
Subsur
5
ONEDIMENSIONAL
TIMEDEPENDENT FLOW
Introduction to Uncoupled
In this chapter
we
treat
the case
is
when
problems
them
in the finite
known,
it
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
effects.
UNCOUPLED CASE
As an
illustration, let us
known
,.
where
a'
f*
The
stressstrain relation
ay
is
known change
column
in tem
[Fig. 5 1(a)],
= *'T,
(51)
of a
effect
strain e yo in the
is
denoted by
given by
= *, 
(52)
= Ee = E(e, yn
e yo ).
(53)
103
104
Chapter 5
EA
(a)
1000 kg
10cm
Area = 1.0 cm 2
a'
v,
=0
10cm
E =
10
2x 10 6 kg/cm 2
cm
(b)
and
mesh
Chapter 5
For deriving
Tl p
= A
element equations,
finite
a,dy
same
as before.
U'
(5 4)
Jyi
Jyi
to
(321)], as
Jyi
first
we need
105
The
(53) as
A C y2
=4
=U +U
1
last
Ee y yo dy
A C
Jyi
Jyi
We
yz
Ee)A\
y2
Ectfy.
Jy
(55a)
term of Eq.
(55a), since,
when
Up
is
being a constant,
differentiated.
it
will
not
Thus, compared to
new term
that enters
is
the
second term in Eq. (55a), which after substitution for e y from Eq. (313) can
be expanded as
U2 =
Ap[Vl
v2
]^^Ee
yo
dy
A{qf
J"
T
[B] [C]{eyo }dy
Jyi
\L)\
aUVi
(v 2 L)
= AEe (v + v
U2 is the part of the strain energy relevant to e
yo
Here
same as
loads.
(55b)
2 ).
yo
strain
x
is
the
and v 2 gives
Mi =
AEey Xl),
(56a)
^=
AEeJY),
(56b)
or in matrix notation,
=AOy
II
!*
106
Chapter 5
In general terms,
{Q
Finally, if
=A
(57b)
j yi
we add {Q
equations are
[k]{q}
{Q}
{Q
(58)
},
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
righthand side of Eq. (58) and does not contribute to the lefthand side,
which contains the unknown displacements.
Initial Stress
As can be
can act as
For
dam; such
pressures below a
Residual Stresses
is
strain or stress,
which can be
called
by a
common
fit,
and other
number of techniques
Example
51
The
details are
shown
and
of
initial
is
change (increase)
stresses for
effects
= aT =
yo
0.0000065 x 100
= 0.00065 cm/cm.
Eyo
2 x 10 6 x 0.00065
1300 kg/cm 2
Chapter 5
and the
initial
is
{Q
The assembled
initial
x 1300
{~ J}
1
1
{Rol
13(
1
is
1300
1
<
1300,
By
107
vector,
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. 51) we have measured y positive in the upward direction.
The boundary condition v = is now introduced, and the solution of the resultx
v2
=0.0015,
v3
V4
The
total strains
due to the
0.000
cm
(specified),
= 0.0030,
= 0.0045.
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
load vector.
It
108
Chapter 5
situation.
may
change
TIMEDEPENDENT PROBLEMS
As
[1],
dT*
d 2 T*
/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];
(59c)
v
dt
'
here/?*
= (excess)
is
m =
cm
/g).
initial conditions,
(i)
and boundary
(iii)
We
is
= f *(*),
(510)
conditions,
(ii)
where h
T*(x, 0)
T*(0,t)
T*(h,t)
= f*(t),
= f (h),
h
t>0,
(51 la)
/>0,
(51 lb)
occur as
initial
initial
or starting condi
Chapter 5
Element
109
(a)
tAt
+ At
(b)
Time
variation of tem
first
derivative.
We
initial
illustrate
1.
2.
is
= [N(x)]{T },
n
(512)
10
where
{T}
= [T T
Here
2 ],
respectively.
Chapter 5
and
T2
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 onedimensional flow.
Step
3.
The following
heat flow and
is
q'
where
q'
= rate
or
Step
(513)
W/mK), and A
= area normal
to
= thermal
x
direction
).
4.
First
= kA 6
conductivity (Btu/hrftF or
(ft
we
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 stressdeformation problems.
By
we have
g = ^[i(ii)r
=
+i(i+L)r2
(515a)
[B]{T}
and
J= t
= [#! 
L)T,
+ ^(1 + L)T2
dT,
dt
[N]
\dT2
"fcl
dt
[N]{f} ;
(515b)
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. (515a) and (515b) into Eq. (514) yield
x
A_l_
2 2
FY[BY<x[B]{T n }dL
+ ^^{t} T lWmTn}dL
(516)
(517a)
1
1
</L
V4 2.
#1#2
</L
#i# 2
f J"
(JV.7,
+N T )dL.
2
(5 17b)
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.
(518)
this 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
112
approach mathematically
dQ.
Ace
WtTxlF'j:
+ 4[j
dQ
_
~ 2l
At*.
dT,
differentiation yields
2T >~ 2TJ dL
+ NyNjjdL  St
(Nit,
dL
= 0,
(519a)
J"
(2T
The
less rigorous.
Chapter 5
+ 2T2 )dL
+ 4[^(N N2 T +Nlt
1
)dL
= 0.
,dL
#J>
(519b)
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 )
(519d)
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=
(520a)
o.
(520b)
r
_l
i_
"2
1"
kr6 .1
2_
f^il
^/
ft
* r
>l
7
_W
_?'
2
(52 la)
1
or
LKKTJ
[kjffj
= {Q(r)}.
(52 lb)
Layered Media
In the case of layered media, the formulation based on Eq. (59b) or
(59c) will lead to the following element equations
Chapter 5
i
r2
a,
1 2
afo
113
(521c)
where A, = ^4/:// and A 2 = Apcl/6 or ^j = Ak/yJ and A 2 = AmJ/6, corresponding to Eq. (59b) for the thermal problem and Eq. (59c) 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. (521a) and (521c), 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. (521) can be expressed as
[K]
A P[Bfa[B]<fr,
(522a)
J XI
[k,]
{Q(t)}
=A
=
<j
\" [W[N]dx,
(522b)
I" [Wdx.
(522c)
J Xl
Solution in
Time
problems
because of the appearance of the second term on the lefthand side of Eq.
(521).
Up
using the
to Eq. (521)
direction, x.
Now we
respect to time.
The
first
[Fig. 52(b)]
where At = time increment. Equation (523) essentially gives the slope of the
chord joining points A and B as an approximation to the continuous derivative dT/dt. By using Eq. (523), we can now write approximations to the time
derivative at the two nodes of an element [Fig. 52(b)]
^r.(' + ^)r.(O
dT\
The time
We
_T +
2 (t
At)
T
2 (t)
is
(5.24a)
(524b)
[7].
use this simple scheme only for the sake of introduction. In fact, a
114
Chapter 5
or
(w + iw)
;i;:3**^w
<>
or
>#:s*
where
= ft.] +
[k]
{0}
{Q(r
^ft]
A/)}
+ [k,]{
At any time
at time
= 0.
The
Tat
known.
Step
4.
We now
Derivation by Galerkin's
Method
consider formulation of
finite
tFt?*
(5  27a >
or
a
(
fe) r * = 
(5  27b)
or
LT*
= 0,
(527c)
= a 0w
T,
(5 " 28)
Chapter 5
On
Assuming
3,
linear
we
115
we have
as in Eq. (512),
r = i(iL)r (0
I
+ i(i+L)r (0
2
= t NT
(529)
t.
peratures
are functions of x,
TV,
We
t,
Eq. (529) gives variation of the temperature along the length of the element.
At
time dependence
only, 7^(0,
Now
T2 (t)
is
[Fig. 52(a)].
3 dx
a3
** &<
first
term leads to
NA
a 3
dx
'L
f
3dt
N dx =
(530b)
or
x
\dTdN
dT M
N dx
dT
= *<Z N
AT
"dxlti^
+L
T=
I
Substitution of
X2
,
NT
t
<
(530c)
<
l,2,y=l,2.
(53 la)
dN1 dN2
dl^^dNi
dx
dx
dx
dN2 dN2
dx
dx
3.
dT,
<Mj.'
dx
Ty
dx
dx
dt
N,N2
>dx
dT2
N
{
dt
\..dT
dx
(53 lb)
dT
dx
as in Eq. (521).
116
Because
xu
N =
x
at
x 2 and
at x,
and A^
Chapter 5
and
at x,
7V2
at
<dT
0i
i(,i
(532)
Step
5.
As an
illustration, consider a
is
L = 1
L = +1
and cross
section.
nodes are
equations as follows
"11
Ace
I
~2
TV
T2
0"
21
01 21 T
01
J
1
>
6At
1_
"2
R*
*2
>
<
6At
_0
t+At
t;
T2
T
2_
0"
_0
t+At
0~
2_
r1
T
(533a)
or in matrix notation,
+ At
= Wt+
+k%
{r}
r}
(533b)
Chapter 5
assemblage nodal
unknown
(temperature) and
vectors, respectively,
117
known
forcing function
Note that we can take out a// and l/6At as common if the
assumed to be homogeneous; then Eq. (533b) can be written as
in Eq, (533a).
material
is
Ar[K*]{r}, +A
where AT =
to be unity)
ccAt/l
(533c)
is
Btu
pc
ft
lbmF
Btu
hrftFlbm
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.(533)
is
7[KJ]{r} 0+Af
The
initial
Now we
^[K*]{r} 0+Af
{R} 0+/
6At
[KJ{r}
(534)
= 0.
=r,(Q,
r =r
4 (/z,
*)
= *i.
'>o,
=S
t>
4>
(535)
o.
A procedure
0"
"4
I
sym.
o
T2
i
T,
sym.
r4
!
f
71
r,
4
Af
+
3
"*"
T,
(536)
_0
_i
^*
_j_ _f_o
"*"
To
li
3
\2
118
It is
Chapter 5
known temperatures
corresponding to the
at
= 0.
now be
setting
KAA
J = 2, 4,
2, 3,
J =
*n =1,
1,
Ku=0,
K4J = 0,
(537)
1,
R =5 U
t
=S
R*
A9
which leads to
0"
'
i
I
I.
Note
its
'Ti
s,
T2
T
Tt
(538)
r.'
A,
A/
it is
economical and
It
Chapter
in
3,
by modifying further
= 1,2,3,
Kn = 0,
Kj, = 0,
and the righthand term
0"
o
01
f
1
Step
6.
(539a)
as
"1
2, 3, 4,
[TV
8.
\T2
To
\t3
r.
k.
At
+
+
<5,
(539b)
i*
<5 4
Unknowns
Example 52
To
illustrate solution
of Eq. (539b),
let
T 5u and
,
S 4 as follows:
f =
Si
S4
=
=
0.0,
10degrees
20 degrees/
f>Q
(540)
Chapter 5
Then Eq.
first
time increment
(t
0"
119
At) becomes
10
5_0
(541 a)
>
1
"I
oo
I
[T4t
2oJo
Therefore,
!^2
S'3
'
7Y
00
3
'
or
 5T =
5T + \6T =
\6T2
2
50,
100,
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.
1
1
0"
Tl
\
i
_5
T2
T
At
fi
(/
[T4
At
on
Lo
2 At
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
(541 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
(541 c)
148.01/6
20
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,
7.
Compute
The derived quantities can be rate of flow and quantity of flow. For
example, from Eq. (513) rate of heat flow q' for an element can be written as
kA[B]{T}
= =^[i
Since
we know T and T2
x
(542a)
!]{}
T =
x
10 and
is
straight
T2 =
5.63.
Therefore
g'(2)
= =^(10
=
+5.63)
4.37^ Btu/unit
time.
(542b)
ONEDIMENSIONAL CONSOLIDATION
Consolidation
the
is
[3, 8, 9]
phenomenon which describes timedependent deformamedium such as soil under applied (external)
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 stressstrain behavior is expressed through an
loading.
effective stress
concept given by
a=o'+p,
(543)
= total applied stress (Fig. 54), a' = effective stress carried by the
skeleton, and p = pore fluid (water) pressure. The stressstrain or
where a
soil
[2]
Aa'
= Ae,
(544)
av
where a v
proportional to axial
or vertical strain
be expressed as
*&
where
cv
<5
45 >
permeability,
eQ
The
is
Darcy's
law:
v=* dP.
Equation (545)
is
(546)
all
steps of
We
an
illustrative
Assume
Length of element,
Coefficient of consolidation,
cv
= 10 cm;
= cm /sec;
= g/cm
1
121
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)
Onedimensional idealization,
(b)
(c)
Boundary conditions,
(d)
Discretization.
By using Eq.
"
1
10
(52 la),
we
r
[2
11
1
i_
{;]+? _1
2_
QM
1),
(547)
or
[k.]{p}
( +
[k,]{p}
= {Q(0}.
(5  48)
Chapter 5
123
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
(549)
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<
54) as
h
,t<0,
t>0,
(550a)
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
dp
(2H,
By
is
is,
the second or
0,
>
0.
base,
It is
Neumann
and then
(550b)
In the
finite
COMPUTER CODE
The code DFT/C1DFE mentioned
At
may
study the portion of the code relevant to these problems. In the following are
described
some
results
Example 53
Figure 55 shows a onedimensional 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
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
is
along
shown
in Fig. 56(b).
Thus,
sional
it
is
possible to
compute
of heat flow can be found at any point in the bar and at any time
level,
giving the
<
The timedependent
temperature (Fig.
124
we can assume
that p(0, t)
=p(2H,
t)
0.0 at
all
times.
56). Often, in
geomechanics,
it is
Chapter 5
125
A"
A
consolidation
U as
[2]
U=
\7
pdy
hM* e
Pody
where p
M=
is
(7T/2)(2/
Tv =
(551)
cv t/H 2
and
1).
Figure 57(a) shows the distribution of U versus time factor Tv Figure 57(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 onedimensional and subjected to (vertical) structural loads. The procedure can yield history
of settlements and pore water pressures with time and the final settlements under
.
Once
From
a\
from Eq.
(544).
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,
t.
(552)
(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)
0.25
0.50
0.75
1.0
(b)
Time
factor vs. degree of consolidation, (b) Pore water pressure vs. depth.
128
Chapter 5
It is
129
An
Example
54.
Media
shown
[10].
any convenient units can be assigned to the dimensions and to the properties. Program DFT/C1DFE was used to solve the problem.
Figure 58(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
Figure
58
Consolidation
layered system
[10]. (b)
in
system,
layered
(a)
Details
of
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
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
130
Chapter 5
0.6
0.8
1.0
(b)
unknowns.
As
it is
easy to find
PROBLEMS
51.
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
.
Chapter 5
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
Solve Example
52 for
L)a 2
initial
7*(0, 0)
T(h/2, 0)
T(h, 0)
54.
131
=
=
=
10,
20,
10.
Derive equations for the three elements in Fig. 53 with the following properties
Element
k, p, c;
Element
2:
2k, p, c;
Element
3k, p,
c.
<x<
= 0.0,
t) = 10,
t) = 20,
T(x, 0)
7X0,
T(h,
Assume
that
ment and
55.
A u &i and A 2
ct 2
t>
t>
h,
< 0,
0,
0.
ele
Consider gradually refined mesh for both space and time, and by using DFT/
C1DFE, study the behavior of the numerical solution for onedimensional
heat flow (or consolidation); assume the boundary conditions in Example
53.
Use
2, 4, 8, 16,
and
0.5
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].
56.
57.
By using DFT/C1DFE,
solve
Example
53
=0.1
unit/unit length.
132
Chapter 5
58.
By
using
DFT/C1DFE,
dary, that
solve
Example
53 for
is,
Ty^H,t)=0.
59.
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
REFERENCES
[1]
Carslaw, H.
S.,
and Jaeger,
J.
C,
Conduction of Heat
in Solids,
Clarendon
[3]
S.,
B., Soil
Mechanics
in
for Analysis
of OneDimensional 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,
R.,
Interscience,
[6]
communication.
New
[7]
Crandall,
[8]
Desai, C.
S. H.,
to the Finite
New
York, 1956.
and Johnson, L. D., "Evaluation of Two Finite Element FormuOneDimensional Consolidation," Comput. Struct., Vol. 2, 1972,
S.,
lations for
pp. 469486.
[9]
Desai, C.
S.,
and Johnson,
L. D., "Evaluation of
Num. Methods
Eng., Vol.
7,
Boulder, 1971.
[11]
Desai, C.
S.,
and Saxena,
S.
Num.
1,
No.
1,
Some
VPI
&
S.,
"OneDimensional Consolidation by
May
1976.
the reader
element method.
3, 4,
and
must have
is
and
deformation [Eq.
(59)]
(345)], steady
flow [Eq.
medium
these
(41)],
in similar
is
The
effect
of a
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.
same program
Flow,
Temperature/Consolidation,
DFT/C1DFE.
listing
provided at the end of the chapter. In the following, details of various stages
of the code are described.
133
PHILOSOPHY OF CODES
same time,
at the
should be
it
is
it
without
code and
in the following.
STAGES
The explanation of
number of
stages,
which
Input Quantities
1.
common
title,
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
(NTIME). The
linear
levels at
many
that
is,
as
(NOPT)
approximation used
which output
number of
is
materials)
or timedependent flow; Table 61 shows the meaning of the terms for the
weight).
water
input in place of
is
RO.
Input Set 2. This set gives the data on the nodal points. Here
M denotes
node.
node
point.
^or
134
VLY
boundary condition
it is
Chapter 6
TABLE
135
61
Terms
AMV
RO
DENS
Remarks
1.0
1.0
Eq. (314)
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. (43b)
Timedependent
Temperature
Homogeneous
Nonhomogeneous
1.0
1.0
Eq. (59a)
ki
Ci
Pi
yi
Eq. (59b)
Eq. (59a)
yi
Eq. (59c)
Consolidation
Homogeneous
Nonhomogeneous
*i
1,2,...,//
Cy
1.0
1.0
ywi
number of
vi
different materials.
KODE.
If
KODE = 0,
the node
is
"free,"
and we
can input
then
gradient or slope
zero,
is
Within input
nodal data. This
set 2,
is
all
if
we
there
input
is
KODE = 0.
i.e.,
we have made
applicable only
is,
if
and
if
the
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
subsequent computations.
Input Set 3.
An
The
total
number of elements
NEL
is
set
equal to
NNP
1.
As
is
identified
if
(p. 140), it is
136
Chapter 6
input. If
IAREA =
2,
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. 62, Example 62, 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. (328). 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.
(TOTIM)
first
specifying uniform,
for
linear,
or arbitrary
In the second subset, input the specific values of time levels TIM(I) at
which output
The
is
desired; the
Here, according
INOPT, data
INOPT =
1,
input
= 2,
first
and the
last
INOPT =
3,
input
2.
Initialize
NCT
initialized here.
TIME =
0.0
and
NCT =
are
denotes the number of time steps and can be printed out at each
is
set
initial
conditions. Matrices
and
(533b)]
AK(I,J)
is
its
value
is
Matrix
is
set
Chapter 6
starting a
new time
step.
For the
stress
137
is
required. AP(I,J)
the
is
3.
Element matrix
computed
for
all
[k]
flux are
computed according
to Eq. (328)
At
4.
Assemble
this stage,
we add
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.
Stage
5.
to
Concentrated Forces
The concentrated
forces
in vector {R}.
Stage
6.
Boundary Conditions
matrices and load vectors [Eqs. (332), (413), and (533)]. In view of the fact
that
special logic
is
used to
Stage
7.
Time Integration
4)
Stage
The
8.
final
Solve Equations
tion procedure.
For
NOPT =
or
2,
this is
138
NOPT = 3 and 4,
steps = TOTIM/DT.
it
is
Chapter 6
is
This
Set {R}
9.
is
{H}
{R} f+Af
NOPT =
and
4.
Output Quantities
Stage 10.
out as
follows
NOPT =
=2
=3
=4
results for
problem,
problem,
problem.
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. (332), (413) or (533)].
ALL =
AP =
AREAEL =
DENS =
DT =
GWT =
H=
=
=
=
2 input
Time increment
[Eqs. (523)
and
(533)].
IAREA =
computed
at
initial
/
A/ time
2: for linearly varying areas; input values for only the first
3: for arbitrary variation, that
is,
and
last
elements.
Chapter 6
IBAND =
139
= 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)
pressure.
temperature,
consolidation.
layered
QK =
QP =
R=
(521).
(533).
Computed
this
RO =
TF =
TIM =
TIME =
TITLE =
TUINIT =
TY =
UAV =
Mass
Time
density of material.
factor.
is
desired.
Elapsed time.
Title
Total of
at a given time.
UINIT =
Values of
USUM =
VLY =
initial
Y=
it
0.
