APPLICATION OF BOUNDARY ELEMENT METHOD FOR SOME 3-D ELASTICITY PROBLEMS by KAPPYO HONG,
B.E., M.E.
A THESIS IN CIVIL ENGINEERING Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE IN CIVIL ENGINEERING Approved
Accepted
Dean of the Graduate School
August, 1985
ACKNOWLEDGEMENTS I wish to express chairman
sincere appreciation
Dr. C. V. G. Vallabhan
guidance during research. Vann and
for his
to my committee
encouragement and
Also thanks are due to Dr. W. P.
Dr. W. K. Wray for their helpful suggestions.
I thank
my parents
for instilling
in me
education and for their financial support. my wife, Miroo,
and brother, Sunpyo,
support and encouragement.
IL
the value of
Also I thank to
for their
unwavering
TABLE OF CONTENTS page ACKNOWLEDGEMENTS
ii
LIST TABLES
v
LIST FIGURES
vi
CHAPTER I. INTRODUCTION
1
Purpose Of Study
2
Scope of Rearch
3
II. BOUNDARY ELEMENT METHOD
4
Introduction
4
Basic Equations in an Elastic Continuum
4
Equilibrium Equations
4
Strain-Displacement Equations
6
Stress-Strain Relations
7
Equations on The Boundary
8
Formulation Of Equations For The Boundary Element Method
9
Formulation of Boundary Integral Equations
9
Boundary integral Equation for an Interior Point Fundamental Solution
13 14
Boundary Integral Equation for a Boundary Point 111
17
page III. NUMERICAL ANALYSIS
21
Introduction
21
Discretization
22
Determination of Internal Displacement and Stresses
26
Numerical Integration
27
IV. EXAMPLE PROBLEMS
33
Introduction
33
Example Problem 1: Prescribed Traction on the Top Surface of Large Elastic Soil
34
Example Problem 2: Embedded Foundation With Prescribed Displacements on The Top Surface of Large Elastic Soil
45
Example Problem 3: a Solid Cube (Uniformly Loaded on one side only)
48
V. SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS
59
Summary
59
Conclusions
59
Recommendations
60
REFERENCES
61
LIST OF TABLES Table
Page
1. Comparison of Vertical Stresses Between Analytical Solutions and Boundary Element Method due to Surface Loading
41
2. Comparison of Vertical Stresses due to Displacements
50
3. Comparison of Vertical Displacements at Internal Points due to Vertical and Horizontal Displacements
53
4. Results For Example 3 Subjected To Axial Tension...
57
y
LIST OF FIGURES
Figure
Page
1. Stress Tensor in Cartesian Coordinate System
6
2. Definition of General Problem
10
3. Definition of the Displacement and Traction
15
4. Kelvin's Problem
16
5. Augmented Boundary Surface
18
6. Definition of Spherical Coordinate System
18
7. Quadrilateral Boundary Element
23
8. Quadrilateral Element In Dimensionless Coordinate System
28
9. Idealized Semi-Infinite Region Problem
35
10. Discretization of Example 1
36
11. Discretization Top Surface
37
12. Boundary Conditions and Applied Loads
39
1 3. Applied Moment in Example 1 14. Comparison of Vertical Stresses due to Vertical Traction
40
42
15. Comparison of Vertical Stresse due to Horizontal Load
43
16. Comparison of Vertical Stresses due to Moment
44
17. Discretization of Top Surface of Example 2
45
18. Discretization of Finite Element Model 19. Boundary Conditions and Application of Displacements
47
20. Comparison of Vertical Stresses due to Vertical Displacements vi
49
51
Figure
Page
21 . Comparison Of Vertical Stresses Due To Horizontal Displacements
52
22. Comparison Of Vertical Displacements At Internal Points Due To Vertical Displacements
54
23. Comparison Of Vertical Displacements At Internal Points Due To Horizontal Displacements
55
24. Surface Elements Arrangements for Example 3
58
VLl
CHAPTER I INTRODUCTION For solving complicated equations representing physical problems, engineers and scientists have been using numerical techniques for several years. Most of them were using finite difference, solving
finite element
these
researchers
problems.
have
come
or
other numerical methods for
During
up
with
boundary element method [1,2,3]. ment
of
the classical
the a
last
decade,
technique
some
called
the
This method is an improve-
integral equation methods
used
in
mechanics and physics. The improvement in these methods lies primarily in the computerization of the method, so that method
can be
used
for
a wide
variety
of
problems
the in
continuum mechanics. Researchers will asic difference,
the
question:
while
the
finite
finite element and other numerical methods have
been successfully used in solving a wide variety of problems in mechanics and while their versatility has been established,
why develop another method ?
for the development of the
There are many
boundary element
reasons
method.