KODE =
0,
it
implies free
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
Chapter 6
Note
1: (a) If the
143
for the
and the
first
last
is
nodes,
(b)
KODE = 0:
Node
is
free,
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
PROP, AMV,
Example
first
61
stresses. Plot
the results.
10 kg
E
u
E~
A =0
element)
Vl
.it
\/h/////,
Figure 61
Example 62
See Fig. 62.
stresses.
plot
displacements and
"
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 62
Example 63
(p
10cm + 10cm
,
'
= 100.0
cm 2
find
+
10cm
k x = 1.0 cm/sec,
q = 0.0
Figure 63
Example 64
at a
/)
=
1
T(10,
.
/)
0,
is
100F.
is, 1 1
nodes,
> 0.
10 units
T(x, 0)
Figure 64
PROBLEMS
See the problems in Chapters 35.
LISTING
DFT/C1DFE
C
C
C
C
Q
C
C
C
20
MAIN
FINITE ELEMENT C3DE FOh ONEDIMENSIONAL DEFORMATION. ^LOW,
MA I
TEMPERATURE AND CONSOLIDATION
MAIN
OFT/CiDFC
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), TIM20I 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 (MN) 50,80,60
MAIN
CONTINUE
MAIN
WRITE (6,1030) M
MAIN
GO TO 910
MAIN
AUTOMATIC GENERATION OF NCDAL POINT OATA * *
MAIN
DF=M+1N
MAIN
RY=(Y(M)Y(N1) )/DF
MAIN
KGDE(N) =
MMN
Y(N)=Y(Nl)+RY
MAIN
VLY(N)=0.0
MAIN
IF (N.EQ.l) GO TO 100
MAIN
****
**
COMPUTE
MAIN
ELEMENT LENGTH
ALL(Nl)=Y(N)Y(Nl)
^AIN
MAIN
CONTINUE
WRITE (6,1040) N,K0DE(N) ,Y(N),VLY(N)
MAIN
N=N+1
MAIN
IF (MN) 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=NNPl
MAIN
N=0
MAIN
REAO (5,963) M , I E M,
MAIN
, 1= 1 , 3
MAIN
N=N+1
MAIN
IF (MN) 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(Nl,3)
(MN) 173,170,130
(NELN) 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(Nl,l)+i
I
160
170
180
))
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=NEL1
(
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=NNPl
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
\!
660
670
680
690
700
713
720
730
740
146
I )
Listing
370
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
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
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+Mi
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
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=NNP1
NR=NNP
HAIN3123
*MN3130
DO 670 N=l,NRS
MAIN3140
M=N1
MAIN3150
MR=MIN0(IBAN0,NRM)
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=N1
MAIN3260
MR = MIN0( IBAND,NRM)
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=NRI
MMN3350
M=N1
MAIN3360
MR=MIN0( IBANUNRM)
MAIN3370
,
149
Listing
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 STRESSDEFORMATION 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(L1,3)
MAIN3620
STRFSS = (P(L)R(L1) *PROP( MT )/ ALL L 1)
MAIN3630
L1=L1
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(Ll,3)
MAIN3760
VEL0=(R(L)R(L1))*PR0P(MT)/ALL(L1)
MAIN377D
VELO=VELO
MAIN3780
Li=Li
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(TIMETIM(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 = MK
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
00 850 I=1,NTIME
DIF=ABS(TIMETIM(I)I
IF (OIF.LT.TOLERI GO TO 860
CONTINUE
GO TO 530
CONTINUE
USUM=0.0
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 IMAIN4400
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
YCOCRD
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
.. STRESSDEFORMATION 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
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.
INPUT TABLE
PROBLEM
1A
PARAMETERS
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
..
YCCORD
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
..
EXAMPLE
62/1D
STRESS DEFOR/SURFACE
INPUT TABLE
PROBLEM
1A
PARAMETER S
INPUT TABLE
NODE
KODE
1
.=
. . .
. .
.=
RO/DEN OF WAT
O.IOOE 01
O.IOOE 01
. .
OR MV
O.IOOE 04
. .
=
=
=
.
.MATERI AL PROPERTIES
. . .
..
MATDENS
0.500E 00
YCOOPD
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
STRESSDEFORMATION 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
62/1D
STRESS DEFOR/SURFACE
INPUT TABLE
1A
PROBLEM
PARAMETERS
Problem
INPUT TABLE IB.
MAT
C OR MV
RC/OEN OF WAT
0.1O0E 04
0.100E 01
0.100E 01
YCOORD
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
..
KODE
1
EL
.MATERIAL PROPERTIES
INPUT TABLE
NODE
(cont.)
3.
STRESSD6FCRMATICN 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
62/1D
INPUT TABLE
1A
PROBLEM
..
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
YCOORD
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
STRESSDEFORMATION 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
63/1D
STEADY FLOW/CONSTANT
INPUT QUANTITIES
INPUT TABLE
PROBLEM
1A
PARAMETERS
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
YCOQRD
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*
0.333F01
0. 333601
0.333E01
2
3
6.
0.100E 03
0.100E 03
0.100E 03
OUTPUT TABLE
PROBLEM
AREA
EXAMPLE52/CHAPTER5/HANDCALCULATION
INPUT QUANTITIES
INPUT TABLE
PROBLEM
1A
PARAMETERS
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.
...=
...=
...=
..
YCOORD
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
IE DEPENDENT
53..
DATA
OUTPUT
Lt^i.
..
iooe ci
=C^
11
==:~LEMS
DPTION
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
7.
INPUT QUANTITIES
INPUT TABLE
PROBLEM
1A
PARAMETERS
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
..
YCOORO
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
SOLUTION TI*E =
30 OE
03
OPTION:
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
8.
INPUT QUANTITIES
NPUT TABLE
PROBLEM
1A
0.400E01
0.200E 00
0.50QE02
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
0.100E02
0.100E01
0.10GE02
0.100E0L
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
YCCORD
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
<
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
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=
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
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.)
..
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
Problem
NODE
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
NODE
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
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
0.9214E01
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
0.8632E01
0.7420E01
0.5837^01
0.4087E01
0.2094E01
O.OOOOE 00
3
4
5
6
7
8
10
11
12
13
14
15
16
17
18
19
20
21
NODE
1
2
3
4
5
6
7
e
9
10
11
12
13
14
lb
16
17
18
19
20
21
NODE
1
2
3
4
5
6
7
FflCTQK= 0.137E 01
TIME FflCTCK=
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
IC1N=
0. 78hE
PORE PRESSURE
O.OOOOE 00
0.6125E01
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
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
0.4439E01
0.8803E01
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
0.2847F01
0.2326E01
0.1772E01
J.1195F01
0.6016E02
O.OOOOE 00
O.OOOOE 00
3.2922E01
0.5 83 5E01
0.8729F01
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
PORE PRESSURE
NODE
170
(cont.)
NUDE
8.
00
00
00
00
00
00
00
JO
Problem 8 (cont.)
13
14
15
16
17
18
19
20
21
** JOB
0.2092E 00
0.1491E 00
0.3507E01
0.1882E01
0.1536E01
0.1170E01
0.7887E02
0.3969E02
O.OOOOE 00
END **
171
BEAM BENDING
AND BEAMCOLUMN
INTRODUCTION
Problems of beam bending and beamcolumn analysis using onedimensional
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
(7la)
where w*
is
p(x).
dx'
and w
as the
We now
Step
1.
element formulation.
172
finite
idealization the
Flumped
beam can be
at the line.
The
replaced by a line
idealized
beam
is
now
Beam Bending and BeamColumn
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
with trans
Figure 71
(a)
verse
shown
Step
A generic element
in Fig. 7 1(d).
2.
it
To
3),
was necessary
to satisfy interelement
we
of
com
174
fulfill
nodal displacement
therefore possible to
Chapter 7
the physical
firstorder (linear)
it is
As
and
slopes, that
a consequence,
is, first
it
derivative (gradient)
becomes necessary
to use
it is
we can add
slope at the
node
as
node
and
vv 2
at
node 2
wu
at
[Fig. 7l(d)].
0A00A0
Local nodes
Elements
Subscript^
SuDerscript
node
element
local
=
A
(a)
(b)
beam
3). (b)
Inter
bending.
at
w (x)
w(x)
N,0 2
(72a)
or
[N]{q}.
(72b)
Chapter 7
Here{q} r =[w! 6
polation functions
175
= [N N N N
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
),
(73)
Nt
In Fig. 73,
we have shown
ls (s
1).
plots of
i9
1, 2, 3, 4.
In mathematical
[2, 3].
^_N
5^*
'3
N4
_^
It is
worthwhile here to
it
1, 2, 3, 4,
[Eq. (73)].
where
a,
= generalized
matrix notation as
= j
f
we can
cc 2
3,
where we started
<x 3
x2
a 4 x
(74)
176
w (x )
X2
[1
^=Q =a
Chapter 7
X2]
(75)
leads to
2a 3 x
= [0
2x
= [*']{}.
3a 4 x 2
3x 2 ]{a}
(76)
Wl (x
"10
= oy
*i(*=0)
w 2 (x = I)
[q]
0"
<x' 2
(77)
<
M* = 0.
_0
2/
a3
3/ 2 _
.4
or
{q}
(78a)
4]{a}.
Therefore,
{a}=[A]'{q}.
(78b)
w(X)
= [*][A]'{q},
(79)
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
Approximation Function
highest order of derivative in the potential energy function in Eq. (7 14a)
below.
It
it
Chapter 7
is
a, in Eq. (74)
177
states of strain
PHYSICAL MODELS
It
For
instance, the
3.
3.
Define StrainDisplacement
[1],
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() )
commonly
is
expressed
M[x)
= F(x}w"(x).
(711)
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
nvTir
/,
,
F SP = Frf? [N]{q}
j
(713b)
[B]{q},
matrix;
(7 " 13a)
'
its
by proper
For example,
= j[6s 
6s 2
1(1
4s+
3s 2 )
6s
l(3s
9,
2s)]
(7 13c)
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.
ENERGY APPROACH
We now
derive
Up =
Substitution of w and
II,
where
F is
[*'
w" from
= \Fl JoV
=
Differentiating
IT^,
pw
element,
(7 14a)
dx.
{q}W[B]{q}<fe
/f
'
(714b)
\N\p{q}ds,
^0
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
matrix notation
as
1
(715a)
Jo
Jo
or
M{q}
(71 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)//
p^JJ
2(^2)
6(2,!)
ffi^iyj^
(? . 16a)
Chapter 7
ri2
[k] = ~~
6/
12
6/
4/ 2
6/
2/
12
6/
sym.
179
(7 16b)
4/ 2
p
where
/7j
=(1 
varies linearly as
^s/? 2
5)/?!
(71 6c)
and p 2
3s  2s
ls(l 2s + s
2
(I
and
2, respectively,
3
)
{Q}
= /f
15]
[S
 2s)
(s  1)
(716d)
s (3
ls
J>2
or
7/?,
3/? 2 \
^(3/7,
f
2/7 2 )
(716e)
=2i<
^(2^
3/> 2 )
The
residual
is
we consider one
3,
generic
given by
(717)
S"
method
p Ntdx
gives
/=1
'
2 3 4
'
(718)
'
first
term leads to
d*w dN,
dx 2 dx
w from
Substitution of
0.
(7 19a)
<****+'*
,d
w dN
dx 2 dx
= 0,
i=l,2,3,4, y=l,2,3,4
(719b)
or
I,
</x
</x
"**
Jxi
^'^
' </x 3
dxl dx
(7 19c)
180
Chapter 7
N'i
J X\
Nim
N'm
N'iN'(
1
N'iNT
N'iN'l
aw
N?
sym.
w,
N'i
(720a)
>dx.
w2
e2
same
x. After
this expression
The
first
(720b)
iVj
dx'
we
(Fig. 73),
(f)\
(720c)
(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
(720d)
Chapter 7
181
or
m.
X
dx>) 2
(720e)
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
(720c)
and
no
moment
Sit
xu
two nodes.
forces at the
when
(720e)] are
assembled for
all
It
may
be noted that
if
beam
will remain,
and minus
whereas
signs.
The
nonzero end terms will denote the boundary conditions; for instance, for a
simply supported beam the end term in Eq. (720e), 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
1.
Steps 5 to 8
We
20
cm and crosssectional
area
= 2 cm
cm deep x
(2
beam with
1
cm wide)
and
E=
10 6
16
15
153
100
15
cm 4 ):
15"
50
'
0,
<
3
15
15
50
15
15
w2
100_
e2
3'
1000
(721)
"
,5,
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 74 Example for
By following
beam
bending.
3,
we
obtain
15
15
100
3
15
15
50
15
15
50
15
The beam
(Fig. 74)
3]
0i
15
w2
15
50
15
50
15
(722a)
15 I*
Wy
kJ
100_
5j
is
(hV
200
16
0"
0"
(w>r
f
15
100
3
15
15
50
15
50
16
15
15
50
w2
1
200
15
50
(722b)
<
r
W>3
15
1
5
0i
100_
kJ
,5J
Chapter 7
183
w,
0j
= 0.0000 (given),
= 0.0500 radian,
=
=
w2
2
0.3125 cm,
w3
0.0000 rad.,
0,
dw_
dx
which
yield the
same
computations. That
method
yields the
[1]
= 0.0000 (given),
= 0.0500 rad.
P x (T3
{L
24F
2Lx 2
p (fZ
24F K
6x 2
x 3 ),
(723)
4x 3 ),
(724)
from the
is,
same
(722c)
finite
element
finite
element
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,
(725)
0.0000
M(at$
0)
0.0500
[6
4/
21],
>
0.3125
.0.0000
' 25F
0.08334
\l
833.4 kgcm
lilarly,
M(at
M
For element
(at s
0.5)
1)
 0/250F 
15F
~f
2
3333.3 kgcm,
5833.4 kgcm.
2,
0.3125
M(ats
0)
= ^[6
4/
0.0000
6
21}.
0.0000
0.0500,
5833.4 kgcm.
Beam Bending
I
and BeamColumn
184
Chapter 7
Similarly,
M(2LtS
0.05F
0.5)
M(aXs
0A25F
1)
I
M=
which
at
//4
solution for
moment
3333.3 kgcm,
833.4 kgcm.
is [1]
\2p
(Lx
24
Fw'Xx)
and x
x 2 ),
(726)
1/2 gives
m(x =
j\
M (x = L\ =
Figure 75(a) shows the bending
3750.0 kgcm,
5000 kgcm.
two elements,
(a)
(b)
forces:
grams.
~j^ 833.4
500 kg
(b)
Chapter 7
is
yields
V (x
^=
0)
= y) =
V (x = L) =
(*
From
which
d3w
dx 3
V
which
V,
185
1000.0
kg,
0.0kg,
1000 kg.
dx
^[12
ds
6/
12
(728)
6/]
/,
[0.00001
K=^[12
6/
12
0.0500
500.25 kg,
6/]
0.3125
0.0000
2,
Figure 75(b) shows plots of shear force from closed form and
finite
element
computations.
yield the
finite
same values of
element method
is
satisfactory.
finite
element computations
show a wide disparity as compared with the closed form results [Fig. 75(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)
MESH REFINEMENT
The beam is now
/,
/,
cm, as shown in
30
12
10
125
100
X 2
x 3
12
30
30"
50
100_
20
cm
Figure 76
0i
w2
250
kJ
2500/12,
cm
cm
Mesh
25
f
2500/12
30
12
ym
cm
refinement for
beam
cm
bending.
"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
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,
forces
from the finite element analysis and those from the closed form solution. It
186
Chapter 7
cm
cm
187
cm
cm
208^
Closed form
Finite element
a
243
(b)
moment
moments and
shear forces:
grams.
now
twoelement approximation.
HIGHERORDER APPROXIMATION
As the next step from the cubic approximation, we can adopt
a fifthorder
approximation as follows:
w(x)
=
=
ol
[1
a2x
X2
a3
A'
.\
J
cc X
B
4x
a 5 x4
a 6x5
(729)
where
{a} 7
The
first
and second
"
[a,
%2
%4
a.
a.]
= [0
=
ax
3x 2
4x 3
5.x ]{ai
(730)
Chapter 7
and
d 2w
dx 2
w
where
vv"
at the
\2x 2
6x
[0
20x 3 ]{a},
(731)
is
"10
'Wt
"
10
w"
{q}=<
and w'
>
IIP
w2
2/
/5
/4
4/
6/
12/ 2
3/
_0
(732a)
5/
20/ 3 _
or
{q}
(732b)
[A]{a}.
Therefore
{a}=[A]{q},
(733)
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
1, 2,
6,
1
(734)
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 ),
(735)
\5s*
),
V
Use of
patibility
these
of displacement w, slope
6,
Chapter 7
189
BEAMCOLUMN
If in addition to the lateral
load
[Fig. 7l(a)].
effect is called a
To
then the
load the
beam
beam
is
also acts as a
beamcolumn.
we assume
is
small in corn
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
190
Chapter 7
assumptions,
Step
As
we can superimpose
the bending
and
is
linear.
Under
these
before, the
beam
is
Step 2
Now
we have
(72)]
and u
Chapter 7
[Fig. 7 1(a)].
191
u(x)
[\
(736)
[NJ{qJ.
s]
The combined
(72)].
interpolation
["Ml
[NJ
'
(1
"
[0]
2)
(1
fe.J
4)
(737)
w(x)
(1
[NJ
[0]
[q b ] are the
same
2)
(1
and
as [N]
4).
{q} in
we have used
Steps 3 to 5
and
(315)
and
and the
and
in Eqs. (710)
(711), respectively.
The
potential energy
bending [Eq.
II,
=/
and
(714)]
is
iF(w") 2<fc
Here we assume
axial
deformation [Eq.
I
energies due to
Hence
wpds
Pu.
is
Substitution for
E@ffds
Al
sum of the
(321)].
[Eq. (72)],
(738)
Pu denotes
and du/dx
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
(739)
set
of
uniform
in
to be
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
(740)
P
"2
I
1
We
78.
COMMENT
What happens
if
the load
is
effects
if
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
For generating a
Chapter 7
and rotations
beam
at
193
like a builtin
is
expressed as
f*
fw,l
(741)
n\N^
l
Q
The
variation of
moment
(742)
1]
[x
Note
that the
moment replaces
\}dx
(743a)
3/
2
2/
21
(743b)
3/
3/13/
relation between the
flexibility
is
[x
The
stress as the
IT
is
given by
(744a)
=IQ X
Mu
(744b)
or
1
ia
2,
1
[G]
(744c)
(744d)
M,
Now
the inverse of
[f ] is
LJ
/
By using Eq.
(745)
b/
2/2
the
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
M
M=
 s)M + sM
(1
= [N m ]{M
(746)
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. (363) for beam bending can be written as
t
n,
=  i j" MMdx +
J"
M^dx  {Qf{q},
(747)
where
{Qf=[e,
2]
and
r
{q)
Substitution of
=K
<?,
e1i
>v 2
IIr leads to
{M.w
AJ
lf[
+ {Mr
r [NjqN'Trfxfa}
 {QF{q}.
(748)
2)
fe = A [Njr[Nk/x{M
+
^g
P [N J
[N"]rfx{q}
1
(749a)
0,
= o + j" [NJ[N"]V.x{M} 
{Qj
0,
(749b)
Chapter 7
195
or in matrix notation,
'
~~[6fJ
1
(749c)
<
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"
(749d)
= rtNJTN"]^
s
Hi'
i//
[6 + 2^
4Z + 6/5
6125
6ls2l]ds
1
1//
0'
(749e)
i//
1//
IF
6F
J_
6F
J_
3F
_J_
j_
1
_1_
l_
e,
(750)
<
o.
Qi
4
As an
elements.
illustration,
we
The assemblage
consider the
r
beam
two elements
is
two
obtained by
196
Chapter 7
1
3F
6F
(f
W\
1
1
11
6F
3F
6F
Mi
H'2
QTQr
M  .\/f
3F
(75 la)
l)
02
6F
>
M<?
0i
Q\
M,
1
/
QT
H'3
0_
03'
2)
or
[K]{r]
By
(1)
and
(2)
(751b)
example of
Fig. 74,
we
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_;
>
boundary conditions
(751c)
10
= 0.
10
1
12
Introduction of the
\r
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
r5id>
0:
2.5
~w
10
Vi
10
1
Note
Chapter
12/
such a system
If
10
3.
we
is
shall
on the diagonal.
Af
*,
d.
= 833.33 kgcm.
= 0.00 cm (given).
= 0.04999 rad.
M_ = 5S33.33 mgcm.
w z = 0.312495 cm.
=0.0000
0
V 3
833.33 kgcm,
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
same
moments
are primary
PROBLEMS
71.
(a)
Derive
stiffness
matrix
if
EI
= F varies
f = y.F  \ F
= s. s = (x x
:
where
(b)
V =
:
Derive
[k] if
and
\"
;
beam
A = A \
:
72.
I.
 A \z
z
Z acting in
"2
.
I
Solution:
4 i2'
1 1
the z direction,
198
73.
Derive
if
by a
matrix
stiffness
beam
the
is
the
[k] for
supported on an
Chapter 7
elastic
(with uniform F)
series
(Fig. 710).
Figure 710
Part/a/ solution:
The
Beam on
elastic foundation.
given by
is
(k f w)wds
Jo
folTNTMNHq}*,
which
stiffness
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_
74.
_1
in Eq. (79) and show that the result yields
Multiply [0][A]
4, in
75.
Eq.
Perform
it i
1, 2, 3,
(73).
all
(71 6a) to
steps of multiplications
Eq.
(71 6b)
and integrations
(71 6c) to (71 6e).
for going
from Eq.
Chapter 7
76.
Derive {Q}
if
199
p{s)
+N
NiPi
p2
+N
where
= 2s 2  35
N = s(2s 1)
No = 45(1 s)
Ni
end fixed against movement and rotation and the other end
settles by 0.05 cm.
78.
One
One end
Solve the example problem in Fig. 74 with the following conditions (a)
(Fig. 74)
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
79.
VV 3
Compute
0.
face traction
p with
body force
to
Z and
sur
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:
kit
For constant
is
'
{Q}
= AZ
Compute
tion
(734)]
/120
>
/10
/120
Pi
1/2
/ 2 /10
/ 2 /10
model [Eq.
/10
1/2
,/
710.
H2
'
111
'
/120.
/120,
(1
s)pi
sp 2
the interpola
linearly as
200
Chapter 7
Solution:
Pi
Wt +
Pi
*lft+A<**>]
Jpj_
Pi
280 >)
*120^
V12C
{Q}={
{% + Y4<J>2Pi)]
[3
1
(Pl.
V120
+ Pi 210
__
711.
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. 74. Comment on the
distribution of these quantities around the junction of the two elements.
712.
Assume
beam element
 N A + N2 A 2
X
and derive element equations for uniform loading. Assuming the area to vary
shown in Fig. 711, compute displacements, slopes, movement,
and shear forces for the loading and properties of the example in Fig. 74.
linearly as
in
moment
of inertia should
also be considered.
10
cm
cm 2
10
*
cm
cm 2
3 cm'
Figure 711
REFERENCES
[1]
Timoshenko,
S.,
Strength of Materials,
1956.
[2]
[3]
Finite
New
York, 1975.
Chapter 7
[4]
[5]
Crandall,
S.
to the Finite
McGrawHill,
201
New
York, 1956.
ONEDIMENSIONAL
MASS TRANSPORT
INTRODUCTION
mensional idealizations.
The
differential
be stated as
where
Dx
[1]
is
is
salt),
the
vx
unknown
is
or state variable
is
FINITE
first
term
is
essentially the
ELEMENT FORMULATION
main
steps, while
labels.
202
some of
the
we
common
and
Chapter 8
203
H
T
\
\
\
\
\
\
\
'a)
(b)
A A A
k
L=+1
L = 1
(c)
and convection,
(b)
Mass transport by
Onedimensional idealization,
(c)
Dis
cretization.
i(l
L)c
i(l
+ L)c
(82a)
or
c
where
{q}
[cj
c 2 ],
[N]{q]
and
is
= E Vet,
4.
U2,
(82b)
and
medium.
it
residual
is
(83)
204
Weighing
with respect to
TV,
yields
r(S
As explained
ever,
Equation
in
Chapter
3,
we
(84)
is
NjCj) N dx = 0.
t
it is
Chapter 8
is
expanded as
[ilD>i&
Nrt wJSEg*
dx
W  jf(LNJ>)Ntdx = 0.
Integrating by parts the
stant,
(84)
first
term
in
Dx
(85)
to be con
we have
JT^icEW }N dx = D
x N,$i:
dn
*/
X\
The first term on the righthand side in Eq. (86) denotes Neumanntype
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. (86) into Eq. (85) leads to element equations in
matrix form as
[E]{q}
[E,]{q}
(87)
{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
(88a)
[E,]= r'[NF[N]</x,
(88b)
and
{Q}
p [WWdx  D N^
x
(88c)
Here [B] is the usual transformation matrix [Eqs. (313) and (44)] obtained
by taking the proper derivative of c [Eq. (82)]. In writing Eq. (88a) we
assumed, only for simplicity, that v x
second term in Eq. (85)
is
is
not included.
We
note that
it is
not necessary to
Chapter 8
[1
[E]
 /[_]

We
ij
205
\}dL
+ 2[
i_
(89)
note here the important characteristic that the element matrix [E]
nonsymmetric ;
mass transport.
[E,]
this
property
is
is
Now
2
"2
[1
L
L]dL
11
(810)
This matrix
[Eq
is
(521)].
Wl\
(811)
1]
(4
The first term in Eq. (811) 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
Combination of element
81).
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 \lv x \2
2D x jl +
(c
D x /lv 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
,
(812a)
206
The contribution of
Chapter 8
is
we assume
that
(A0 =
(813a)
and
c(0,t)=c.
(813b)
[K]{r]
where
and
[K]
KSjt)
{R},
(812b)
{r} is
is
the assem
unknowns.
Solution in
Time
number of
For
form as
difference
[2]
 0)W,)
+
RK{rU +
+ o^W^{*L
We
where 6 is a scalar.
ing on the value of 6.
= ^,
we
(8  14)
If
1,
method
is
obtained, and
if
and
stability
which can
aspects
If
is
we
text.
first
derivatives in
 ^[K ]){r} r
(816)
Eq. (814),
^[K,]){r} f+A
~ 2{R} f+A 
{r},
([K]
are
is
at given
CONVECTION PARAMETER v x
For the solution of Eq. (816), the value of the convection velocity v x
should be known. It can often be obtained by using available data or formulas.
Chapter 8
One
it is
207
4),
On
if
Example
To
81
illustrate a
in the solution of
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.
field situation.
Boundary conditions.
aO.n =
(ML
1.0.
3
)
at
node
1.
>
0. (Fig. 81).
Initial conditions:
ci.v.
At time
i
i
1
At,
from Eq.
"2
0"
i
4
4
_3
0)
(.816).
0.0.
we have
0~\ 'ci
14 10
14
J_
j
l_
_0
1
= 
i i
4
!_
c:
c3
2_/ [C4
Ar
o"
21
i
\
r<0.
04
2
1
f
4_
x
x
~2
:_
_0
"\
\C\
c2
Cl
ie4
208
Because c
c2
c3
C4
0at
o,
0"
Cj
_7
20
Cl
=r
>
37
20
i
Chapter 8
6
1
13
,c A
<
>
c3
Ar
gives
Cl
20
J
1
20
13
i
C2
<
c 3
[oJ
 lc 4
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 82
A solution to Eq. (81) 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 onedimensional.
The
initial
v x t/L,
medium.
conditions are
c(x, 0)
0,
is
c(0,f)
That
is,
Co
(0,0
Table 81 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
Chapter 8
TABLE
209
81
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
REFERENCES
[1]
Bear,
J..
Dynamics of Fluids
in
New
York,
1972.
[2]
Problems. WileyImerscience.
[3]
Guymon. G.
L..
Finite
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
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.
210
Chapter 8
Desai, C.
and
S.,
Salt
L., Scott, V. H., and Hermann, C. R., "A General Numerical SoluTwoDimensional DiffusionConvection Equation by the Finite Element Method," Water Resour. Res., Vol. 6, No. 6, Dec. 1970, pp. 16111617.
Guymon, G.
tion of
Gray,
W.
W.
G.,
"A
Smith,
I.
No.
2,
Wu,
No.
3,
T. H., Desai, C.
Salt
ONEDIMENSIONAL
OVERLAND FLOW
INTRODUCTION
Overland flow can include problems
in hydraulics
runoff due to rainfall, flow in shallow open channels and in rivers in flood
plains,
is
threedimensional, but
we
this
it
can often
introductory
onedimensional case.
and
(i)
dQf.dA
+ dt
dx
(ii)
Momentum
(91)
0.
equation:
dQ
*+&5) ***>*
<>
dx~
Here
Qf
is
flow
bed
(r )
slope,
in the
for overland
and Sc
for channel
Sf
is
is
(r c )
is
;
the
is
211
212
Chapter 9
Ac
Section CC
00
Section
(a)
due to
rainfall, r
(b)
Channel inflow
I
rr
(c)
Figure 91
Representations
direction
is
overland
for
(b)
and channel
flows,
and
is
the
Sf
(92b)
We
Chapter 9
K=!^/y
n
213
S 1/2
(93a)
or
Q f = h^.RpS 1/2 A
V
where
is
Rh
is
(93b)
area/wetted
channel flow
As
In the
first
and onedimensional
a simple approximation,
stage, the overland flow
we can
solve the
due to excess
problem
rainfall, r
in
two
stages.
[Fig. 9 1(b)], is
computed by solving Eqs. (91) and (93) for the overland flow. The results
of flow from the first stage are considered as the input flow r c in the channel
[Fig. 9l(c)]. Then Eqs. (91) and (93) 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
arbitrary flow
The
channel
is
Step
2.
idealized
the discharge
Qf
A(x9
t)
= (ls)A (t) + sA 1
l
(t)
[N]{A(r)}
= jlN Afi)
t
(94)
214
Chapter 9
of topographic
and other properties
(a)
r
~~
Overland
0)
.c
*
 <>,<
^
*
>1
L_
(b)
(d)
(c)
= x//
(e)
Figure 92
Finite
element
discretization,
Overland flow,
(d)
(a)
Actual
Channel flow,
(e)
region,
Generic
element.
and
Qf (x,
t)
is
= [A
where {A} T
A 2]
is
(95)
= [Qfl Qf2
Chapter 9
215
we
area
as the timedependent
Qf
are
known
at
unknown.
Step
4.
We
(96)
i
Use of Eqs.
(94)
and
(95) gives
(97a)
[BHQ,.}
and
^=
=
When
[1
s]{A]
[N]{A}
[N](^>
(97b)
we obtain
n[NHN]rfjr{A}+
J x\
JfX\
[Wrdx
(98a)
J X\
or
[15
s]ds {k " ]
r[
(2/1
Ir
V2
11
'
'
Qn  Q ;
Ir
;
dA
_1
2j [dt
(98b)
rrrh
[dAA
dt
 Qf i)ds
'
<
(99a)
>
7
J
or
[k]{A}
Step
5.
(99b)
{Q}.
Assembly
as
/,,
/2
and
/3
[Fig.
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
(910a)
<
hr 2
hr z
IVs
or in matrix notation
[K]{A}
{R g }
{R,}
{R}
(910b)
confusion with
r that is
approximation for
first
deriva
tive.
A A
(b)
(a)
+ At
Local node
/l\ Element
Chapter 9
We
use the most simple time integration, often called Euler formula to
approximate the
first
jA^jf
iAy]r+Ar
lAtfJ
If
we
(911)
At
^[K]{A^J f+Af
[R Q }
now
The
+ JptKKA*},.
(912)
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,
(91 3a)
at all times
that
2/(0,0=0.
The values of excess
rainfall for
is
zero at
is,
(913b)
= F(x,
r(x, t)
(913c)
t),
With
J.
0"
"2/,
/,
Time
1,
2/,
6At
2/ 2
2/ 2
.0
/,
2/ 3
2/J
Mil
A2
Si
+ hh
At
is
(914)
Ay
UJ
I3P3
are
now used
[K]{Aj
The procedure
is
2Af
,}*
{R,h*
+ ^[kha*}*.
is
reached.
(915)
Step
Solution for {\ N }
6.
Example
As a simple
91
procedure
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
0"
x 0.166
At
A2
A3
2_
A4
field situation.
as
At
_0
assumed constant
0.166 hr.
we have chosen
m 2 /hr,
1
1.0
x 10"
2
,
2
2
1
At
or
0~
~2
A3
2_
A4
0.5
'Ai
A*
(916)
'
_0
Comment.
'
.0.5. At
At
The
3),
is
A = A2 = A3 = A4 =
l
0.166
(99b)] repre
in
At:
2
.
Now we use the Manning formula [Eq. (93)] to find nodal flows Q f by assuming
the following data:
= 0.3,
5=
0.1;
at
nodes
at global
= 0.5 m,
= 0.6 m,
w3
= 0.75 m, u> 4 =
0.166/0.5
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
x 0.166
0.125
/hr,
1 .3
m.
Chapter 9
1.49
Qf:
x 0.166
0.110,
x 0.166
0.095,
0.3
Q f3 =
1.49.
(0.128)
Qf
To
219
667
(O.l) 05
x 0.166
0.066.
0.3
evaluate the
first
necessary to
it is
convert the above flows at global nodes into flows at local nodes, Fig. 93(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. (910a) and (915),
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
(917)
x 0.166
0.1667
10.1667
.0
2A/.
The procedure
is
repeated
number of steps.
Example 92
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
dure.
It
consists of
three channels.
The
finite
element formulation
is
overland flows and three channel flows and the results are added to obtain computed
flow at the gaging station. Special provision was
modate
it
made
in the
formulation to accom
contains three overland flood plains (OFPs) that are idealized as 24 one
flow planes,
strip,
and element
first
flow plane,
Table
91.
The
discretized similarly; the properties are given in Table 92. Values of excess rainfall
220
Chapter 9
Legend
Overland flow plane boundary
Surface element boundary
Channel element
Primary channel
Secondary channel
Flood detention structure and reservoir
IIB1
watershed
mesh
[10].
which
3 to
1 1
in.,
19, 1969, in
in
Figure 95 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
number of other
Chapter 9
TABLE 91
221
[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
92
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)
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
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
stability
may
the derivations
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)
gaging station, At
and
900 sec
[10].
REFERENCES
[1]
Stoker,
J. J.,
Problems, Report
ical
Science,
[2]
[3]
[4]
New York
Abbott, M.
University,
New
York, 1953.
/.
J. A., and Woblhiser, D. A., "Difference Solution of ShallowWater Equations," /. Eng. Mech. Div. ASCE, Vol. 93, No. EM2, April 1967,
Liggett,
pp. 3971.
[5]
Prince, R. K., "Comparison of Four Numerical Methods for Flood Routing," /. Hydraulics Div. ASCE, Vol. 100, No. HY7, July 1974, pp. 879899.
Chapter 9
[6]
Lighthill, M.
Movement
in
and Whitham, G.
J.,
Long
B.,
223
I:
Flood
May
1955,
pp. 281316.
[7]
[8]
versity of
[9]
[10]
[1 1]
Alabama
Taylor, C., AlMashidani, 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 LandUse
Changes on Flood Hydrographs," M.S. thesis, Virginia Polytechnic Institute
and State University, Blacksburg, Va., Nov. 1975.
ONEDIMENSIONAL
STRESS WAVE PROPAGATION
INTRODUCTION
5, 8, and 9, timedependent 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,
Now we consider
medium
as onedimensional.
Consider a homogeneous bar of uniform cross section (Fig. 101). A timedependent force Px (t) acting on the bar causes vibrations in the bar, and a
(stress)
wave propagates
to
and
given by
where a x
is
[1,
The governing
known
as the
differential
wave equation,
2]
is
direction,
is
is
Chapter 10
225
(a)
= +1
(0
(b)
and
discretization,
(c)
Generic element.
force;
acceleration
and
Px (i)
is
= pii
derivatives,
is
the elastic
is
is
the
assumed to be
is
=E^.
L
dx
(102a)
we obtain
fo x
dx
where
law
Ee
Lx
ox
first
Taking
The
=E
dx
modulus and e x
(102b)
is
dx 2
dt 2
+ PM.
(10lb)
FINITE
classified as a hyperbolic
is
equation
[3, 4].
ELEMENT FORMULATION
= N u, + N u = [N]{q],
N = s, and s = (x x
u
where N,
5,
t )/l.
line
(103)
Step
4.
{x f8{*x
JJJ
work equation
W=
[4, 5]
{F x Yd{u}dV
JJJ
JJJ
{Px Yd{n}dV,
Fx
The
(104a)
=P dt
is
the equiva
(104b)
6x
dx
(105a)
1]
or
{}
[B]{q]
duy
dt
du 2
\dt)
(105b)
[N]{q],
and
{d
uA
dt
d u2
[dt 2
=
since the
Af
(105c)
[N]{q};
(JJJ
{e}, {u},
and
{u} into
[BF[C][BJ^){q}
W
+
226
(JJJ
{<5qf
JJJ
P[W[^W)W
[N]{P^K.
(106)
Chapter 10
JJJ
[BF[C][B]^K{q}
JJJ
p[W[K]dV{q]
we have
{X\ T
JJJ
F*W
(107a)
227
or
M{q}
We
[m]{q]
= {Q}.
(107b)
principle
[5]
1
M = AE
(108a)
1
AIP X (Q
{Q(01
(108b)
and
w^f'f/ki
s]ds
pAl
(108c)
The matrix
in Eq. (108c)
is
it is
may
Often,
it
Apl
divided equally
among
On
the
two nodes:
Apl
"1
0"
(108(1)
2
is
diagonal,
it
5.
more accurate
is beyond
this aspect
for
the
text.
Then
distributed to
first
tages.
is
is
{r}
4) (4
1)
[M]
X 4)
(4
(f)
(4
1)
{R(0},
x 1)
(4
common
nodes.
we have
(109)
228
L = 30
Chapter 10
cm
j1
kg
t
A A A
Figure 102
Mesh
for onedimensional
medium.
where [M] is the assemblage mass matrix and the other terms have the same
meanings as in Eq. (332).
Equation (109) 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
,
As
in the case
of Eqs. (521),
(87),
and
(99),
Time Integration
a simple approximation,
linearly
L near approximation
Quadratic
/r
n
t
1t
+ At
(b)
(a)
t
(0
Figure 103 Approximations for time integration, (a) Acceleration,
(b) Velocity, (c) Displacement.
Chapter 10
229
M'U*
(1010a)
{},*,
where
[K]
{R},.^
= [K]^L[M].
(1010b)
 ^[M]({r} 
{R},.*
Ar{f} r
^(Ar) 2 {r}
(1010c)
).
are
known from
l*U
= 5<M* 
m =
Wo)
is
2{f}
^{f}
(101 la)
^(M*  Wo)  5 Wo  m*
the given
(ioi lb)
is
desired.
The assumption of
Example
linear acceleration
101
between
Assume
wave
elastic
elements
Element length.
10cm;
/ ==
cm :
==
==
Density,
= = 10"
Wave
Cx
Element area.
Elastic
modulus.
velocity,
lOOOkg'cm 2
11
= Jf
cm 4
kg'sec :
109
cm
illustrate the
sec
procedure; they
may
not
An
approximate
size
At
=^=
10
c
7
106 sec
(1012)
This size
is
[Eqs. (109)
and
numbers
(1010)] leads to
230
10
_
x 10~ n
[M]
21
01 21
01
"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
(1013)
_0
1_
is
2_
times; that
Wl (0,
zero at
CONDITIONS
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)
(1014a)
0.
0:
v(x, 0)
u{x, 0)
ii(x,
0)
=
=
=
0,
(1014b)
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
node
4, as
a constant
force
F4 (t) =
With
(1014c)
kg.
is
~2
(R}r + Af
[R}f+Af
6 x lO" 11 x 10
10" 6 x 10" 6 x 6
_0
10 12 {r},).
0"
2_
(M,
(1015)
Chapter 10
10
At
At as
(1016)
<
ir 3
12.
.0
lw 4
Ar
u { (At)
in
10 ur
14 10 w? =
14
1
Now we
0"
2
10
at
231
T600
600
(1017)
300.
prescribed) gives
u'
1
(1018a)
U3
uj
Ar
and
300 }
initial
computation of
velocities
and accelerations
at
a)
At as
o\
dt
()
fl
du 2
dt
du 3
>
{
>
>
1Q 6
2<
(1018b)
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<
>(1018c)
dt 2
d 2 uA
300
dt 2
Ar
now
/o
We note here that the foregoing numerical calculations are meant only as an illustration of the procedure
as acceptable solutions.
optimum
spatial
Example 102
Now we
solved by
The
[7].
wave propagation
follows:
L or
Number
of nodes
Number
of elements
500
mm
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.
Modulus of
velocity,
Time increment.
=:
500
mm
*
mm/ms
[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
mations, respectively.
mass
sistent
It
and
can be seen that for the characteristic time step, the con
is
not that
accurate.