Most
numerical methods such as finite difference, finite element, etc.,
are
domain
methods,
and
discretize the whole domain. method,
the domain
they
But in the
equations are 1
require boundary
converted
that
one
element
into boundary
integral equations. If these integral equations are boundary integrals, be
then only the boundary of the continuum needs to
discretized.
The method
can be applied to solve a few
problems involving infinite continua easily The method in the
is gaining
literature
popularity
and
[3,5,11,13,15].
and accurately.
is ;videiy
reported
The advantages of
the
boundary integral methods over the domain methods are: 1. There
is a considerable reduction
unknowns to solve a problem; dimensional method
continuum
requires
surfaces of the domain.
the number of
for example, in a three
problem
elements
in
and
the boundary element nodes
only
on
the
This becomes very convenient
for input data preparation. 2. For many problems, one can get more accurate
results
using the boundary element method for the same number of unknowns and computer time. 3. For many problems in mechanics,
for example in fluid
mechanics, the input-output flow data are very nicely and accurately represented. 4. Modeling certain semi-infinite and infinite domains having linear properties is comparatively easy. Purpose of Study There
are
many problems
in solid mechanics where the
boundary element method can be laore advantageous
than other
numerical methods. relates
One such series of engineering problems
to those
in soil and rock mechanics
mass of solids is subjected on the surface. tions,
to forces
where a large
in a restricted area
Classical mathematical
closed form
solu-
such as Boussinesq, Kelvin, Mindlin, Cerutti, etc.,
[8,16,17] are
applicable
only to a limited geometry of the
continuum. If the loading surface is embedded into the semiinfinite continuum, structure, solve
the
continuum. to
like a footing or any other
one has to resort stresses The
and
to numerical
displacements
foundation
techniques
induced
boundary element method has
in
to the
been reported
be very successful in solving such problems. Scope of Research In this research,
the boundary
for solving a three dimensional linear,
elastic,
solid continuum,
homogeneous, and
program is developed
element method is
isotropic.
for this purpose.
which
used is
A computer
A detailed descrip-
tion of the theory of the integral equations in the boundary element deals
method with
the
is presented numerical
formulating the discrete Chapter IV
discusses
in Chapter II. integration
Chapter III
procedure
used in
boundary element system equations.
results of a few example
problems to
validate the overall methodology.
All conclusions
from this research are
in Chapter V
summerized
some recommendations for future research.
derived
along with
CHAPTER II BOUNDARY ELEMENT METHOD
Introduction It is found convenient
to present the
basic equations
in the theory of linear elasticity before the integral equations
for the
summary
boundary
of the basic
element
equations
method are showing the
presented.
relationships
between stresses, strains, displacements, and surface tions in a three dimensional Using these equations, derived.
continuum
the boundary
required in the
trac-
are presented here.
integral equations are
Finally, the fundamental solution
static problem
A
for an elasto-
formulation of the boundary
element method is also presented at the end of this chapter. Basic Equations in an Elastic Continuum The reference coordinates used here to represent a three dimensional
elastic
continuum
orthogonal cartesian axes denoted or simply by x. (1=1,2,3).
consist of
three
mutually
by x, y, z, by x^, x^,
Throughout this work,
x^
cartesian
tensor notation will be used for convenience with a range of 1 to 3 for the index notation. Equilibrium Equations Consider an infinitesimal parallelepiped body
referred
to a cartesian
coordinate system
x^,(i = 1,2,3).