The
size
of the time step can have significant influence on the numerical solu
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
2.0
 0.08 ms
:
o
1.0
......
oo
o
a
11
21
31
Node number
(b)
Analytical solution
2.0
t= 0.08 ms
1.0
<
.a
,E
oo
11
31
21
Node number
(c)
Analytical solution
~E
2.0
>
= 0.08
ms
1.0
>
I

o
11
21
31
Node number
(d)
mass: At
At = 0.5
=
I
l'c x
cx
(d)
[7]. (a)
(b)
Consistent
Consistent mass:
233
Damping
[k]{qj
where
[c]
material
([4]:
damping
the
is
is
[c]{q}
[m]{q}
(107c)
[Q(t)\.
matrix. Determination of
and
discussed
is
in the
damping properties of
various publications
in
the bibliography).
PROBLEMS
101.
Compute
where
\L(L 
),
( 1
 L2
WL
uQ
)w 2 ,
Solution:
Apir\
[m]
UL
J.,
L(\
Apl
102.
Compute
1)1
2
)
\[L(L
(1L
1)
L{L\)]dL
2
)
1)J
84"
16
8
64
L4
120
L
(1
8
16.
u:
(1
3s 2
5 2 (3
2s 2 )u
ls(s
l) 2
^
OX
 1)^
Solution:
[m]
103.
104.
22
54
22
13
3
22
Apl
420
54
13
156
13
3
22
in
4.
Example
101.
mass system.
It is
time steps.
234
13"
156
Assume
properties as in
Example
Px
as
shown
in Fig. 106.
101
Compare numerical
results
Chapter 10
235
P.tlii
m
Figure 106
where p 1
respectively.
105.
The
initial
=p
y(l 
cos pt),
it(x,
0)
spring,
0.
Prepare a computer problem based on Eq. (1010) and solve Probs. 103 and
104.
REFERENCES
[1]
Love, A. E. H.,
New
York, 1944.
[2]
[3]
Timoshenko,
York, 1970.
Carnahan,
Wiley,
[4]
New
Desai, C.
S.,
B.,
and Goodier,
Luther, H.
[6]
and Wilkes,
J.
New
York, 1969.
and Christian,
S.,
Engineering, McGrawHill,
[5]
A.,
J.
Numerical Methods
York, 1977.
J. T., (eds.),
New
to the Finite
in
Geotechnical
Wilson, E. L., "A Computer Program for the Dynamic Stress Analysis of
Underground Structures," Report No. 681, University of California, Berkeley,
Calif., 1968.
[7]
Yamada,
Y.,
and Nagai,
Y., "Analysis of
5,
May
BIBLIOGRAPHY
Archer,
Div.
J. S.,
ASCE,
Vol. 39,
Structural Vibrations
Oden,
J.
T. eds.), University of
/. Struct.
1963.
Alabama
236
Chapter 10
Desai, C. S.
(ed.),
Desai, C.
S.,
and Lytton, R.
Parabolic Equation,"
Idriss,
I.
I, II, III,
M., Seed, H.
B.,
Num. Methods
Num. Methods
Geomech., Blacks
in
Geotech. Eng.,
1976.
Int. J.
and
ASCE,
in
ASCE,
113.
Kreig, R. D., and Key, S. W., "Comparison of Finite Element and Finite Difference Methods," Proc. ONR Symp. Num. Methods Struct. Mech., University
of
brich,
W.C.
eds.)
Academic
Press,
New
York, 1973.
/.
Eng.
TORSION
INTRODUCTION
Until
now we have
sional.
unknown
only one
at
line
beam
bending, had
idealized as twodimensional.
To
We
choose torsion
unknown
first
problems (Chapter
the torsion problem.
field
We
because
it
often called
displace
stress or equilibrium,
behavior of the cross section of the bar [Fig. 11 1(b)]. In the semiinverse
it is
is
composed
of the rotations of the cross sections of the bar as in the case of a circular bar
[1, 2].
As a consequence, no normal
is
constant for
stress exists
all
cross
237
238
Torsion
Chapter 11
y, v
(b)
(a)
Warping
(c)
Figure 111 Torsion of bar. (a) Bar subjected to torsion, (b) Cross
section of bar. (c) Warping.
Chapter
12.
case, a linear
two
12).
we
As
discussed in Chapter
3, it is
finite
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
areas,
A l9 A 2
and A 3
local coordinates
(Fig.
L l9 L
Then
112).
\
L2
\
L3 =
Since
in the
1,
(111)
1,2,3.
two independent
y.
local coordinates
The
relationship
given by
is
(11 2a)
=L
3
x
+L
+L
y3
= 2= Ajn
i
*
where [N]
is
= m*&
(ll2b)
{x n } T
is
it
[x M
x2
x3
y2
y3
for coordinates at
y).
1, 2, 3.
240
Chapter 11
Torsion
The
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
(ll3a)
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
'
(ll3b)
.*.
2
a b2
x
whose
is
vertices are
nodes
2, 3
as differences between
a2
a3
=x  x
=x x
=x  x u
2
is
b3
b2
where [N]
=y  y
=y y
=y  y
2
ly
general symbol
= A^w, + N
= [N]{q},
u2
+N
(H3c)
w,
can be expressed as
u3
(ll4a)
FINITE
ELEMENT FORMULATION
is
the same,
we can
Displacement Approach
First
torsion
we
[1, 2], in
body forces, the differential (Laplace) equahomogeneous and isotropic bar can be expressed
the absence of
as
dx 1
dy 2
(115)
'
where x and y are the global coordinates in the plane of cross section of the
bar (Fig. 111). 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
(Fig. 111):
w=6y/(x,y)
where
is
moment,
(ll6a)
>
the twist of the bar per unit length under the applied twisting
[Fig.
ll(a)].
components, u and
v,
are given by
= 6zy,
= Qzx.
u
v
(11 6b)
is
given by
(>)( + *)=
where
is
(117)
is 0.
3.
By using Eq.
(116),
is [2]
'dw
dx
[J
where
{e}
[y xz
du
dw
y
yz ]
+
is
(%  >)
dz
>
,dy
r
<
dv
dz
>
(118)
<
+
*)l
compo
nents of stress
[2].
r xi
Gy X2
Gy yz
.0
(119)
where
{<r}
[t xz
r yz ]
is
GJ
components and
is
4.
For the displacement approach, the potential energy function correspondis given by [3, 4]
where h
is
denotes area.
242
Chapter 11
Torsion
of
we have
y/(x, y)
= N y/ + N
= [N r ]{q r
1
2 y/ 2
+N
y/ 3
(ll4b)
},
where
y/{x, y) is
[y/ i
y/ 2
y/ 3 ] is
ing function.
As
discussed in Chapter
3,
form of
linear
y/
which
a,
as
a2x
a 3 >>,
(ll4c)
is
The
is
approximation for
is 1;
1=0, that
warping functions. The linear approximawarping functions across interelement boundaries. This is illustrated in Fig. 1 14. 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
^2 = warping
at
node 2 of element
1,
and so on
Chapter 11
Torsion
Note
fulfilled.
243
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. (ll4c) provides for rigid
:
in the
order or degree
(ri)
of polynomial
n=0
x
To
x zy z
*r
approximation
and including n
to
xy 2
x 2y
x~
Wc can
xzy
x3
up
XV
constant
Linear
Quadratic
Cubic
Quartic
terms
are provided.
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
(/
1, 2,
^L
dX
dx
iiai
Uina;
dy/
(1 11
dN
lb)
3)
and
y/
with respect to x
in Eqs. (113)
= 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^iXU:  X
azy)1
+&> + &>
etc.,
y/ t
(i
y/ :
NiWi)
X
y/ 3 )
(1M2a)
Nzz
*****
1. 2.
244
Chapter 11
Torsion
functions of
and
(x, y),
Similarly,
dy/
Y1
ar.+a*
2A Y2
fy~2A
'
(1 11
>
dy/
(ll12b)
2A YZ
'
2A
~b,
b2
b3
J*\
a2
a3_
i
<y/ 2
Vs.
dy,
(ll12c)
Here
[B]
is
Hence Eq.
(118)
becomes
By
OX
to =
(ll13a)
1
Ox
dy\
By
(ll13b)
Ox
Now
n,
Gh6 2
GhO 1 [[
2
+b
JJ
dx
[dy/
dyf]
Idy
dy] dy
dy/
JSx
dy/
dy
w
(ll14b)
\dxdy.
x]
n,
Ghd
r
{q,} [B]^[B]{q,}
(ff
[y m
xm ]
)dxdy,
(1
l14b) gives
t
y>
2{qwY\Bf
(1115)
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. (112) 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
(1116)
dy/ 2
dUp
dy/ 3
which leads to
GhO 2
[B]
[B]dxdy[%}
term
last
Equation
results.
JJ
is
JJ
[Bfi
(11 17a)
_M
dxdy.
(1
l17a)
= (<U
fcjfar}
where [kj
GhO 1
(ll17b)
[kj
Ghd 1
[B]
JJ
[B]dxdy
(ll17c)
and {Q^}
is
m=
They can be evaluated
Ghe>\\{w\_
6,
(ll17d)
\dxdy.
as follows
~b t
a,
o2
a2
6,
L^3
Since a and
ym
a
"3.
LM
b,
dxdy.
(ll18a)
hence
or
G/z0
fc
6
4^
sym.
a\
b2b 3
b\
+
+
a 2 3
a\
(ll18b)
246
Torsion
Chapter 11
and
b
>
0\
b2
a2
GhO
[Q,
jy
y,
dxdy
x,
(ll19a)
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
is
(ll19b)
common on both
lem
is
Step
and can
and the prob
is
isotropic
(115)].
Assembly
5.
Example
111.
(1
l18b) and
(1
l19b) can
now
all
G =
h
The
A =
triangle
=
=
cm
N/cm 2
cm,
rad/cm.
2
.
in
order to
quantities a t
Element 1:
x2 =
 1 = 1
2 = 1,
b = y2  y 3 =
= 1,
a2
62=^3^1=10 = 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
(ll20a)
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 115 Torsion of square bar. (a) Square bar and mesh,
(b)
Element 3:
a,
a2
tf 3
and * m
(2
= 1 2 = 1,
= 2 1 =1,
=22=
+ l)/3 = f,
b2
= 1 0 = 1,
1,
(ll20c)
=  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
(ll20d)
248
Torsion
Chapter 11
Substitution of these values into Eqs. (ll18b) and (ll19b) gives the following:
Element
1:
Global
(ll21a)
Element 2:
Global
Local
1
^3
^1
y\ = vi
2
L2
yi
(ll21b)
Element 3:
^2
Global
Local
i
^1
2
(ll21c)
L2
Element 4:
Global
Local
i
1
(ll21d)
3
2
2
we
obtain
249
7br^ K7/2
Chapter 11
>\
Global
"
(2
(2 
(2
2)
(2
2)
(2
(2
_(2
2)
2 
 2)
+ 44
(2
2)
:22) (22)
2)
^2
2)
(22)
2)
4
5
2)
(4
^4
4)J
i=o
'
+ f=4
(ll22a)
f + f =
+ 22 +
,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)
and [R]
is
is
Since the values of the warping function are relative, the boundary conditions
Then
4"
>
4
14
= cm
Step
6.
/rad (given),
y/ 2
8
(ll22d)
<
4
16.
y/si
0.
in
8'
>2
4
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 (119) 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 ^:
(1123)
^V:
0+1X2
1x01x2 + 2
iflX
i
HJH1
Element 2:
Element 3:
Txi
T...
Element 4:
The shear
TWISTING
MOMENT
The expression
Mt
for twisting
G6
moment
y
JI i~ fc
Often
M /G6
t
is
w ^ x2Jr
moment
bar
in the
is
given by
y2 dxdy
and
'
[2, 3, 4]
(11_24a)
is
in the bar.
If
first
Chapter 11
Torsion
251
 , *.  , [y.A$l + XmA(gj
+ *. + ,)}
(,1.24b)
xm
7(x\
gC^?
*1
x\
XjX2
*2* 3
Xi* 3 )
and
is
the total
number of
elements,
3 ).
tm
denotes
moment by element m.
V'V
Stresses
shown
at
centroids of elements
cm
approach.
The
twisting
total twisting
moments. Substitute
for dy//dx
as the
sum of element
(1112),
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
1x02x0] =
J[_l x (2)
^ = }[lx (2) M
tl
= \
0]
1,
1,
l+Jx(l) + =~t + f.
Element 3:
^ = J(lx2+lxO2xO)=l,
= 1,
x + x
^ = J(i x
2 4
0)
Element 4:
* =
jS!
Afr
iKl
JI1
1
X (2)
= i[lxOMx
= 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 Ncm.
(l
24
24)
Moment
M, ~0.1 406(70(2 a) 4
(1125)
>
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
torsional constant
l)
16
4
.
2.666; hence
is
is
em)r
inn
= 2.25002.6666 vXl0
12500
18.5%.
Shear Stresses
From
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
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
(ll26c)
where
a:
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 higherorder approximation,
shear stresses would vary from point to point in the element. In view of the
them
it is
often customary, as
shown
in Fig.
254
Chapter 11
Torsion
we compute
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
It
The
element are
this
= 0.667
and
= 0.000.
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. (1126). 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%.
mation, and the rather crude mesh, the computed results yield
realistic solu
STRESS APPROACH
The
[17]; this
Chapter 11
Torsion
The governing
torsion.
geneous medium,
in
255
differential
[1, 2], is
V^W ^
where
is
26
(1127)
known
function of
(x, y).
terms of
stresses in
>
Ty
<
between the
(ll28a)
>
dcp
TL yz
relation
are
d(p
1 X2
The
cp
5iJ
stress function
is
(ll28b)
*&+)
The energy function associated with Eq.
(1127)
is
the complementary
w =
n =u +
c
u 
w,
or
2
=JJi[ + (g)>^JJe^'
*"*
or
dx
n =
2b
dxdy
'
Here
tial
/c is
The unknown
T
[q s ]
[cp
cp
is
the
work of
= Nl9l
(p
where
IT (20)prfjcrf>\
(11 29b)
dcp
cp 2
cp 3 ].
N
The
N
2 cp 2
is
is
now
expressed as
[N s ]{q s ],
gradient^? relation
(ll4d)
is
dcp'
gj
{g}
2 <p 2
external loads.
>
dx
[B]{q,!
<
(1130)
dcp
g,
Substitution of Eqs.
n =
c
(1
l4d)
idyl
and
(1 130)
JJ {q s y[BY[D][B]{q s }dxdy
jj (20)[NJ{q,}^,
(ll31a)
256
Torsion
Chapter 11
where
1
P>]
is
By invoking
(minimum)
sn c =
(ll31b)
*S,0.
3
or
JJ
[BY[D}[B]dxdy{q,}
(20)[N]Wy,
JJ
(ll31c)
or
M{qJ =
(H31d)
{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.
(1132)
and
IN,
U2JJ N
The
integrations in Eq.
formula
(1
133)
2 }
dxdy
(1133)
can be evaluated
in closed
[4]
a!jg!y!
N\N{N\dxdy
(a
++y+
2A,
(11 34a)
2)!
where
[ N.dxdy
IT
For example,
N\N\N\ dxdy
1!0!0!
(1+0 +
2A
2)!
=
3
2A
X 2~
A_
3
(ll34b)
Chapter 11
257
Torsion
Hence
{<W
26A
(ll34c)
is
three nodes.
Example
112.
We now
The
Approach
properties
G and 9
are as before.
Our
cm x
cm
(Fig. 117).
is
(0,2)
(2,2)
\a/
a\a
/ A \
L_
(0, 0)
(2, 0)
_l
4< ;m
'"
258
Chapter 11
Torsion
element
stiffness
is,
how
41 9i
4
r
4
4
4
16_
<P*
.<Psj
.4,
(ll35a)
or
[K]{r]
[R]
BOUNDARY CONDITIONS
According to Prandtl torsion, the boundary condition
(p
If
we choose
(p
we have
0,
is
(ll36a)
(Fig. 117).
<Pi
<Pi
<^4
(ll36b)
0.
Equation (ll35a) is now modified by deleting the rows and columns corresponding to <p 2 (pi, and (Pi, to yield
.
L4
\(p
16Jl^ 5
(ll35b)
'
?l
= =
<p 5
<?l
2.66667 TV/cm,
1.33334,
and
?2
Step
6.
0.
Secondary Quantities
SHEAR STRESSES
The
[B]
stresses
can
now
(p.
Use of the
dp
dx
dcp_
J_P>i
2A i_fli
b2
b{
a2
a2
(1137)
.
[dy
Substitution of at and b t from Eqs. (1120) 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
dcp
dy
>
N cm
2
.
d(p_
Txi
Element
dx
2.
N 'cm
Element 3:
N cm
i:i
Element
:
.
4.
and
l4dj]
is linear.,
stresses in the
element
call this
MOMENT
TWISTING
According to the
stress

tpdxdy
= f
(p
moment M,
m dxdy,
is
given by
[2]
(ll38a)
'
where
m=
1, 2,
. . . ,
Af,
"a
M = number of
twice the volume under the stress function distribution over the cross section of the
(p
tu
we have
m dxdy
[Ntf 
N2f2 
Nsfftbcdy
yP?
Summing
(ll38b)
(P\
2x1/8
2 x
3
2 x
f
0+3
4
3
00
00
2 x
3
2 (32
7.1111 Ncm.
(ll38c)
260
Chapter 11
Torsion
Therefore, total
moment
M, = 4 x
Step
l8(a)
and
28.4444 Ncm.
(ll38d)
7.1111
8.
Figure
is
approximation
(b)
cm x
is
at the
element centroids.
approach,
(a)
stress
function
2.360
Exact distribution
of
\p
2.666
?<:
Section
AA
.2.666
3fc.
1.333
Section BB
(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
l26a)
is
f
Therefore, the error
2.360
is
err0r
2.3602.666
inn
X 10
1360
13.00%.
T = 0.246 AVcm 2
T
1.060.
error in x yz
= 0.246
= 25.00%
0.000
1.0601.333
26.00%.
=gg
inn
x 100
^~
~
From
0.1406(2 x 2)*
36.0000
cm 4
computed
is
= 36.000028.4444
=7
Jo
torsional constant
is
inft
x 100
^21.00%.
a finer
mesh
is
The computed
Hence,
if
used.
BOUNDS
Figure 119 shows the computed values of torsional constants from the warping
(displacement) and stress function approaches in comparison with the closed form
For
this
Example
113.
Figure ll10(a) shows a refined mesh for an eighth of the square bar in Fig. 117.
In view of the symmetry,
it is
261
262
Torsion
Note: The
number
Chapter 11
44.000
Warping function (Example 111)
42.000
40.000
38.000
36.000
8
10
Number
12
14
16
18
of nodes
34.000
Stress function
(Example 114)
32.000
Stress function
30.000
(Example
13)
Stress function
(Example 112)
28.000
26.000
now we have
bar instead of four as in Example 112. Figure ll10(b) shows local node numbers
for the elements.
In the following
we
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)
mesh,
(a)
Mesh
17. (b)
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
11
1
0"
421
02
01
02
210
021 4
(pi
03
<P*
1
[*
f
(ll39a)
2
2
<Ps
1_
\<P6.
<Ps
<Pe
0.0
=h
+4p 2<p = $,
 2<p 2 + 4^3 = ,
<Pi <Pi
<Pi
(ll39b)
<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
>
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
1.500
2.333
y^
(a)
Computed
(p
and shear
refined mesh.
COMPARISONS
Torsional Constant.
(1138)
r/7
2 [(
L\ 3
=
Then
The
is
'
4J5
3 )
'
V3
3/"
'
3.8888 Ncm.
XM, = 8x
3.8888
is
31.1 100
Ncm.
"*"
3 J
266
Chapter 11
Torsion
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. (ll26a), the closed form value of (p at node
in
Fig. 119
(x
0,
0)
<p
found to be
is
is
e rr or
2.360,
2.3602.333
1.00%.
(x
3,
5/3,
x lnft
100
2/3),
error int*,
error in x yz
= 0.194 = 19.00%
0.000
1.9001.500
21.00%
j^
inn
x 100
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
necessarily at the
in the
error.
The numerical
meshes. Such refinement should follow certain criteria for consistent comparisons;
these are briefly discussed in Chapter 13
Example
To examine
114.
Computer Solution
(Example
135)
for Torsion of
and
in Ref. 4.
Square Bar
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.
this
rilateral element,
which
is
covered
in
is
compared
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]
functions in parentheses,
(c)
Computed
shear stresses.
is an improvement in the
assumed approximation within the element.
Figure ll12(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.
COMPARISONS
Torsional Constant.
constant
The
error in 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
in Fig. 119
and
yields a
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
0.590,
is
error
0.5900.620
^^
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%,
0.2380.240
1.00%.
fyrvi
24
1AA
x 10
0.25)
 (0.100)
^^
0.104
error in i X2
'
4.00%,
0.8530.870
2.00%.
error in t yz
nfcTa
<p,
inn
x 100
x inn
^^0
improvement
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.
or warp
ing functions. These functions are chosen such that they satisfy physical
continuity of the
body or continuum up
and
its
Up
is
satisfied
Chapter 11
Torsion
269
In the case of the stress approach, the assumed stress functions that
complementary energy
II C
Its
Thus
in
compatibility
retical
We
assumptions.
procedures.
HYBRID APPROACH
Since
it is
it is
possible to
[9].
in the results
we
shall choose the elements on the boundary, shown shaded in Fig. 1114, 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. 1114). The stress
function is assumed within the element as
<P
=N
=
q>
+N
2 <p 2
+N
2 q> 2
(ll4d)
[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
y/
rad/mm
'a)
Hybrid
v^

44
1
[,
A.
i
!/
\/\/]/\/\/V
2 L
Displacement
(b)
edge AB.
270
[9].
111
Torsion
Chapter 11
Y
Boundary element
731
^Bou
W 7
CD
~^W
Inner element
=
",2
in
hybrid
approach.
y23
si
^12
^ +N
relation
y + (1  N )y/ + N y/
= N ll +
y2 + (1  tf )^
= (1  ^ Vi + ^2^2 +
y/
definitions of
+N =
3
Nu N
and
(1140)
as in Eq. (113)
and the
1.
As suggested by Yamada
et al. [9],
we
shall also
their procedure.
Step
4.
Element Equations
we
[9, 11]
u = ue + wph = u ch
h,
"AJ['+(K)>**iA
= iHere
jj
{*}
\P]{"}dxdy
j {TYMdS
(1141)
we have used
the subscript
is
ch it
= [M
r 23 t 31 t 12 ]
stresses
is
on the boundary,
{\\f b
T
}
moment
[6 y/ 23
272
Chapter 11
Torsion
Wn
is
W\i\
defined in Eq.
is
(1 l28a),
0"
[D]
1
which
is
We
assume
linear elastic
behavior.
evaluate
for the
is
Uxz
w=
dy
cp 2
(ll42a)
lies
b,
dx
a3
d(p '~2A
[*
boundary,
a2
01
= constant,
<p 3
specializes to
f*l
f
"
(ll42b)
<*>.
or
}
(ll42c)
[PHK,
where
[P]
and
{P}
= {<p,} =
<p
is
2A\ b
=m jj^ffow\dxd m
T
(ll43a)
wmm
2
(ll43b)
<o
X 2A X 2A //
,
'a
_Ari (a] +
SA 2
G
where
l\ 3
side 23
= (a +
2
b\)
3U:H
b}) _lh(t>\
(ll43d)
&GA
= (x
3
x2 ) 2
(11 43c)
f (y 2
2
3)
and
[H]
=
_
JJ
\PYtPmdxdy
*23
4G^
(1144)
Chapter 11
Torsion
in
273
and can be
expressed as
IV. ,
M.(
7r.Vr.dS
'iiViidS
r
y,
'^ dS
(ll45a)
(ll45b)
;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.
(ya
f4
xmx^dxdy
(1146)
and by
~.
~_
(a.b; b.a.)p.
2Al ::
x: sin
r,,
cos a
'
'
'
_ (a.b:
a b
imponant
'
It is
b a
:
(1147)
l
ba
)(p
0.
normal
to the
boundary
\f
b;
identically
xjt
y.i
is
we have
Ul
1 f3
(114&0
r::
*12
hi
{f}
3 /::.
::
The
performed
Eq.
1145
easily.
For
{\f/ b }
u :: from
.
W+
can be
gives
because JS
di4Sb)
[R;p
dN
l23
lNJvt + N&JKhs
along side
(ll49a)
of the expression
274
in
Torsion
Chapter 11
J2
y/ 23
= k^{y/i +
dS
(ll49b)
Vi\
Then Eq.
(1145)
becomes
W = {tY\{y }dS
ph
(ll50a)
(P}Wy
(ll50b)
\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
(ll50c)
m [G]{%] = {P}W[L]{q,}.
T
(ll50d)
Here
[G]
(ll50e)
PT[L],
and
2y m A
{L}
Now, we
relations a
substitute for
a2
and
b2
b3
a2
a3
(ll50f)
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.
(1 151)
ch
to zero,
ch
we obtain
dUeh
[HP]  [GHqJ
0,
(ll52a)
or
[H]{P)
(ll52b)
[G]{q r ],
or
}=[H][G]{q r },
which expresses the relation between nodal
ing functions for the boundary elements.
(1152C)
stress functions
Element
Matrix
Stiffness
We now
approach. For
we
this,
P = tf*FhK*}where
{q w }
= [6
y/ x
y/ 2
is
y/ 3 ]
(H53a)
e,
after substitution
m
=
u =
c
\n\m
Mq,}W([H]y[H][H][G]{q,}
{qjqGF[H]^[G]{q
2"14^
(1
144)
is
(ll53b)
usually symmetric,
we have
([H]
_1 r
)
= [H]'.
Equating the above two expressions for
stiffness
matrix \k h ] as
[k*]=[GF[H]i[G]
[Gf
AGA
(ll54a)
[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~ )
j(2A)(2A)
j(2A)(2A)
\(2A)(2A)
(ym ai +
x m bi
x m bi)~
(ll54b)
1
01
a2
2A\
xm b\)[
^(2A)(2A)
relations
= 0,
f b 2
63
= 0,
and
a 3 b2
a b2
a2b 3
2A.
Inner Elements
By following
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 ]
relations
b 3 ym)
+ 01^3
+ a 2a
+ a\
(1155)
dis
and
M{qJ =
{Ql
(11 56a)
WW =
(Ql
(11 56b)
where
Ay m
~^*m
&i
flj
{Q}
'M,'
^2
2
*3
*3
Ft
ft
(ll56c)
F2
F3
1
and
(1142)
2A
_:;h
AL;;Hw
(0
a
l\i\_
{a
^ b x x m )
l ym
b {a y m
x
b
xm)
0i
(1157)
6,
3
Assembly
For elements on the boundary and inside the bar, we use element equaand (ll56b), 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 (ll56a)
Chapter 11
277
Torsion
we
puter solutions by
Example
We
Yamada
hand
calculations
and then
results
first
from com
et al. [9].
115.
cm x
cm shown
example the local node numbering is changed as shown in Fig. 1115 to comply with
the numbering for the boundary element in the foregoing hybrid formulation.
Global node
Local node
/j\ Element
(0, 0)
(2, 0)
lie
boundary elements;
it is
(1
It is
in general,
l54b) and
will occur.
Hence
(1 155), respectively,
matrices for the boundary and inner elements. In the following are given salient
details of the
Element
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
G=
N/cm 2
i>
278
Chapter 11
Torsion
to element equations:
2<
3<
Mi = ~
Local
(0)
Mf,
f
}V1
2
1
Lf
is
2\
We assume that
Global
lj ly 3
is
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.
of nodal
y/'s
208
12
12
3
12
3
1
1
12
T
1
W1
1
4
0_
the solution
now
m=l
tm )
1
1
1
.s.
,
,
y/^
0.
Then
11
with
is
= 1,
1
(prescribed),
2 =2,
s
*
s
These
=
=
=
2,
0,
0.
We
same
as in the case of
have considered
Example
this rather
simple example
mainly to
illustrate
SHEAR STRESSES
For the four elements
at the
stresses:
Element
1:
2(2 x
Oi
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 111 (Fig. 116); they show that
boundary is zero. It was only incidental that with the
fourelement 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
computed
results for
281
Torsion
Chapter 11
M^ [fxl
=
This value
is
same
the
1.7777 Ncm.
as that
from the
stress function
of about 21.00%.
Example
116.
method
is
is
presented by
discretized.
Yamada
The
et al. [9].
Due
to
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
AB
[Fig. ll13(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
111
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. ll13(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
methods. The
latter
boundary but
TABLE
112
results
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)
AB.
(b)
Shear
stress xyz
[9].
(a)
Shear
stress
283
MIXED APPROACH
In the mixed formulation, both the displacements (warping) and stresses
(stress function) within (including the
to be
unknown. Hence,
[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
(ll58b)
(q c
3
= [i
(ll58a)
3]
lyzt
Tyz*
*yz 3 ]
is
is
is
and
is
0"
(1159)
1
Step
We
4.
[12, 13].
expressed as
n* =
Noor and
mixed approach can be
procedure presented by
dU
the
w w  *ww  JJJ^. 
is
[f xz
If
vm^ds,
(i
leo)
Si
dU
{a}
essentially
[4, 12]
JIf
V
where
follow
shall
Anderson
fy2 ]
is
X2
TJ^J ^
*XM
dV
is
the vector of
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.
is
x\VN N N
2
~~
yJ
and
[B]
is
Nt
(1162)
N Nj
2
The
(1161)
its
stationary value.
and simul
taneously with respect to nodal stresses and nodal warping functions; thus
dUM
<5ru
Wq.
(1163)
<?{q
fc]
ftJ'
LfcrF
{Q.:
{q.
J {q
[0]
(ll64a)
{0}
fcj
(6
JJJ [NJ
(6
6)
PU = jjj [NJ r
(6x3)
(6
[B]
2) (2
^^z,
[NJ
[D]
2) (2
2) (2
(ll65a)
6)
(ll65b)
dtofy<fe,
3)
and
{Q.}=0jJ{N.f
(6
1)
(6
2) (2
^{4
[N]
6)
(6
(ll65c)
1)
286
Chapter 11
Torsion
stresses
tangential
last
We now
illustrate
*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
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
wfj
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
117.
the
I lar
:Mixed Approach
(2x2
cross sec
288
Chapter 11
Torsion
of the computations of element and assemblage matrices based on Eqs. (1164) and
(1165).
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
(ll64b)
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
=0,
bi
=2,
"2
=
=
1,
b2
1,
b3
=
=
=
a3
Xm =i,
ym
1,
1,
1,
(7, 8, 9)
(
289
Torsion
Chapter 11
T xz3< r vz3< ^3
(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
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,
inter
element compatibility.
It is difficult
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
is
Problems are
exercises in
left
to the inquisitive
and advanced
reader.
is
STATIC CONDENSATION
Often
it
may be
particularly
single
when
element
is
it
is
nodes. For instance, Fig. 1118 shows a line and a quadrilateral element
line
primary nodes and one inner node, and the quadrilateral has four corners
or primary nodes and one inside node.
(a)
lateral.
Inner node
'a'
292
Chapter 11
Torsion
The
finite
element equations
[k]{q}={Q]
are derived
on the
(ll66a)
freedom
Thus
by the use of the inner node. In the case of the quadrilateral [Fig. 111 8(b)],
the element equations are often obtained by adding individual element equations of the four
unknowns
in {q} are 3
unknowns
this
in place
is
at the inner
node
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. (ll66a) in a partitioned form as
[q,}
IK]
=
""
fCQJl
(ll66b)
ItQJJ
{q,]J
{q p },
at the
{q,}
= [k,,]'({Q,) 
dk,J
first
[k,,F{q,}).
equation of Eq.
frJfcJfcjFXq,}
= {Q,} 
(1
(1
l67a)
(1
l67b)
l66b) leads to
fcJfcJ'tQ,!
or
where
[k]
(stiffness)
vectors,
if
desired,
it is
possible to retrieve
Chapter 11
the
unknowns
static
293
Torsion
at the inner
many
[4].
PROBLEMS
Invert the matrix [A] in Eq. (112) given by
111.
in
x2
x3
yi
yy
EHB
Eq. (ll3b).
112.
the
bar.
twisting
113.
114.
moments.
in
in
approach.
115.
tion approach.
\A
A\
a\
A
\A\A \
A\
(0, *)
'
\A
\*,
*i
*
(0,0)
G=
6 =
(2,0)
Figure 1119
116.
117.
Solve Prob.
8 as
4, 6,
and
This and some of the other problems will require use of an available equation solver once
294
118.
119.
Torsion
Chapter 11
Consider the mesh in Fig. 1120 and solve for torsion by using the stress
it may be necessary to use a computer. The code
FIELD2DFE can be used for solving the entire problem. Alternatively, the
(2,0)
(0,0)
Figure 1120
1110.
in Fig.
Figure 1121
Chapter 11
1111.
For the torsion problem in Fig. 117, 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:
1112.
295
Torsion
accuracy.
<p 5
0.6667;
M /G9 =
t
1.777, error
21.00%.
For the circular bar of 2cm diameter, compute nodal stress functions for
mesh shown in Fig. 1122. 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 1122
1113.
For the
bar a
elliptic
= 2 cm
code.
Assume G
and 6
cm
Figure 1123
1114.
By
using the warping function approach, find shear stresses and torsional
constant for the triangular bar divided into two elements (Fig. 1124).
Assume G
and
296
Torsion
Chapter 11
^
(0,0)
Figure 1124
1115.
Use
the stress function approach to solve for torsion of the triangular bar
inProb. 1114.
1116.
By
FIELD2DFE
mesh
and obtain
vs.
with a triangular element (6 nodes) and derive the element matrix and
Assume
W
Note
tyl
=
<P2
all
[N]{qJ,
9*
<P*
<P5
terms up to n
9s\
2 in the polynomial
Figure 1125
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
(4L31)63
l)a 2
(4I 3 l)fl3
4{IiZ)2I:6i)
4(L z b 3
L
b2)
4(Z. 3 6,
+,63)]
and
fc]
= [(4L,l)a,
(4I :
4(I.,<i2
+2*1)
rele
REFERENCES
[1]
Elasticity,
Dover,
New
York,
1944.
[2]
[3]
Timoshenko,
York, 1951.
Herrmann.
S.,
and Goodier,
New
J.
91,
[5]
[6]
Murphy.
R., Elastizitat
to the Finite
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,
286295.
[8]
Yamada,
Kawai,
Y..
ElasticPlastic
[9]
Yamada.
Y.,
Nakagiri.
S..
298
[10]
Chapter 11
Torsion
Yamada,
Nakagiri,
Y.,
S.,
"ElasticPlastic Analysis of
Int. J.
Num. Methods
Mechanics,"
in
Advances
in
in
Continuum
New
York, 1972.
[12]
CM.,
Noor, A. K.,
(private communication).
6,
for
1975,
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)
kx
>%)
l + kjjgt, +
x
= p(t)
k^Ll,
<*(<p
on S
cpo)
(122a)
iff)
on S 2 and S 3
0,
(122b)
Here
cp
is
the
unknown
299
Potential, Thermal,
300
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 twodimensional steadystate problems; that is, the problem is independent of time and the righthand side of
Eq. (12la) vanishes. Also, for simplicity, only homogeneous materials are
kx %g
+ k*j* + Q =
to
=f
q>
(12lb)
on 5,
(122c)
and
kx
/x
lx
+ k,fy, + q =
(1 15)
and
(1 127)
on S 2
(122d)
POTENTIAL FLOW
The
is
dx 1
dy 2
or
V> =
(123)
0.
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>
the
same manner
respectively.
as
the
y/.
problem
(p
and
is
y/
similar to
are used in
Chapter 12
The
Potential, Thermal,
relations
301
q>
and
y/
are given by
[2]
dx
(125a)
and
57
(125b)
tfy
"' =
Substitution of i\ and
p, into
Eq. (123).
Boundary Conditions
is
and
the unit
Vw
(126)
normal
and the
wall, respectively,
and
0
d(P
n
Sv
(127a)
S,^'
Figure 121
From
Flow
in pipe.
Fig. 121.
dx
,
dn
dv
_
~
dy
(128a)
ds
dx
(128b)
Potential, Thermal,
302
Hence Eq.
dx
dx dn
is
Chapter 12
(127a) transforms to
dtp
which
,dydy
The
3).
the flow or
is
(127b)
0,
dn
dy dn
Neumanntype boundary
boundary condition
potential or Dirichlet
is
<p
onS
=<p
(128c)
Often both the flow and potential boundary conditions occur together, which
is
called the
FINITE
mixed condition.
ELEMENT FORMULATION
For twodimensional
idealization,
we can
one described for torsion in Chapter 11. Hence, we present a formulaby using a fournode quadrilateral element (Fig. 122). It is possible to
to the
tion
use either
(p
or
y/
We
first
con
(p.
(x 4
y4
M.+1)
/T
@7( +1
(1,1)
<
+1
^^^_~/^
(x 2
y2
>
(+1,1)
~
The velocity potential is a scalar and has one value at any point. For the
fournode quadrilateral, there are thus four nodal degrees of freedom, and a
bilinear
model
for
q>
at
(p
= g, +
= mm*
CC 2
<x>
&*xy
(129a)
or
(129b)
where
[<J>]
coordinates.
Chapter 11;
pt
(Pi
^3
^4
in
however. Evalua
nodes yields
at the four
<p
303
[1
tion of
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
(1210a)
4i
or
{q,}
where
[A]
(1210b)
[A]{a],
is
^4]
We
r
{q,,}
{*}=[Ar{q,}
and
and
(12lOc)
The product
[<D][A]
(1211)
and
N
N
N
=i(ls)(lt),
=#l+s)(lt),
which
by
(1212)
=i(l+s)(l+t%
N<=(ls)(l+t).
3
Here,
s, t
forty,
The
1,2,3,4.
global coordinates x,
at
any point
in the
i9
S Nx
t
(1213a)
= 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,
(1213b)
y 3 y4].
unknown
N _ \{y
Potential, Thermal,
304
is
cept offers a
Chapter 12
number of advantages
As
in
1, 2, 3, 4.
integrations.
in
satisfies
the
problem governed by Eqs. (121) and (1214). It does not, however, include
all terms in the polynomial expansion, see Chapter 1 1
Step
4.
Use of
problem gov
erned by Eq. (123) will yield essentially the same results. We consider the
following variational function for the twodimensional idealization
JR[@' +()>*
We
d Jt
d
\_ JP.
2 dx dx
and
i^z
2 dy dy
(1214)
Chapter 12
Potential,
305
are similar to those in Eqs. (321) and (46) and can be considered to represent
a measure of energy.
The
of
tp
x and y
_
~
gx
dtp
dx
_
~
dtp ds
ds dx
_
~
dtp
dtp dt
~di
dx
(1215)
dtp
g>
Since the
dy
ds
dtp^dt_
ds dy
dt
dy
s, t,
we
use the
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'
d)
id]
dx
[*1
Ts
dx
dy
ds
ds
dx
[dt,
Tt
dy
d
Tt_ Idyl
>
The matrix
[J] is
dx
(1216b)
[J]
d
Idyl
Id]
dx
>
[J]i<
rule gives
>
dt
dy
dyl (d\
dt
ds
ds
dx
dx
(1217)
dt
Jl
The terms
in
dx dy
~dJdI
{dt,
dx dy
(1218)
dt ds'
57
ds_
Jacobian,
[Eq. (1212)].
For instance,
"
5)(1
i(l
0*4
(1219a)
306
Potential,
Chapter 12
i(l
ds
(1219b)
evaluated as
/ is
iisitewf")**")]
(1220a)
By
setting
4 and j
1, 2, 3,
1, 2, 3,
4 for each
the
i,
summation
in Eq.
\t s +
l+t
1+j
st \s
15
1 t
S +
l+s~
X
"gLM
Xt
X,
Xi
st
yi
l+t
(1220b)
Expansion of /
4)
3)
4 )]
i}
Use of Eq.
(1220c)
and dt/dy
in Eq. (1215) as
o
ds
dydk_
(iyis
dy
($*N
ai . 2 u\
Similarly,
dy'
dt
?i
from Eq.
(1211),
fdN y \
= ~\T[\k~~SF
1
dy \J\\M
Now
(1221b)
Xi
\J]\&i dt
t
(1221c)
(1221d)
Xi )'
ds
we have
dN
% = 
dtp
m
and
dq>
dNim
f^fff*
(1222)
Potential, Thermal,
Chapter 12
and (1222)
307
d<p
dx
(1223a)
dtp
&
The
indicial notation
instance, the
setting
first
and/
dN
m V(dN
x
+
,
When
(dN dN
1, 2, 3,
dN
6N,
(dN dN2
dN,
is
obtained by
dN
2\
dN dNA
~W) y \
(ON, dN,
W
x
this is
1, 2, 3,
row
"
3\
first
4 as
dN, dN,\
\\T\\
in
Xi
setting
4,
(1223b)
=2
we have
and j
d<p/dx.
#12
#13
#14
"23
B"24
M{q,},
#11
B,21
A
rr
"22
(1224a)
or
{g}
(1224b)
where
#u
8/
0>24
^34^ y23*)>
#12
#13
8J
(^13 +^34^+^140.
8T7t(
l
#14
8/
>
24
+y^s
yut),
Ol3 ^12^+^230.
(1224c)
#21
8/
8/
\X 13
X 34 S
87
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).