The derivation
isotropic continuum is given
as shown in Figure 1 with of the
aquations
for
an
in standard text books [6,7,8]
and, hence, is not produced here.
The differential equation
of equilibrium at an interior point in the continuum
can be
written as: 3a 3T XX ^ xy 3a
3T
3x 3T
^
xz
"iT Here,
3T
3T
3y 3T
xz
^
^^'^'
^ ^ " °
3a zz
yz
* ^
3z
* -^T
* 2 =0
a / CT,,,, t^„„ ^re normal stresses XX yy zz
are shear stresses.
and
, T , T xy' yz' xz
T
X, Y, Z are body forces per unit volume
at that point. These equations can be symbolically rewritten using index notations as a. . . + b. = 0 13/3 1 where
a.
represents
the
stress
components
on a
face
perpendicular to the x. axis and in the direction of the x. 1 3 axis as shown in Figure 1. The positive directions of a,. 13 is also shown
in Figure 1.
volume in the i-direction.
b. is the body force
The comma after ij in the above
equation represents differentiation respect
to the corresponding axes
script following the comma.
per unit
of the stress a., represented
with
by the sub-
Figure 1. Stress Tensor in Cartesian Coordinate System Strain-Displacement Equations If u, V, w ment
or
functions in
displacements at
u. (i = 1,2,3) are continuous displacean elastic
continuum
representing
the
x., assuming that the body undergoes small
displacements, the components of strains at any point x. can be represented by derivatives of the displacement components
7 These relations
between
strain
and
^xy " 3x
3y
c- - IZ
^,
3v
S - 3y
V
displacement
are
as
follows [9,10]: X
3x
3w z ~ 3z Here, e , e , e X y z are
shear
3w
/O ox
= 3? " 3l
^^-^^
_ 3_u ^ ^zx ~ 3z "^ 3x are
strains.
normal
strains
and
Y / 'xy
Y / Y 'yz 'zx
These equations
can
be symbolically
rewritten as £.. .) 13 =Tr(u. 2 1,3. + u .3,i' where £..
is a strain tensor.
If i = j ,
£.. represent components of normal strain and
if i ^ j, they represent components of shear strain.
It is
to be noted that Y-• is the engineering
shear strain
which
are the corresponding
shear
is equal to
2£.. , where z. .
strain components. Stress-Strain Relations The next step in solving a problem of elasticity is
to
develop the stress-strain relations. Assuming a homogeneous, isotropic, linear, and elastic material, the general form of Hooke's law may be expressed as 'ij = ^ ^kk ^i: * 2G e,.
(2.3)
8 where A. and G are Lame's constants given by X = .....y^
...
,
(1+v)(1-v) '
where
E
is
the Young's
G =
2
^ ~ 2(1+2v)
modulus
of
elasticity and
Poisson's ratio of the material in the continuum.
v is
A general
form of the linear relationships of stresses and strains can be written as a. . = C. ., , e, , where 13 13 kl kl material property tensor.
C. ,, , is a fourth order 13 kl
Equations on The Boundary To solve an elasticity problem, we need equations on the boundary.
There
are
two
types of
boundary
conditions
prescribed: 1. Prescribed displacements on the boundary, T u.=u. i
l
,
o n f . .
i.e., (2.4)
l
2. Prescribed tractions on the boundary
f^.
These pre-
scribed surface tractions p. are related to the stress tansor as follows: Pi = ^11 ^1 * ^12 ^2 ^ ^13 ^3 ^2 = "^21 ""l "• ^22 "^2 " "^23 ^'^3
^'^
^2
^^'^^
P3 = '^31 ^1 * ''32 ^2 ^ ^33 ''^3 or symbolically p.=p.=a..n. on F-, '^i ^1 13 3 2 where n. are the direction cosines of the surface
F^.
Formulation of Sguations For The Boundary Element Method To derive
the boundary
integral equation,
a weighted
residual method is used here.
This is one of the classical
approximation
in numerical mathmatics
techniques used
and
can be applied to solve many types of equations. Explanation of this method [8].
along with some comparisons can
In the formulation of the boundary
be found in
integral equation,
an influence function is used as a weighting function.