308
Now, we
np =
Qp
(1225)
l\lBY[H[B]dxdy{q,l
{q,}
Chapter 12
where
[I]
0"
[C]
1
en,
that
<p
we
obtain
(1226a)
0,
is,
d<p x
dQ,
=
)==> sq p
(1226b)
o,
d<p 3
dQ,
which leads to
r
{H} JJ[BnB]rf^{q,}
(1227a)
or
\\lW[R]dxdy{%}
(1227b)
or
Mi,} =
where [kj
is
(1227c)
("I.
[kj
jj[BY[B]dxdy.
(1228)
Numerical Integration
The
coefficients
k9ll
= ffah + Blt)dx4y
JJl(8J
(j 2 4
y 4Sy 23
3
t)
)2
</

(^24
*34*
X 23
]^^
(1229)
Potential, Thermal,
Chapter 12
and so on.
It
can be
difficult to
and
309
it
often convenient
is
efficient to
P F(x)dx = 2 F(
)W
Xi
(1230)
t,
1=1
Jxi
[K]
f_JjBY[B]\J\dsdt
m
_n
^S E[B(*f
j=l j=l
y )F[B(*f,)]
/(*,,/,)
W,0,
(1231a)
where
Wt
[Bfe ORBfe,
Example
It is
121.
t t)]\J(s t , t t)
(1231b)
t.
tables
[3].
Numerical Integration
many
finite
element applications.
it is
used in
integration over a (square) quadrilateral [Fig. 124(c)]. In view of the fact that
hand
calculations of
all
terms of matrix
[k,,]
;
ku
inEq.
From
(1229).
Eq. (1220c):
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 =
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)
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
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
(1X0)]
r[(0Xl)
 (OXDD
square, which
sfdXO)
constant for
all
its
Note
value
is
Now, numerical
integration in
*,n
=S
r=
?ife, U)
tt)\
{,
1, 2, 3, 4.
From
[3],
we can
t.
tabular forms
B\\
B2i =
(?24
87( A'24
J34S"
>'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)
577 ) 2
(0.423) 2
= ^^ =
(1577)2
Chapter 12
(0.423)2
(1.577) 2 ]
0.667.
manner.
Assembly
5.
Equations (1227) are assembled such that the potentials at common nodes
The
are compatible.
final
= {0}.
tKJ{r,}
Under
(1232)
(1232)
(p t
Applied Fluxes
Then
the
2?
dn
If
we provide

for
and
q,
(127b)]
(1233)
0.
becomes
Use of the
becomes
fl^  J>*
<
12  34 >
WW =
C235a)
{Q3.
where
{Q}
JJ
[NY{Q}dxdy
[Wl$ds
(1235b)
=
is
{Qi}
{Q 2 }
Chapter 12
Potential, Thermal,
{Qi}
\\lW{Q}dxdy
Ni(st
N (s h
N (s h
N4(s
Here we assumed
ii
[W{Q}\J\
ti
dsdt
)\w
t t)}
t t)
tt)
i9
313
= i:ms ,t )Y{Q}\j(s
i
Q\J(s
,t t
)\w
(1236)
t t)
to be a uniform flux.
relevant to the
is
boundary
only.
As an
illustration,
N
N
{Q 2 }
(1237a)
q x dS.
We
1,
i(l
and
s varies
 s\ N =
2
{(W
(1
from
12,
w
*)
*)
coordinate
to
+ s), N =
10
qx dS =
K 1
N =
{
*)
fcl'IrfS.
to
(1237b)
Potential, Thermal,
314
Since s
= 2S/l
l9
we have
dS
where
is
lx
Chapter 12
ljds
\J\ds,
(1238)
{Q 2 }
f,
s]
(1237c)
>q x ds,
which reduces to
(Q 2 }
(1237d)
This indicates that the total flux on side 12 in the x direction
equally between nodes
the
x and y
and
2.
is
divided
(p.
The
differential
g+$=
as an
y/
is
similar to that
vy = o,
or
unknown
y/ is
(1239)
is
am >*
+
(1240)
Stream function
y/
can
now be
V=[N]{q}
By following
expressed as
N,.^.
(1241)
are
(1242)
[Mq,}={0],
where [kj
is
and {%}
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
(125).
potential approach,
Pi
^11
#12
^13
#14
.#21
#22
^#23
#24j
fi
(1243)
(Pa)
and
V:
#21
#22
#23
_
L #11
""#12
_ ^13
Vi
^24
~~
(1244)
#14.
Wa)
now found
can be
a section
AA
in
an element
(Fig. 126)
as
QfM =AV
(1245)
n,
where
the horizontal.
The
is
Vy,
cos 9,
fj
(1246a)
2,A V
t
(1246b)
AA and
can be found as
is
the total
number of elements
315
Example
Potential
122.
fluid flow
Figure 127 Analysis for potential flow around cylinder, (a) Flow
around
(0, 0) etc.,
mesh
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
(46)
(gl)
form solutions
Potential, Thermal,
Chapter 12
(p
317
upstream boundary,
nodes 15,
downstream boundary,
nodes 5155.
(1247)
Appendix
4)
was used
NTYPE =
2 in the code.
tions
from
domain. Table 121 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
0)
<p(x
8 [Fig. 127(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,
(1248)
0.50,
in
the negative
x direction
domain
and 6
is
=
=
the angle
axis.
TABLE
121
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.
(b)
2DFE
NTYPE =
3 in the
code FIELD
permits solution for the stream function approach. For this solution, the fol
W =
y/
nodes 555,
(1249)
Figure 129 shows the distribution of computed values of y/. Table 122 gives
comparisons between closed form solutions and numerical predictions for y/ along
the line of nodes 2630.
For the
(0.5
line considered,
0.0)/4
0.125.
solution
y/
= u(r y)sin#.
90 deg, and
(x 2
is
given by
[2]
(1250)
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
Chapter 12
122
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.
or
1
The numerical
+~
(1251)
The comparisons
in Tables 121
1210
show
satisfactory comparisons.
The
velocity
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
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
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
,
T
k**
+ k*
dx
dy
y
2 =
(1252)
0,
T= 1
on S
(1253a)
and
kx
tx
d~x
where a
The
is
>o\
C>
ol(T
T )q =
onS
and 5
(1253b)
finite
will result
term in
i<x(r
which
 T YciS.
(1254)
For evaluation of
[kj
{<2 3 }
[k a ]
a[N] r [N]</S,
(1255a)
*\Nr{T
(1255b)
we need
]dS.
= ^^lW[K]dS
c
[K]
*
i(l
*)
>[(ls)
i(ls)
0]dS
Potential, Thermal,
322
i(l
i(l
* )
i(l
Chapter 12
s )
s)
dS
0.
"W
olI x
o
(1256)
Computation of {Q 2 }
is
(M +
where
{q r
f  [^ T T
Example
As a
IW){qr}
2
123.
= {Q} = {6,} +
{Qi}
{Qil
(1257)
rj.
12 nodes.
=
7X3, y) =
7X0, y)
100 deg,
(1258)
25 deg.
The thermal
FIELD2DFE.
TABLE
123
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
323
3 units
SEEPAGE
Seepage
is
is
meanings of
Unknown,
(p
k x ky k 2
,
is
(1259)
directions, respectively,
Q =
q
medium
[4].
we encounter
conditions.
When
1212(b)],
we have
the
q>
[Fig.
(1260a)
and
**
911
= o.
(1260b)
Potential, Thermal,
324
Chapter 12
U
(a)
Free surface
Surface seepage
0,0,0
(2.0)
(0.0)
:
M: :0

(2.0)
(0.0)
,a5,
(c)
(c)
Steady
element solution,
it is
is
finite
can consult
Refs. [4][6].
Example
124.
Figure 121 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.
unit.
q
O
=
=
are
=
=
atx
at
at
<p
525
0.
(1261)
2 units.
(1261)
parentheses.
velocities are
shown
in
TABLE
124
Element
0.005
0.45
10"*
0.005
0.1S
10*
As expected, the
:
=
S
dx
0.005
0.005.
\2
;.
Example
V,
'.,
to zero.
125.
The foregoing problem is rather simple and can permit hand calculations.
consider a problem that is more difficult and needs use of the computer.
(dam. sheet
pile)
shown
is
kx
k:
this
included in the
The
.'
we
of a structure
assumption
finite
soil
Now
is
itself is
assumed
to be
impervious. The twola>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.
sides, respec
a =
g, =
m.
(1262)
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
<p
<
3<
(b)
^z
7^
V//////////////A
"
i
\
\
\
i
\
1
50
40 30
20
10
Percent
(c)
326
52"
Potential. Thermal,
Chapter 12
m (equal to the
Fig. 12
ten! ia
ill
be app:
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.
needed
>r
at the interface
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
medium
is,
al
mis bounda:;
Since the
:
Then at the centerline we can assume a natural boundf/dx = 0. This is valid since at the centerline the rate of
[Fig. 1213
it
From
on the
structure.
Moreover, the
we :an
finite
easing
uplift
at selected secticr.s
AM
this
information can be used for analysis and design of structures founded on porous
soil
ELECTROMAGNETIC PROBLEMS
In the case of steadystate electromagnetic problems the governing differential equation, in the
absence of
in
equation
or
if
kx
A.\.
Vr =
where
is
(12
0.
V:
is
publications
Thus,
[7, 8].
phenomena
in
328
Potential,
Chapter 12
COMPUTER CODE
finite
element
FIELD2DFE
This
is
field
Problem
Code,
Torsion
NTYPE
1
Potential flow
Velocity potential
Stream function
Seepage
Heat flow
In
fact,
the code can be used easily for other steadystate problems such as
and magnetic
electrostatic
Appendix
4.
PROBLEMS
121.
Derive in detail
(see
122.
Chapter
[k]
and {Q}
11).
if
respectively.
123.
By using numerical
integration,
compute the
coefficients
for
By hand
calculation, find the element matrix [k] for steadystate heat flow
T(x,y)
100 deg,
7X2,>0=Odeg,
and ky
ky
unit.
Potential, Thermal,
Chapter 12
329
t(9)
(0,2)
(2,2)
A,
A A
(0, 0)
(2, 0)
Figure 1214
125.
Tat nodes
4, 5,
and 6
shown
in
Example
50 deg.
126.
5,
and
6.
Q = 0.1
Hint:
units at
You may
node
and find
evaluate {Q}, which will need numerical integration similar to that used in
Example
127.
121.
soil
kx
and applied
fluid potentials
ky
(p{2, v)
128.
129.
(c)
=
=
nodes
10 cm,
5
cm.
4, 5,
and
and
6; (b) velocities v x
and
vy
heads at nodes
1210.
and
cm/sec
0.1
equal to
(p{0,y)
Compute
with
cm 3 /seccm
Q = 0.1 cm
at
fluid
6.
code,
330
Potential, Thermal,
1211.
Chapter 12
shown
in
Fig. 1215.
Insulated
T=
T=
100 deg
0.0 deg
8 units
Figure 1215
1212.
By
using
FIELD2DFE
Figure 1216
1213.
Derive the
finite
field
(Fig. 1217).
problems by using an
eight
Potential, Thermal,
Chapter 12
331
(1,1)
(1, 1)
(1,0)
(1,0)
,
(1,1)
(0
Figure 1217
Partial results:
Assume
u
/
[A7"=<
where
[iV]{q},
i(l
5)(1 /*)
id 
{qf
/1(1
&"<
Usee)f
u3
u4
u5
u6
t)(2s
JJ
5)(2/5)\
1(1
5)(2r5)
t)
i(l
5)(2r5)
i(l
pr
=<
t)
y
/(1J)
then leads to
[brc][B]^
and
{Q}
5)
)
etc.,
id  *
"/(I ~ *)
2
id  *
t)
S )(2t
U B ],
/i(l
u
t)
t)(2st)
[k]
f)(25
u2
[i
t)(2st))
#1 1(1
1(1
**xi
id 5)(1^)
>>
\\[XY{Q}dxdy
[NF[
REFERENCES
[1]
[2]
[3]
Streeter, V.
L.,
to the Finite
New
York, 1948.
[4]
Desai, C.
and Christian,
S.,
[5]
Desai, C.
J.
T. (eds.), Numerical
New
Engineering, McGrawHill,
Methods
Desai, C.
S.,
[8]
98,
M.
IEEE
in
Geotech. Eng.,
/.
Soil
PAS
FEM
Geotechnical
S.,
in
York, 1977.
PAS
Trans., Vol.
P.,
IEEE
Trans.
BIBLIOGRAPHY
Desai, C.
S.,
ments
"Finite Element
in Fluids,
kiewicz, O.
Doherty, W.
C,
P.,
Vol.
(eds.),
Methods
for
Flow
in
Porous Media,"
Wiley,
Wilson, E.
L.,
New
J.
T., Taylor,
and Taylor, R.
in Finite Ele
C, and
Zien
8.
L., "Stress
Analysis of Axisym
Ergatoudis,
I.,
C, "Curved,
No.
1,
Int. J.
Isoparametric
1968.
332
TWODIMENSIONAL
STRESSDEFORMATION
ANALYSIS
INTRODUCTION
number of onedimensional problems and twodimensional
problems with only one unknown or degree of freedom at a point, we
are now ready to consider a different class of twodimensional problems. This
class involves analysis of stress and deformations with more than one degree
of freedom at a point.
After studying a
field
PLANE DEFORMATIONS
Plane Stress Idealization
beam and
compared
The
thickness
is
334
y,v
Chapter 13
y,v
pJ
(a)
(b)
membrane
referred to as inplane or
(stretching).
Under
these conditions,
is
constant,
it is
z,
that
is,
and
across
T xy This idealization
.
nonzero
stress
is
components,
(131)
x,
only.
(132)
[y,
directions, respectively.
If
we
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
v)'
Thus
,2'>'>
(133a)
Chapter 13
335
or in matrix notation,
w=
rate]
1
],
(133b)
v
2
(Fig. 132)
w=
and a z
= v(a x +
a y). The
(134)
is
expressed as
1v
W=
v
1
[C]{]
(1
v)(l
v
2v)
sym.
(135)
2v
y, v
z,
y~^
x u
'
UL
(b)
(a)
Long
underground tunnel.
Axisymmetric Idealization
its
centerline axis
and
subjected to a load symmetrical about the axis. In view of the symmetry, the
336
'
Chapter 13
'
>z,
[2]:
'Or
^r
\>
{}=rT
(136)
[yj
Dr this
case
[C]{]
(1
v)(l
is
V
V
V
1
2v)
sy m.
v
[el
1
(137)
2v
StrainDisplacement Relations
From
we can
idealizations [2]:
du
dx
dv
dy
t>
du
y xy
(138a)
dv
\dy
dx
Axisymmetric:
du
dr
<
(138b)
<
dw
dz
du
Initial Stress
initial
dw
dz^
',
dr)
in
Chapter
5, it is
possible to include
is
applied.
The
of temperature,
where dT
may be
effects,
= T T
is
T
\
*dT,
(139)
and a
is
the coefficient of
thermal expansion.
We
sum of
where
{e
} is
strain case
[C]"
= [q({e}{e
which
[D], in
{},
(1310)
and
to
The matrix
=[C]" '{}
= [q{}
})
(1311)
plane
is
v
1
[D]
v
(1312)
337
ELEMENT FORMULATION
FINITE
As shown
in Fig.
dimensional elements such as triangles and quadrilaterials (squares, rectangles, trapezoids) in the
(z) is
generally included
stress.
y.v
P(x,y)
(L 1( L 2
,L 3
P(x, y)
(s, t)
ments.
been covered
in
Chapters
11
element.
As
v(x,y)
We
= i + &
= fi F
fi 2
+
+ Prf +
r
cc 3
cc 4
u,
u,
v at a
v at a
xy,
(1313a)
fi A xy,
or
{u}
where
338
ju}
[u
v],
[a}
= [a
/1
(1313b)
[<D]{<x},
cc 2
a 3 a4
f} 2
fi 3
fi 4 ],
and [O]
is
the
Chapter 13
~i::.\
::'
Di'j
.oD:
coordinates. Evaluation
2;X
Pz x
P:
2:V.
Pis
:::
':Yr
oces 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
(1516)
= K1  *Xi + 0
The geometr;.
1515
ons hV.
"*'"*
dehned n Eg
V;
tne vec
:s
,.
matrix of nocal
"A"
V;
5.
v =
Here
1514:
\v v
the square iS
:s
the
::.
;.
and 154
coordinates
i:
any point
erpolation functions
r=VYx
ine
emeni. car
1=174
il3l"ai
V
"
'
::
il5lb>
(2x8)
(2x1)
wnereW = [x,
This leads
xa
:o :ne
jc3
xJandfoF
dehninon
o:
(8x1)
>, y2 yl
:he elemen: is :
dement.
Requirements for Approximation Function
polynomial form as
is
in Eq.
sutisrieu since
>
1515
we have
z r zy\i: r ::
Two Dimensional
340
Chapter 13
is
is,
the dis
tied in with
the highest order of derivative n in the energy function, Eq. (1321) below.
Since n
bility is
equal to n
minimum
=0.
element boundaries.
common
It
two elements afe compatible since only the straight line can
pass through two nodal displacements common to both elements. For the
sides of
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 135 Interelement compatibility.
we
Chapter
_ du
dx
_
~
duds
ds
\J\
du dt
dx^Ttfa
&ML'\ds
dt
dt
dsF
(1318a)
Similarly,
dv
dy
_
~~
dvd
dvdt^
ds dy
dt dy
(1318b)
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.
(1318c)
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}
(1319)
B 14
or
(1320)
[B]{q},
where the
Step
We
is
is
[3]
t If wmm* *
6
where {X} r
Ty
{q}
JJ
[Tx
and
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
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
(1322)
1)
u t9 v lt
etc.,
and equating
to zero,
6U,
(1323)
0.
d{<\]
M
(8
(q)
8) (8
1)
(Q)
x 1)
(8
{Qil
x 1)
(8
{Q 2 }
(8
1)
(1324)
342
where
[k] is
Chapter 13
= h ft [Bf[C][B]dxdy,
(1325)
and {Q}
is
{Q}
{Qi}
Evaluation of
[k]
{Q,}
JJ
\NY{X}dxdy
js
\NY{T}dS
and [Q
M  h S M*, tdriQMst,
where
{st ,
t t)
The
is
first
t,)]\J(s l9 tt)
and
t,
(N
2 or fourpoint
= 4)
and so on.
integration
used.
part
{QJ
{Qi}=AStN(*0] r {X}V
If
136(a)],
it is
(1327)
t.
(1326)
X = 0,
force intensity
is
0~
~N,
(1328a)
N,
N<
N
N
N_
)\W
J(st9
N,
tt
(1328b)
t.
i;
t
}
_0
The
subscript (s (i
t t)
(Sl,tt)
component of
1(5)
evaluated at points
ti
)\J(s h
ti
)\W Y,
/,).
For
(1328c)
(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
Chapter 13
343
(a)
tttttttttt
(b)
tractions
(b)
Surface tractions.
on
(a)
Body
force,
Then
0"
r^
N
N
{Q 2 }
Txi
N<
h
dS.
(_'
Nt
N
N
_0
N,_
(1329a)
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),
+ ^),
(1329b)
+0=0,
+0=0,
{Q 2 }
We
hU
"
s)/2
s)j2
dS.
(1
s)j2
(i
(1329c)
y\
integrations,
Upon
required
we have
IT
T.
{Q 2 }
1
where
lx
= length of side
12.
(1329d)
y\
of the quadrilateral. If
results
may
we
is
is
is
distributed
a consequence of
approximation,
procedure
is
essentially the
same
common
and
boundary conditions
in
Two Dimensional
Chapter 13
Then
are
(1320);
and
(133), (135),
345
strains
and
and
stresses
(137), respec
tively.
Triangular Element
1 1
it is
relatively
using the triangular element (Fig. 112). Use of the linear function, Eq.
(ll4a) for both u
and v
satisfies
plane deformations).
[N]
~N
(1330a)
v],
_0
W=
"2
["l
b2
~b t
a2
#!
V2
(1330b)
3
V 3 ],
(1330c)
0"
b3
M=A
where the
N,
a3
fl t
fl 2
a3
b2
b3_
(1330d)
in
Eq.
(ll3c).
is
quadrilateral element [Eq. (1324)]; only the orders of various matrices are
different.
The
stiffness
[k]
and
is
given by
= h[BY[C][B] j\ dA
hA[BY[C][B].
6x1.
(1331)
Y (X
= 0)
[Fig. 137(a)],
[Q
The uniform
T
l
= ^[0
surface traction
is
divided equally
on
side
[Fig. 137(b)]
{Q 2 y
is
(1332a)
7].
among
applied.
For
instance, for
Txl
we have
^ZkZi[0
110
0],
(1332b)
Two Dimensional
346
Chapter 13
(b)
Body
face traction.
in the
directions
131.
Figure 138 shows a square isoparametric element (see also Example 121, Fig. 124)
subjected to a surface traction
are
E=
First
we show
we have
(1220c),
fx
equal to
jc
13
*24
X34
12
*23
x 14
kg/cm on
computation of matrix
=o  = 1,
= 1 01,
= l 01,
= 01  1,
=
=
= 0.30.
= 0,
= 0,
[B].
= 1,
7,3=01 = 1,
 = 0,
y 12 =
y =  = 0,
7 14 =01 = 1,
^ 2 3=0l = 1.
y 1Af
Referring to Eq.
'
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
Then
/
= i[(D(D +
/[(0)(i)
= id +
= 0.25,
which
is
s[\
(1X0)]
(0)(i)]
1)
onefourth of the area of the square. Therefore, use of Eq. (1224c) 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\
=
lxs0xt]
\{\
~
2
x
x/
xn
,j
'}
B 2 * = y[l + ( 
/]
= y(l 
[1
"
/),
/),
*23
Now
(D(l) +
(_1) *
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*
i
oq
oqoq
cq
cqoq
cq
t>
<Sh
N
cq
oqoq
+U
r*
oq
348
**
**
cq
_L
Chapter 13
[qf
As
{q} as
u2
vl
[u l
u3
v2
we need
349
to
w4
v3
v 4 ].
at
each
of the four integration points (Fig. 138) and add them together to obtain the matrix
[k].
As an
Buisi,
Bn(s2
Bu(s 3
t2)
t3 )
The
"49.45
[k]
tj
found as
Bui**, U)
tt )
tt)
Bn(s h
of
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.
0.50
0.0
[o.O
0.0
0.50
0.0
0.0
(1334b)
0.0].
v\
v2
v3
W4
v4
two equations as
4945w 2
4
549w 3
549!
4945w 3
=
=
0.50,
0.50.
u2
Use of Equation
{<y}
u3
=
=
1.000
kg/cm 2
0.300,
0.000,
stresses
Ox
are
x 10" 4 cm.
ax =
0.91
1.000 kg/cm 2
0.300
computed from
^ i V(<7*  O Y +
y
(1334c)
Xly.
for analysis
Example
132.
Evaluate element
Triangular Element
stiffness
139).
(0, 0)
t<l,0)
11(0, 1)
v(l,
i)
=
=
v(0, 0)
kg/cm per
unit thickness.
0.0,
=0.0,
v(0, 1)
=
=
(1335)
0.0,
o.O.
Thickness h =
cm
Element 2
1
Local node
Degree
350
of freedom
Chapter 13
We
351
1
(Fig. 139). 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. 139 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. (1330d>.
"I
0"
B =
1
1
For plane
[k]
we use Eq.
stress conditions,
1
1336;
133(b)J
and
1331) to yield
MBFicm
l
ooo
1
= A^
1
1
J
!
10
1
133
"a.)
and hence
f
Element
Global
1
I
Local
1
d 
V)
<1
  V)
(
  V)
V
1
vj
;
1
2.1
W=
133r.
(1
"  VI
svm.
Local
Element 2
4431obal
352
The
stiffness
2,
Chapter 13
element
element
2,
at the
bottom of
Eq. (1337b).
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
24 yields
Element 1
Global
Element 2
Local
1
h
1
[Q]
X
2
Element
assign
E=
Global
Local
we
(1337c)
'
.0.
If
Global
10,000 and v
0.3,
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
(1338)
Local]
Element 2
Globall
Assembly of the
follows
stiffness
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
Chapter 13
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.
u2
These
u4
x 10" 4 cm.
0.91
Comment on Convergence
In the case of the displacement formulation, algebraic value of the
potential energy in the system
As
is
is improved by using higherorder 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
trace
is
defined as
2K
ih
that
is
stiffness.
is,
is
is
7418 x 8
quadrilateral element
COMPUTER CODE
The computer code PLANE2DFE (Appendix
4)
The
133.
The
m wide
may
Chapter 13
erties
355
= 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
shown
as
in Fig. 1310(a).
By using the code PLANE2DFE 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
first
value of "wall
0.21
Beam
10,000 kg
1
I
44
!!
1
4
4II
JBeam'
m
20,000 kg
h
Column
Wall
Opening
Wall
n
Column
'///////////////////A
y////////////////////y
8m
(a)
356
Chapter 13
y
,
10,000 kg
20,000 kg
A
A
Origin^ W/a
A
w/AW/A
wA
w//,W/A
(b)
and
offer
comments concerning
on the
loaddeformation behavior.
Figure 1310(b) shows a
mesh
is
two conditions.
1310(a)].
Chapter 13
357
moments
It
allows computations of
Example
134.
Analysis of
is
Dam
on Layered Foundation
dam resting on
n
Scale:
Horizontal:
'
'
cm
3
1.5 x 10"
Vertical:
'
cm
= 0.788
/
/
E^O^xlOHg/m 2
(a)
Figure 1311
shapes for
Computed
Two Dimensional
358
Scale:
Horizontal:
Vertical:
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)
dam and
foundation.
Concrete
in
The
stresses
dam:
E=
v =
y =
Layer
and
432 x 10 6 psf
(=
2.1
x 10 9 kg/m 2 ),
0.3,
150
1:
).
10 6
kg/m 2 ),
Chapter 13
Scale
cm
for a
= 1.57
104 kg
359
+ Tension
 Compression
A 0'
B 0
C 0
Figure 1312
Computed
Layer
distributions
w ,u =
0.21
x 10
of c >
kgm 2
at
A A, BB,
CC
2.
E=
144 x 10 5 psf
(=
7.00 x 10 8 kg
=0.3,
2
),
j.
Partial Results
Figure 1314 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. ( 1339)
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
Chapter 13
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
The crack at the crest of the gallery is introduced by providing node numbers on
both sides of the crack. In Fig. 1315 is shown the distribution of vertical displacements along the ground
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. 1314 and 1315 and the results
stress distributions
results,
it is
dams on
check
is
Moreover,
if
stress
is
if
it
may
be possible to
it is
method can
possible to find
galleries.
3
Gfl
a
c
a
363
Example
As discussed
results that
Beam Bending
135.
is
refined con
Note
mesh
that a refined
is
required for
mathematical convergence. This and other requirements for mesh refinement are
discussed in Ref.
[3].
TABLE
the
maximum
difference
131
N D B
==
(/>
10
25
14
81
10
22
1)/
NB 2
Point
Point
//
C
V
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)/
(1339)
The quantity
NB 2 is proportional
equations, which constitutes a major portion of the time for the finite element solu
Figure 131 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 tradeoff between accuracy and
,
that
computer
is,
(b),
and
(c)
element solution
show
[4].
o x for
the
refinement.
Comments. The
results of this
example
number of elements
As noted
364
mesh
we can improve
is
and computer
time.
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 1316
mesh,
(c)
Beam
bending,
(a)
Fourelement mesh,
(b) 16element
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)
Displacement
vs.
number of nodes:
366
NB 2
points
points A, B,
'
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
stress,
ax
at typical sections.
367
368
Chapter 13
we can use
nodes
and
for the
quadrilateral there will be eight nodes. There will be a corresponding increase in the
wide
The
in
in using
is
scope and usually needs parametric studies for a specific problem on hand.
analyst
may need
use
PROBLEMS
131.
Compute
the
132.
Compute
[A]
load vector
initial
triangular element
1
in
number
{Q
for the
in Fig. 139.
133.
Fill in
134.
fy
[Fig. 136(b)].
135.
Evaluate the
stiffness
136.
[u l
u2
vt
v2
u3
v 3 ].
stiffness
stiffness
an additional
kg at the
and stresses
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 1315 shows the deflected shape of the ground surface with hydro
40
ft
(12
the
PLANE2DFE
static loading.
1310.
Vary the
relative values of
and
For the beam in Fig. 1316, consider a concentrated vertical load of 5000 kg
at point A, and study convergence of displacements vs. TV and NB 2
1312.
finite
beam
in
Example
35,
with
<3
<8
E
cr>
cn
II
II
MOO
II
LU
C
41
<$
^3
369
Two Dimensional
370
Chapter 13
a hole around the centerline. For the concentrated load shown, obtain the
finite
medium
The medium
and
0)
at
vertical
directions, decide
properties of the
Change
the boundaries to
6B, 8B, and 10B in both directions and examine the displacements near the
discretized
Q_
=4,6,8,10,
_:
T3.
Discretize
X 2 =4,6,8, 10,
1314.
is
excavated in a (homo
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
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
REFERENCES
[1]
Popov, E.
P., Introduction to
Cliffs, N.J.,
Mechanics of
and Goodier,
Timoshenko,
York, 1951.
[3]
Abel,
J.
F.,
S.,
and Desai, C.
Bending," Proc.
2148.
Englewood
[2]
[4]
Solids, PrenticeHall,
1968.
ASCE,
S.,
J.
to the Finite
New
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
threedimensional building frame and its foundation (Fig. 141), it is convenient and economical to treat the building frame as composed of onedimensional beams and columns and twodimensional 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. 141); this approach is often referred to as a Winkler foundation.
Thus, the system contains three components beamcolumns, slabs or plates,
and foundation.
Many other situations in stressdeformation analyses and field problems
require such multicomponent idealizations. We shall illustrate here the
problem of a building frame (Fig. 141). 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
[13].
Chapter 14
Building
373
zation of supports
Idealized
%>
column
VARIOUS COMPONENTS
As
BeamColumn
In Chapter 7 we have already derived the element equations for a beamcolumn. However, now we need to consider the possibility of loads causing
lateral
axes, respectively.
374
Building
Chapter 14
w2
A,E,I X ,T,I 2
/I
Bending
Axial load
Bending
in
xdirection
in
ydirection
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 ],
(141)
we assume
= 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
(142a)
(142b)
(142c)
Chapter 14
Here
u(z)
Building
and
v(z)
375
directions, respectively,
(end) loads.
As
discussed in Chapter
7,
express w(z), v(z), and w(z) in terms of interpolation functions and their nodal
values as [Eq. (737)]
= 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
(143a)
(143b)
(143c)
where
Nxl
+ 2s
ls(\ 2s + s
5 (3  25),
Is (s  1),
1
>!
yl
Nx3
Nx = N,
3* 2
3
,
2
),
(143d)
>'3
and
AT,
N =
9
where
is
By
(143e)
s,
where f
s,
=z
and z
=r
(144)
is
1.
7,
we can
derive the
mii)
'{Q]
(145a)
{Q}
{Q.3
where
a,[kj
[0]
[0]
[0]
ajkj
[0]
M
.
[0]
12
[kj
[kj
[0]
6/
4/
(145b)
fcJJ
12
6/"
6/
2/ 2
12
6/
(145c)
sym.
4/2
(145d)
[kw]
/
1
(145e)
376
Building
{QF = A W{X}</z
Jo
Chapter 14
fW{f}</z,
(146)
Jo
axial
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 143 shows the two effects, which can be superimposed if we assume
tions, bending,
Membrane
Effects
For the inplane or membrane loading we can assume the plane stress
and use the element equations derived in Chapter 13, Eqs.
(1324) and (1331) for the quadrilateral and triangular elements, respectively.
idealization
Membrane
patibility
elements
L
(b)
(a)
e
(c)
(a)
Com
Bending
For isotropic
flat plates,
Dn /(9
vv*
{d^
D=
where
v
is
Eh
/I2(l
w*
2d w*
d w*\
d^W + d^) =p
+
,
),
is
written as:
, tA
E is
is
(14_7)
is
and p
is
the applied
surface traction.
Our aim
herein
is
to
show
multicomponent
system for a building frame in which plates or slabs occur as one of the main
components.
problem.
bending
It is
It will suffice
is
considered. Here
we
much detail
problem.
much
difficulty.
as
[3, 5]
w(x,y)
yl
yl
yl
where [NJ
Nx3 N
yl
is
etc.,
y,
b]
where
377
378
Building
+ 2V
l)
btit
Ny3 = (3 20,
W = bt\t  1).
{a
and
1,
1),
{q^F
is
[Wl
0*1
1,
y4
[Fig. 143(b)], s
= x/a,
<s
and
^2
^yl
(148b)
of the element
3t
<
Here
N.
Chapter 14
0*2
H> 3
0,2
0*3
0,3
0*4
yi]
The
N N
interpolation functions
N = Nxl N = (1 
etc., satisfy
3s 2
yl
l9
we have
25 3 )(1
2t
=
= 0,
= 0,
= 0,
=h
=h
 3t +
2/
3r
(149a)
for
5=0, '=0,
5
=
=
=0,1
_
^, _
1,1
1,1
#,
f=0,
f
AT,
1,
#,
l,
tf,
,>
AT,
s=i,i'=0,
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
(149b)
).
(rad)
5_
16
i
iin Figs. 144(a), (b),
and
(c).
In
Chapter 14
Building
379
(a)
of
is
unity. It
interpolation
typical
N2
(c)
functions
for
plate
dN2 jdx.
all
The
d w]
z<
d2w
dy 2
d2 w
w yy
<
z[Bb ]{ qi
Mxx
"1
is
"
Eh"
~ 12(1
v
w
w
v
2
[C]{]
is
yy
Here h
(1410)
2w xy
dxdv>
[6]
w xx
dx 2
(]
bending is given by
w
J
(1411)
380
Building
The
Up
=
"T
If
problem
Chapter 14
is
expressed as
(tfMd&
(1412)
The components of
{e}
dx
= >
and
= [BJ{q
Substitution of {e}
= =.
=
dy
a ds
(1413)
v
7
b dt
A}
= 0=>&L =
(?n,
partial
(1414a)
0,
KK*} =
(14I4b)
{Q b },
where
\kb ]
{Q,}
\\
[B b Y[C][Bb ]dxdy
\\
[Wmdxdy
and
+
The
coefficients
[N] T {t}ds.
[4].
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].
(1415)
Here we have considered only one of the many possible and available
fact, the function in Eq. (148) 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 beamcolumns 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
difficulty.
membrane behavior
for the
beamcolumn
The
final
[K]{r}={R}.
(1416)
These are modified for the boundary conditions and then solved for the nodal
strains,
stresses,
[7].
column,
of
foundation
by
sp:;zs.
[a]
Beam
can be added directly to the diagonal coefficients of the global matrix [K]
[Eq. (1416)]. 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
is
described in Appendix
Some examples
4.
Example
141.
Load
Figure 146 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
(1417)
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.
finite
element analysis
is
0.01568 cm.
25.40
cm
Example
142.
Two
Two
Fixed Edges,
Figure 147(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
The value of
[3, 8].
This
E=
7.0
x 10 5 kg/cm 2 was
assumed.
finite element mesh for the plate. In Fig. 148 are shown
computed values of transverse displacements at center section CC of the plate
[Fig. 147(b)].
We can compute the displacement by using the results from strength of materials
[9] if
the plate
is
distribution of bending
computed from
this
382
maximum
is
Chapter 14
Building
383
0.453 kg
cm
1.905
0.3175
jL
26.67
cm
cm
(a)
&
*A
(b)
beam bending
[8].
(a)
Beam
with
from
solutions.
PL
\92EI
~
3
W max
m ax
~~
453 x 26.67 3
192 x 7.0 x 10 5 x 0.00508
(1418)
0.0126 cm,
where
l
Example
143.
1.905
12
x 0.3175 3
0.00508.
12
384
Building
Chapter 14
1.510 kgcm
1.510
1.510
(a)
1.510
1.510
EI
(b)
(b)
Conjugate beam.
732 cm
732 cm
(a)
[10]. (a)
Plane layout.
Figure 1411 shows the computed deflections and the deflected shape of the
frame. Table 141 shows the computed values of bending
bers in comparison with those
moments
in various
mem
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)
cm
= 0.0394
cm
~~7
/
/
/
/
/
J
_J
/
/
/
/
/
/
/
1
/
/
/
385
386
Building
Chapter 14
the latter
a member.
is
It
in the frame.
TABLE
141
Member
1
2
3
End
6
7
8
9
10
Example
144.
Method
x (kgcm)
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
Figure 14 12(a) shows the layout of a building frame idealized by using a one
dimensional beamcolumn, twodimensional plate membrane, and bending elements; Fig. 141 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
;
Area of cross
section,
Table 142 shows a comparison between moments (Mx ) at typical nodes for the
It can be seen that without floor slabs,
the middle section of the frame will have higher loads (moments) that those at the
Chapter 14
Building
TABLE
142
Comparison of Moments
(Mx )
387
at Typical
Nodes
Plates
Node
Not Included
(kgcm)
Plates Included
(kgcm)
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
[10]. (a)
Details of
Building
388
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
first
and second
floor levels.
edges. Moreover, the displacements at the central section are significantly reduced
if
is
0.24393
cm and
Example
To examine
145.
For
instance, the
is
y displacement
at
node 12 with
plates
0.40225 cm.
Chapter 14
Building
389
A
A
A
732 cm
732 cm
At, k.
Figure 1413
Frame
with
rotational
restraint
represented
by
spring.
very high value produces results close to those with total restraint (see Table 141).
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
(kgcm/rad)
lOis
lOio
10 9
108
107
10 5
103
143
Moments
x (kgcm)
End
Member
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
390
Building
We
Chapter 14
aim
common
is
to
coordi
nate system.
Although the
For
element equations. Hence, both these instances did not involve transformation
of coordinates.
For certain
beamcolumns, nonorthogonal
and curved structures (e.g., shells), it is often
systems whose directions are different from the global
Then
it is
is
achieved
z,
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,
(1419a)
and
global axes; for instance l x represents the direction cosine of the angle
between
As
x'
at
i
a simple illustration,
x.
[t]
given by
m,
cos a
sm a
sin
cos a
(1419b)
[t]
J,
The transformation matrix
[t] is
orthogonal, that
is, its
inverse equals
its
transpose
[t]"
We
{Q ( }
can
now
=[t] r
(1419c)
and
and load vector evaluated
Chapter 14
Building
x, y, z
391
Global
Local
x', y', z
(a)
XcL
lb)
of coordinates,
(a)
Threedimen
Twodimensional.
W=
z',
then
T
[T] [ki][T]
(1420a)
and
{<W
= imft)
(1420b)
where [kj and {Q g } are the corresponding matrix and vector referred to the
global system, x, y,
z.
is
is
composed of [t]
in Eq.
given by
[/]
(1420c)
[71
[t]
[/].
m = mw
written as
(1420
392
Building
where
[u,]
and
[u g ]
Chapter 14
and
where
{q g } is the vector
now
expressed as
(1422)
{Q,}
(1421).
by using
com
unknowns.
The foregoing concept can be used and extended for other onedimensional, and two and threedimensional elements. The process is usually
patibility of the
straightforward. For further details, the reader can consult various references
PROBLEMS
141.
Derive the load vector {Q} in Eq. (146) for uniform body force
form traction f causing bending. Hint See Chapter 7.
X and uni
142.
stiffness
axis.
Solution:
[k,]
where
G is
the shear
= GJ
1
!]
REFERENCES
[1]
[2]
S. S.,
Struct. Div.
Analysis,
Structures," /.
ASCE,
[3]
Desai, C.
[4]
L.,
and Schmidt, L. A., "The Generation of Interand Mass Matrices by the Use of Interpolation
Stiffness
Chapter 14
Building
Formulas,"
in Proc.
393
in Struct.
Mech., Wright
[6]
[7]
[8]
Desai, C.
S.
to the Finite
B., Soil
Plate Bending
Mechanics
in
Shells,
McGraw
Timoshenko,
S.,
Strength of Materials,
1930.
[10]
MalekKaram,
project report,
tute
PRELUDE
TO ADVANCED STUDY
AND APPLICATIONS
It
is
method
lies in its
we have
necessary to
finite
element
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.
dissimilar media,
convergence,
stability,
and consistency
of various
formulations, and a
and mathematics of the
394
Chapter 15
engineering.
The bibliography
at the
395
THEORETICAL ASPECTS
1
Formulation Procedures
a.
Weighted residuals:
collocation,
least
squares,
subdomain,
and
Galerkin.
2.
3.
Isoparametric Formulations
4.
HigherOrder Approximations
5.
a.
b.
c.
Mesh
d.
refinement.
Mathematical Aspects
a.
Basis of formulation
b.
Accuracy, convergence,
c.
Initial
linear
spaces.
bounds.
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.
freesurface
problems.
e.
Loading.
Static,
ii.
f.
7.
Linear
elastic.
Higherorder
elastic.
as structure
396
Chapter 15
d. Hypoelastic.
e.
Elasticplastic.
i.
ii.
iii.
Strain softening,
iv.
f.
Viscoelastic.
g.
Creep.
critical state
concepts.
h. Thermoviscoelastic.
i.
8.
Endochronic.
Nonlinear Analysis
a.
b. Incremental, iterative,
c.
9.
procedures.
Coupled Problems
a.
Thermoelastic.
b.
Consolidation.
c.
Liquefaction.
10. Fluid
Flow
Environmental Problems:
Mass
12. Fluid
13.