The
details of the influence function will be discussed later. Formulation of Boundary Integral Equations For the general
three
body shown in Figure 2,
dimensional
isotropic
elastic
the field equations and correspond-
ing boundary conditions are: a. . . + b. = 0 ,
in the domain
Q
(2.6)
with boundary conditions u. 'i = u"i
on
r^ ' 1
(2.7)
p.. =- p. M.
on on
FF-, -,
(2.8)
2
1 ^ 1
where
F. is the boundary where displacements are prescribed
and F^
is the boundary where tractions are prescribed. The
entire boundary is F = F. '^ ^o If
u. is
assumed as the weighting functions,
weighted residual technique,
in
the
then the equilibrium equations
10
Figure 2: Definition of General Problem can be rewritten as + b. ) u. df2 = 0 (a iD/D 1 1 JQ In order to get the expressions the integral,
(2.9) for the above part
of
consider another integral of the type
(a.. u.) . d^ ID 1 /D M
(2.10)
Differentiating the integrand of equation 2.10, r (a., u*) . df2 =
f
(a.. . u*) dQ +
f
we have
a. . u. . dfi (2.11)
11 and (a. . u. ) . d^ = 13 1 ,3
JQ
(a u.) n. dF 13 1 3
• / ,
From equations 2.11
X
Q
{o
u.)
13r3
Substituting
p. u. dF r ^1 1
(2.12)
and 2.12:
dQ =
I
1
p.u! dF
a
u. . d^l (2.13)
J r ^1 1
equation 2.13
into the
equilibrium
equation
2.9: p. u. dF ^ ,F ^
/
a. . u. . ^^ ^'3
Q
d^
b. u. d!^ = 0 (2.14) Q ^ ^
Using the symmetry of the stress tensor, i.e. a.. = 13 1 ^ ^ I a . . u * . d ^ = I a. ^ (u, . + u. .) d^ 2 1,3 3/I a. . e. . d^ Substituting
equation 2.15
into
31
(2.15)
the
equilibrium equation
2.14: r
p. u* dF -
r
a.. e*. d^
b^ u^ dJ^ = 0
(2.16)
^ijkl
Using
^
For a linear elastic material, a.
13
%1
this identity,
f
a. . e * . dfi =
f
C.i. 3.,k l, £,k l, £.13. df^
Q
\l
\ l ^"^
(2.17)
12 Once again, using the Divergence Theorem;
I:„ '^ij '^i'-j ^" = j r °tj "i "j ^' 'L p* u. dF
(2.18)
and differentiating by parts: I
(a. . u. ) . dS^ =: f
f
a*. . u. di:^ +
a*. £. . d!;^ ^^
Q
(2.19)
^^
Substituting equations 2.17 and 2.18 into equation 2.19: a.. £.. do =
/ p. u. dr -
/
a.. . u. dQ
(2.20)
Now, equilibrium equation 2.16 becomes
p. u. dr +
/
p. u. dr
a. . . u. di^ +
^ r Q
b. u* d^ = 0 ^
(2.21 )
^
or a. . . u . di^+
r
p . u , dr
/
=
r p. u* dr +
jr ^ "
r
b. u* dn
(2.22)
Ja ^ ^
In order to remove the first domain integral in the above equations, we can assume a*. . + A. -, (s,q) = 0 13/3
11
in
^1
(2.23)
13 if s ?^ q,
A^^
if s = q,
A.^ is a Direc delta function
such that
where
= 0
A.T d^
= 1, if i = 1
A^-j^
= 0, if i j^ 1
s is the source point, q is the field point, 1 is the direction of force at s,
which means that the solution to this equation in the domain is the solution due to direction.
one
of
solution, rather
at node s in the
i-
For every 1-direction (1=1,2,3) in source point,
we have 3 sets of any
a load applied
equilibrium equations and the solution of
equilibrium *
*
p.. and
u..,
equations which
from
the
fundamental
are second order
tensors
than a vector. Boundary Integral Equation for an Interior Point
For any internal within the domain becomes
-u., due
Q,
point
's' where
the domain
to equation
a force is applied
integral in equation 2.22 2.23,
and in
case of
the
absence of body forces, u*(s) + r p*.(s,q) u.(q) dr(q)= [ u..(s,q) p.(q) dF(q) 1 J r ^^ ^ JT -• -• (2.24) where
i is j is
the direction of a unit force at s, the direction of displacements or tractions on
14 the boundary, p^.(s,q) and u^.(s,q) are the traction and displacement on the boundary in the j-th
direction due to a
unit load at s in the i-direction The equation 2.25 gives
the displacements
at any point
inside the domain fi in terms of the boundary values u.
and
Fundamental Solution The fundamental solution at the field
point 'q' given by
Hi
u..(s,q) represents displacement field
point, q,
source
point, s,
due to
in the j-direction
a unit force
as shown
at the
in i-direction
in figure 3.
at a
This notation is
applicable to tractions as well. For an which is
elastostatic problem, the fundamental solution,
written
as equation
2.25,
has been
derived by
Kelvin and his solution can be found in several books [5,6]. * 1 ,,-, .,.v . u. . = . ^ „ , ^ 1— {(3-4v) A. . + 13 167rG(1-v)r ^ 13 * 1 r9r , ,. ^ , Pii = 2 ^ ^ {(1-2v)A
3r i:— 3x. A +
3r , ;c } 3x. ' -, 3r 3r , 3 T^p T^p}
-{(1 - 2v) T ^ n. - ^ n. }] 3x. 3 3x. 1 1 3 (2.25) where r is the distance between the source point 's' and the field point 'q' (see Figure 4), and
15
Figure 3: Definition of the Displacement and Traction
16
components of u,, and p^.
Figure 4: Kelvin's Problem
IK
17 / J/2 2 2 2 , , r = (r^ r^) = r^ + r2 + r3 = |s - q| r^ = x^(q) - x^(s)
(2.26)
^ r. r = — i ^ — = -i ,i 3x^(q) r
Boundary Integral Equation for a Boundary Point In
the
preceeding
sections,
the
boundary
integral
equation for an interior point was derived. Considering that the unknown tractions and displacements on the boundary must be calculated
to obtain the solutions in the domain,
equa-
tion 2.24 must be modified in order to use on the boundary. Equation 2.24 is valid when the source point is the domain ^. r,
But when this point is moved to the boundary,
one of the integrals in equation 2.24 becomes
i.e.,
if the source point is equal to the field
such a case,
within
special techniques
must be
singular, point.
used to
In
evaluate
these singular integrals. Assuming Figure 5, as the
that
the body can be represented as shown in
the integral
sum of
on the surface
integral on the
These integrals are
evaluated
can
be considered
boundaries, at the
F-e
limit as
calculation of this integral e is faciliated spherical system of coordinates (Figure 6 ) .
and
e-)-0.
e. The
by employing a
i —
^
part of surface containing the point 's' Figure 5. Augmented Boundary Surface
^x
^2= ^ sin 9 sin ;^
Figure 6. Definition of Spherical Coordinate System
19 Consider
the
first
integral in
equation 2.24 as two
parts:
F-e and let
I =
lim
f
£ u.
p*.
dF ,
(2.28)
e *
Substituting for p.. from equation 2,25: I =
lim
u. ^ [- I ^ -[^ {(l-2v)A. . + 3 1 ^ | ^ } ^ f ^ {(1-2v)A. -» L^"^ •a^ I \ dx. ox. / I F^ 87r(1-v)r^ ^-'.j
- (1 - 2 v ) { ^ n. - ^ n.}dF] (2.29) 3x. 3 3x. 1 1 ^ 3 Now, for the particular case of the hemispherical region, 3x.
n. "1
r
=
e. '1
1
where e. are
the projection of the unit normal vector on i-
coordinate axes. r,.
The second term in equation 2.2 9 becomes, r,.
-
r,. r,.