14.
15.
16.
BIBLIOGRAPHY
Brebbia, C. A., and Connor,
J. J.,
Finite
New
York, 1974.
S., and Abel, J. F., Introduction
Nostrand Reinhold, New York, 1972.
Desai, C.
Desai, C.
S.,
neering,
to the Finite
in
Geotechnical Engi
Gallagher, R.
1975.
H,
Finite
Cliffs, N.J.,
Chapter 15
Finite
Oden,
J.
T., Finite
J. T.,
New
Fundamentals and
and Reddy,
J.
N.,
Variational
Methods
York, 1975.
Oden,
397
in
New York,
1972.
Theoretical Mechanics,
Pinder, G.
demic
F.,
in
Press,
New
York, 1975.
Strang, G., and Fix, G. J., An Analysis of the Finite Element Method, PrenticeHall,
Englewood Cliffs, N.J., 1973.
Zienkiewicz, O.
Hill,
New
in
Engineering Science,
McGraw
APPENDIX
VARIOUS
NUMERICAL PROCEDURES
SOLUTION OF
BEAM BENDING PROBLEM
INTRODUCTION
To
illustrate the
appendix we
shall solve
we
MWR
MWR
is
expressed as
(212)
,?,.
Jd
398
R(x)W (x)dx
i
= 0,
1, 2,
W
,
(x).
sense by weighting
Thus,
(Al1)
Appendix
D is
399
the body.
Wu and depending
we
ti
on the choice of
W = 3(x  x
(Al2a)
t).
Then
RixdSix
= 0,
x)
t
1, 2,
1,
=x
[0,
^x
(Al2b)
n,
where
3
is
selected
number of
A 11 (a),
the residual
The
means
if
t9
is
equated to zero at a
equated to zero at n
shown
in Fig.
5 points.
total
1(b)],
,+1
x 'J
< x and x
0,
'
>
(Al3a)
xt+l
Then
R(x)dx
= 0,
1, 2,
...,
method
1.
(Al3b)
weighting
*g.
and
(Al4)
this leads to
f
Jd
R 2 (x)a dx =
t
{
(Al5)
0.
Jd
*(*)?,(*)=
0,
1,2, ...,.
q> t
are
made orthogonal
(Al6)
to the
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,
Subdomain.
functions are
functions
t ;
(c)
1, 2, 3,
commonly
then the
details are
shown
in Fig.
is
is
R(x)
F^ 
P(x),
(Al7)
Appendix
401
Collocation
Finite difference
Subdomain
3L/4
L/2
Galerkin,
Least squares,
(c)
Ritz
0L
(d)
Finite element
Figure Al2
methods,
different
Subdomain.
(c)
(a) Collocation,
where F = 7 is the
p(x)
is
flexural rigidity,
and x
the coordinate.
is
W (X
= 0) =
d w,
dx 2
The
first set
tions
w(x
=L)=0,
Ax
0)7
dx
L)
(Al8a)
(Al8b)
0.
set [Eq.
boundary condi
conditions.
We now choose
unknown w*
a, sin
^+a
sin
^+
a3
sin
^+
a4
sin
= t *0M,
where the a are the undetermined parameters and the
t
(Al9)
(p t
are the
known
402
functions.
whereas
Appendix
in
interpolation functions
t.
Note that the function w in Eq. (A 19) satisfies the boundary conditions in
Eq. (A 18) 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. (Al8a)] at the ends were used to modify the assemblage
boundary conditions
the natural
3,
To
w and
its
derivative,
we
differentiate the
T;
+
where X
256A 4 a 4
16/ 4 a 2
sin
j
+ 8U
a3
sin
j
sin^,
(Al10)
= 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,
in Fig.
1, 2, 3, 4,
at the supports.
instance, for Xj
X4a
sin
4.77
103<xi
7.72
7.72
x 10 3 ai
4.77
103<xi
103<xi
j
= x )=0,
= 1,2,3,4.
(Allla)
= L/5, we have
~ + 16A a
4
256A 4 a 4
is
Then we have
R(x
For
sin
sin
^^ 
^ y + 8U a
Pb
Pa
sin
^~
Pa =
(Al lb)
+
+

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
(Al12)
800,
900.
Appendix
403
a3
= 0.11374399,
= 0.00033150,
a2
a3
=
=
0.00105974,
(Al13)
0.00001564.
0.11374 sin
^  0.00106
0.00001564
sin
^+
0.0003315
sin
is
^*
4nx
(Al14)
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
(Al15)
r
LI 4
J 1L 4
L
[
3Z.
x lCPai
1.83
10 4
1.83
x 10 4 ai
7.57
x 10 3 ai
ai
+
+

119
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,
(Al16)
2030,
2340.
0.12388 sin
^
0.00004724
0.00151 18 sin
^+
0.00078719
4nx
sin
^
(Al17)
sin
LeastSquares Method
&*
(Al18)
Therefore,
W sinp,
x
3nx
3
sin
W,
2nx
sin
W = sin Anx
fjjr
>
404
[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
(Al19)
R(x)
The
final
4.06
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
(Al20)
^
(Al21)
Galerkins Method
It
is
method
are the
same
same
as the functions
(p t
in the leastsquares
solutions.
Ritz
Method
In the Ritz method [1,2] the potential energy in the body (beam)
trial functions, and the resulting expression
a,.
is
is
<
A1  22 >
with
p{x)=Pa
+ ^(PbPa)
(Al23)
Appendix 1
405
dll p
= 0,
~
da 2
'
(Al24)
dUp
0,
da 3
dll p
= 0.
(?a 4
For
this specific
leastsquares methods,
is
the
same
as in
Eq. (Al21).
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.
method (Chapters
11,
is
Finite
Element Method
The
"
12
30
30
100
12
30
F 12 30
24
125
30
30
50
50
12
30
200
12
30
30
12
30
5030
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
(Al25)
<
417.0
2312.5
1875.0,
gives
= 0.0000000,
(A 126)
0.0383335.
Method
finite
method
and mathe
finite difference
406
However, we
The
are
this
book.
method
Appendix 1
method
finite difference
[1].
is
For
finite differences.
Fig.
approximated as
(see
Al3)
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 i2
i
Substitution of
2Ax
w i+1
2w,_
2w i+l
(Al27)
w i+2
2(Ax) 3
yv t
4w
_,
6vv,
4w /+1
Ax 4
4w _
t
6w
4w i+l
w i+2
= 0,
1, 2, 3, 4,
(A 128)
Pf
Ax'
tion of the
Ax
2w
w,_ +
p Wj2
Ax 2
d3 w _
dx 3
^ w w _i w
Mw
~~
beam domain
finally gives
four simultaneous
i1
divided
+3
Appendix
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 129)
Solution.
with
==
w5
= 0.070656,
= 0.115968,
w 2 =0.114432,
w4
(A 130)
= 0.073344,
0.
results
^(L  2Lx + x
2
) 
)X
(/7
f80ir^l
(3^
10L 2 * 2
7L 4 ).
(Al31)
beam by
using
various methods and the closed form solution [Eq. Al31] are compared in
Table
AM.
TABLE
Al1
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
last
and
by substituting various values
(Al31). 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*
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. (72). 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
loading characteristics.
REFERENCES
[1]
Crandall,
[2]
Abel,
[3]
Desai, C.
J.
S.
F., (private
S.,
[4]
Timoshenko,
1956.
S.,
York, 1956.
communication).
and Abel,
Nostrand Reinhold,
New
J. F.,
New
Van
York, 1972.
Strength of Materials,
APPENDIX
SOLUTION OF
SIMULTANEOUS EQUATIONS
INTRODUCTION
Most problems
= {R}.
(A21)
we have
in this text
essentially
is
= {b)
(A22a)
or simply as
Ax
where
[A]
is
= b,
(A22b)
the vector of
*n*i
021*1
a Hl x
(J
is
unknowns such
as
+ 012*2 +
+ 022*2 +
+
an2 x 2
{r},
+
+
a ln xn
a nn xn
a lH xn
=b
=b
l9
29
(A23)
=b
n,
409
410
Appendix 2
where n
the
is
number of
the total
equations.
METHODS OF SOLUTION
The two common methods
iterative procedures.
relaxation (SOR),
examples of the
these techniques.
Gaussian Elimination
This
is
equations.
common method
unknowns x
lent systems
The equiva
sequence by superscript
initial
+
+
iV*i
fllV*i
tfi
tfixt
ffl*a
*i+.ff*a
+
+
+
+
+
..
flfite
is
simpler
Eq. (A23)
(1) as
= M
= W\
a[xn
in
+ ax =b
n
(A24a)
>\
or
=b m
A (1) x
(A24b)
In the
multiply
first
it
(a +Xi*iV)*i
(flffl
*2ia 2 l)x 2
(affi
Aaiaffitei
+ hiW)
(A25)
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
first
I'M
+ a^ixi +
+ a%x +
+ ax +
2
+ a$x +
2
we
+ aiite
+ a%xn
+ flff*.  o
/>< 2 >
/)
'
aff x.
(2 >
3
(A26a)
.(2)
b
/32
42
wr
~wr
x *
 sgf
iV*i
sequence at
step of elimination as
(l)
41
!)
+
+
+
+
iy*a
(2) Y
"22M
+ fl,Vxj
+ a&Xt
+ JV*
+ ax
S
+
+
+
+
which leads to
+ affjc. = 6
+ t&x. = 6i2
+ a>* = W,
+ a%x = Ai,
(1)
'i
>
>,
(A26b)
+ fl& *3 +
Finally, at the end of n
elimination steps, we shall have
a.'M" + ai*2 + <*,V*s + JV*4 +
+ ai'.'x, = M
2
+ Sx2 + ai 3>x + a$JC +
+ tfi*. = b{ \
+
+ a$x + aflx, +
+ a%x = 6<
+
+
+ a^xt +
+ ai?x = b\
}
.',
>,
(A26c)
fl
We
(l)
fl
(l)
A(nl)
note that Eqs. (A26a), (A26b), and (A26c) are equivalent to the
(A26c)
is
x n can be obtained
Back
The
all
from the
unknown
last equation.
Substitution
first
(A26c) as
step here
is
x n from
412
Appendix 2
= fcu
(A27a)
n.n
Now
*.>
(n
l)th equation as
(A27b)
^fe^*
Since x is known from Eq. (A27a), 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
tions [13].
banded and symmetric. In banded matrices, nonzero coeffioccur only on the main and adjacent diagonals, and other locations
have zero
a tj
=a
jt .
coefficients.
For
is
banded
is,
(tridiagonal)
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
(A28)
2i
^>4>
significant simplifica
is
necessary to store
only the nonzero elements in the computer and only the coefficients on the
[4].
SOLUTION PROCEDURE
Almost
all
is
not
finite
element analysis
element solution
fact,
is
a major
spent in the
shall
now
of the foregoing
Example A21.
we
3,
we
solved these
procedure
100 (1) jc!
+
+
100 (1) * 2
^!
100 (1) x 2
(1)
*3
100 (1) *3
200 (1) *3
=
=
=
7.5 (1)
15.0 (1 \
15.0 (1)
(A29)
Details:
+ 10Q(i) _
Therefore
we have
(200
100) (2) x 2
100 (1) * 3
15.0 (1)
100 (2) * 2
100 (2) * 3
22.5 (2)
7.5 (1)
or
lOO^*!
 100
+ 100
 100
(1)
(2)
x2
(2)
Now
A 32
+100/100
*2
+
 100
+ 200
(2)
*3
(1)
*3
=
=
22.5 (2) ,
15.0 (2)
1.
x2
first
(200
100) (3) X3
15.0 (2)
37.5 f3)
22.5 (2)
or
+
Hence, the
final
100 (3) x 3
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
(A210)
100
60
100
7.5
100
100
100
60
100
67.5
100'
413
Iterative Procedures
made
is
for the
is
The
unknowns x and
iterative
is
procedure
number by which the final solution differs from the solution in the
previous iteration. The simplest iterative procedure is the Jacobi scheme [13].
acceptable
To
it
illustrate this
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
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
(A2llb)
'T,
033
j_ ^3
...
Xv (0)
3
first iteration,
_ a 2n XY
(0)
n
(A2lla)
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
\<e,
1, 2,
(A212a)
,n,
X2
x (m>
ai\
an
= _q
fl33
x nm)
(
414
an
y (ml)
2
'
Cl\n
# 23 v (m* 3
1)
ain v (ml)
32 **2
v (ml)
033
Qnl *2
v (mQnn
1)
Q/3
a nn
v (ml)
an
an
x (ml)
Ojjymi)
an
x (ml)
Qn\ v (ml)
x
a nn
fll2
x (ml)
3
a22
an
a 3n v (ml)
x"
7.
#33
as3
an
(A2llc)
a nn
Appendix 2
m ~ l)
x\
m)
415
is
<
1, 2,
(A212b)
n.
a large
it
may
require
obtained.
The
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 A22.
first
A3
iteration
*<>
^(0.6125)
Second
(i)
( o.575)
2)
_ A^^fml)
an n
__
bi_
aii
_ 1^1) +
^_,
(A213)
A33
**.
fl/m
x <s
x
We
x 2>
x 3>
0.50,
1^0(0.5)
^=
+ i
a575
(o)

=
=
0.50.
0.1125
0.04375
>
>
>
'
0.6125,
(A214a)
0.38.25.
is
0.005
0.005>
not acceptable.
(A215a)
0.005J
iteration:
^.=^(0.6125)
+T55
100
_
_
we have
x a)
x (m)
3
a nn
b\
All
A22
0.50,
4o
we choose
..
xv (m)
2
*i"=Jo> 5>
Tf
x<>
for the
A2 3 ^(wil)
3
A22
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,
03796875.
(A214b)
416
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 }
(A215b)
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.
(A214c)
Convergence check:
\
X U)  x (2) =
0.003125
xy =
_x =

0.002344
0.001172
.12)
Hence the
x (3)
j2)
< 0.005]
< 0.005 >
< 0.005J
end of three
Xl
x2
x2
=
=
=
(A215c)
acceptable.
iterations
is
0.6844,
0.6070,
0.3785.
tions,
it
Jacobi:
x (m
>
=(L+
U)x lm
>
(A216a)
fi.
GaussSiedel.
Lx (m)
L and U are
the lower
Ux (m
>
A22, at
m=
1,
100
Too
ioo
200
(A216b)
fi.
[A]
in
and fi is
Example
Appendix 2
100
100
417
100
200
U
7.5
fim
15
15
We
where
can
now
x(m>
= X (mD
write
_j_
^[(^(m)
ux (m
l)
P)
x {m
X)
(A216c)
co is
on the righthand
Rearrangement of terms in Eq. (A216c)
gives
x (m)
which
=(/
is
coL)
[{\
a statement of
For
significantly.
(o)I
SOR
coU]x (m
SOR
[2].
may
l)
co(I
 coL)
/},
(A217)
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
Detailed descriptions
REFERENCES
[1]
[2]
L.,
New
York, 1965.
2,
Bathe, K.
J.,
PrenticeHall,
[4]
to
[3]
An Introduction
Fox,
Finite
to the Finite
Element Analysis,
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
stated in Chapter
plastic.
As shown
1,
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.
RTV
The models
Rubber (Silicone
made from
Silicon
New
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 A31 shows a plastic mold used to cast the models. Figure A32(a)
shows a model of a continuous beam. The model in Fig. A32(b) was cast
Figure A31
Mold made
beam.
beam
of plastic.
Physical Models
Mesh with "rigid" elements: linear apMesh with deformable elements: higherorder
proximation,
(c)
approximation.
element.
may
With
Appendix 3
Physical Models
421
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. A32(b).
Under
load,
the
indicates
improved
approximation for w.
[Fig. A33(a)].
made out of
(thin) plastic
A33(b)]
is
model
of the base plate such that the top of the screw touches the bottom (at the
center) of the four elements.
Physical Models
422
By moving
sively)
the screw
Appendix 3
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
junction.
COMMENTS
The foregoing
and
concepts in
finite
element analysis.
APPENDIX
COMPUTER CODES
INTRODUCTION
Descriptions of a
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
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
may
CHAPTER
made
CONS1DFE: OneDimensional
Consolidation
This
is
dependent
a modification of
settlement
DFT/C1DFE
analysis
of foundations
with
nonlinear
material
properties.
423
idealized as
foundations and earth banks are available. Both linear and nonlinear (plasticity)
CHAPTER
also exist.
BMCOL1DFE:
BeamColumns such as
Piles
axially
and
laterally
loaded
piles,
and nonlinear
soil
7),
and
stress
foundations can be
included.
The output
in the
is
rotations,
and
bending moments.
CHAPTER
MAST1DFE:
This code
is
OneDimensional DiffusionConvection
in
Chapter
8.
This
is
salt
CHAPTER
10
WAVE1DFE:
OneDimensional
Wave
Propagation
424
media can be
solved.
CHAPTERS
AND
11
FIELD2DFE:
Field Problems
12
The following
problem and
its
corresponding
option code.
Option
Code
NTYPE
Problem
Torsion
Relevant Problems
(Chapter 11)
Potential
Output Quantities
Nodal
stresses, twisting
Flow through
Nodal
moment
Flow
Velocity
channels
Potential
Stream
Function
Seepage (flow
through
porous
media)
Heat Flow
Flow through
velocity potentials,
Nodal stream
functions,
channels
quantity of flow
Nodal heads,
of
dams and
sheet piles,
velocities,*
quantity of flow
and
plates, slabs,
bodies
The formulation
is
element with linear constitutive laws; for instance, for seepage, Darcy's law
assumed to be
is
valid.
Codes
and dam.
Code
plates
and
bars.
425
CHAPTER
13
PLANE2DFE: TwoDimensional
Plane StressDeformation:
Linear Analysis
This code
is
13). It
stress,
elastic analy
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
structures.
In the case of pile foundations, the code also has the facility of
outputting design quantities such as bearing capacity, wall shear
friction
in addition to
stresses.
2.
This
is
(elasticity
interface elements.
426
CHAPTER
STFNFE:
14
Analysis of
Frame
Linear Analysis
In a general building frame type of problem the beams and columns are
approximated as onedimensional beamcolumn 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
foundation media.
427
INDEX
for
torsion, 242
6163, 396
Basis functions (see Interpolation functions)
Beam bending, 15, 172, 173, 364, 369, 398,
400,418
approximation models, 173
physical models for, 418421
Beamcolumn, 172, 173, 189, 373
Boundary conditions,
279,301,317,325,395
categories of, 27
concept, 59
Dirichlet, 60
essential, 27, 31
explanation, 5962
first,
60
429
430
Index
Boundary conditions
(cont.):
317
in seepage, 325
258,264,279
wave propagation, 230232
in torsion, 249,
in
mixed, 60
modification for, 32, 33, 6062
natural, 31,60
Neumann, 60
nonzero, 61
types of, 60
zerovalued, 60
Bounds,
7,
11,8385,261,320,321
lower, 11,84
symbolic representation, 84
upper, 1 1 84
Buildingfoundation systems, 372, 392
spring supports, 372, 388
,
Calculus of variations, 24
Collocation, method of, 26, 27, 47, 398408
Comparisons: numerical predictions, and
closed form solutions, 6567, 183185,
252, 261, 267, 282, 283, 316325
in column problem, 6567
in beam bending, 183185
in potential flow, 316321
in seepage, 325
Compatibility, 7, 30, 43
interelement, 30, 43
for approximation functions, 43
Complementary energy, 24, 47, 7679, 192,
271,295,296
for beam bending, 192
for onedimensional problem, 7679
in torsion, 269
modified expression for hybrid approach,
271
principle of, 47
stationary, 24
Computer codes,
Convection, 202
Convergence, 7,
354,
364,418
mesh refinement, 87
monotonic, 44
onedimensional column, 87, 88
physical models, 419
Coordinate systems:
element, 36
global, 36
local,
36
Coupled problems,
Damping, 234
Darcy'slaw, 15,24,45,95
Degree of consolidation, 124, 125
Degree of freedom, definition of, 21
Derived quantities, {see Secondary quantities)
Diffusionconvection, 202210
Direct methods, for solution of equations,
410,413
Direct stiffness method, 58
Discretization,
1, 5, 13, 18,
213,225,246
Discretizations of "infinite" boundaries, 329
Displacement approach, 39, 240, 281, 282,
293,295,338,375381
for torsion, 240
Displacement vector:
assemblage, 58
element, 54
nodal, 54
Dynamic problems, 224236
beamcolumn, 424
buildingfoundation, 382, 427
field problems, 266, 328, 330, 425
flow problems, 102, 124,425
heat flow, 158164
Conformable functions, 44
Consistent mass matrix {see Mass matrix)
Consolidation, 108, 113, 121132
onedimensional, 108, 121130
layered media, 112, 113, 129
Constantstrainline element, 46