And noting the fact that —^—
= 1,
=0 equation 2.30 becomes
I, = lim [ - r u.{(1-2v) A + 3r,. r,. } ^ ^ dF ^ €^0 -^ r ^ ^^ ^ ^ 87r(l-v)r^ (2.30) and
expanding
I = lim [ - I
£-0
equation 2.3 0
for the instance
when j = 1,
{u.^(1-2v) + 3u.^e.^e.^+ 3u2a.^ e2 +
JF^
. . sin8d8__d8 ,„, ^^3^1^3^ 87T(1-v) ^^^
,-, ^. . ^^'^^^
20 The integral is now independent of r and may be expressed in terms of
9 and (j) only.
with respect to The
same
9
result
and (J),
Integrating the equation 2.31
we get
applies for
I.= -0.5 u.
i = 2 and i = 3,
giving the
combined result as : I = -0.5 u.
(2.32)
1
Limitation of the right does
not affect
into equation
side integral
the integral.
2,24,
of equation 2.24
Substituting equation 2.32
the boundary integral equation
on the
boundary becomes C(s) u.(s) +r p*.(s,q) u.(q) dF(q)= ru*.(s,q) p,(q) dF(q) 1 Jr 13 3 Jr 13 3 (2.33) where C(s) = j
for a smooth boundary.
It should be noted that the smooth linear
boundary.
or higher
necessarily
lie
C(s) is equal to 0.5 only
And for the
order elements, on
a
smooth
general
case of
where the nodes do
boundary,
C(s)
for using not
cannot be
calculated easily in the case of three-dimensional problems. However,
calculation of
C(s) can be done
using a rigid body motion criterion.
alternatively by
CHAPTER III NUMERICAL ANALYSIS
Introduction In chapter II,
the boundary integral equation required
for the boundary element method
was derived.
The equation
2.34 is reproduced here for further development. In equation 3.1,
the
displacements at a
point s
on the
boundary are
related to displacements and tractions over the boundary and body forces over the volume: C(s) u^(s) + r p*j(s,q) Uj(q) dr(q)=
ru*j(s,q) p^(q)dr(q) (3.1 )
where
p^.(s,q)
and
and displacements,
u^.(s,q) are the fundamental tractions and
p.(q) and
and displacements on the boundary. tion
' j , ' either the
In each cartesian direc-
displacement
u.(q) are
precribed.
and
on the boundary,
0.5
u.(q) are the tractions
C(s) is equal
p.(q) or
the traction
to 1.0
in the domain
if the boundary
is smooth
as
mentioned before. Because the extremely
analytical
difficult,
integration of equation 3.1 is
a numerical approach
obtain a solution of the equation. in a three dimensional domain, 21
To solve
is necessary to this equation
the boundary which becomes a
22 surface
is discretized
using surface elements.
variations of tractions and displacements over can be obtained through The order of
kept
the
same
be the same in
each element
the use of interpolation functions.
the interpolation functions
and tractions can
Differeni:
this
for displacements
or different,
research
for
but they are
computational
and
programming efficiency. Using
Gaussian
integrations collection
are of
quardrature
performed
boundary
point in one element.
formulas,
sequentially
elements Then
the
for
over
tions on the
equation,
boundary
boundary
displacements at an internal point are necessary,
the entire
conditions
are
A X = B,
and
unknown displacements
are determined.
surface
a particular source
applied to form a system equation of the type by solving this
the
The
and trac-
stresses
calculated
and
wherever
using equations which are discussed later.
Discretization For a three dimensional
solid mechanics problem,
boundary is a surface and it can be divided into
the
a discrete
number of surface elements. In this resesarch, these surface elements are assumed to be plane quadrilateral elements with four nodes, which are further assumed to be on one plane. For
a typical
element
boundary values of u. and
as
shown
in
Figure
7,
the
p. defining the displacement
and
23 tractions
on the boundary are assumed
interpolation functions u. and
p. can
to vary according to
within the element.
be assumed
The values of
to vary depending on the
of degrees of freedom that the analyst
would like to imple-
ment in the model.
(x
number
31 ^32 ^33^
(^41 ^42 ^43^^
'^^U "^12 ^13^ (x 21 ""22 ""23^
Figure 7: Quadrilateral Boundary Element
24 Consider the case of boundary values of u. and discrete
o. on the '3
J
surface element given by an interpolation function
such that: u . =
3k
^j ^
k
jk Pk
(3.2)
where u^ and p, are the nodal values of displacements and tractions at the node n (n=l,2,3,4). Simply by choosing the appropriate interpolation function the
$,, or 4/., , 3k 3k'
surface
of
quadratic, etc.
the
u. and p. can have any 3 3 element
such
as
constant,
These functions are standard
functions similar
variation
on
linear,
interpolation
to those used in the conventional
finite
element formulation. Supposing
that for
the discretization
of the body,
M
elements on F are considered, and applying the equation 3.1 on the whole boundary with equation 3.2, we get: M ^ ^ C(s) u.(s) + Z { p..(s,m) ^.,(m) dF(m)} u,(m) "• m=1 J F "-^ ^^ ^ m M . (s,m) ^., (m) dF(m) } p!J(m) m=i 1=1
Ji J F_
'^
^^
"
m where LM
is the total number of boundary elements, and
F is the m-th boundary element, m
(3.3)
25 In
equation 3.3,
integrals
on the discrete surfaces.
are replaced by integrals
For the constant element in which
tractions and displacements are
constant over each element,
the equation 3.3 becomes
^
r
*
C(s) u.(s) + 2 {
^
p..(s,m) dF(m)}
m=1 JF M = 2 { m=l
m F
u.(m)
^^
^
u.,(s,m) dF(m)} ^^
p.(m) ^
(3.4)
For a domain with 'M' nodes, the equation 3.4, in matrix form,
becomes CU + fiU = GP.
If H = fi + C I, then the whole
equation becomes HU =: GP.
(3.5)
Equation 3.5 can be written where 3M comes
from the
as equation 3.6
number of
node multiplied by the number P are global vectors containing
in matrix form
degrees of freedom
of elements
M. • Here, U and
the displacements and trac-
tions on the boundary. ^ H H H
H
l.rl
H 1,2
H
2,1
H 2,2
H
3,1
3M,1
r
1,3
1 ,3M
u1
2,3
2,3M
u.
H 3,2
H 3,3
3,3M
u.
H
H
3M,2
3M,3
per
H
u 3M
3M,3iyi )
26
'1,1
'1,2
1,3
'1 ,3M
'2,1
'2,2
'2,3
'2,3M
'3,1
'3,2
'3,3
'3,3M (3.6)
'3M,1
^3M,2
Note.that tractions
^3M,3
M^ values of displacements and
(M^ + M^ = 3M) are
hence in the U and P vectors, may be
'3M,3M
gathered
into
known
3M M-, values of
on the boundary,
there are 3M unknowns,
a left-hand
side
vector X.
and which After
reordering the equations, we obtain, AX = B where A is
a square
vector X is tions.
matrix
formed by the
of 3M, unknown
The contributions of
included in vector B. yield
(3.7)
all remaining
the
fully
populated,
and
displacements and tracprescribed
Equation 3.7 may
values
now be
solved
are to
unknown displacements and tractions on
the boundary. Determination of Internal Displacement and Stresses To find the internal displacements discrete point equation.
within the domain,
and stresses at any
we can use the following
27 °VHn ki3 = -^a4 - ^ 1 - 2 ^ ) [A,.^3. r ,3. +A,k:3. r ,i. - A.13. r ,,K' ] + B r . r . r ,] ,1 ,3 ,k^
1 A r 4aiT(1-v)
(3.8) ^ '
. (1-2V) (6n^r^.r_. . n.A.^. n.A.^)
- n-4v)n^A..l
m=!l
,^J(,.,,
(3.9)
m=l
where D,ki3 ..
and
S,ki3 .. are the contributions of the k-th direction to a.. at internal point due to trac13
tions and displacements,
respectively.
a.. is the stress tensor at the discrate 13
point, and
u (m) is in i-axes at the .m-th boundary element, n r . 3r / 3x. 1
/I
ci= 1 , 3= 2, Y= 4
for two dimensional cases, and
a= 2, 3= 3, y= 5
for three dimensional cases.
Numerical Integration In assumed
this section, to
the shape of the surface boundary is
be a quadrilateral
on a plane
with four nodes
28 as shown
in Figure 8.
The quantities to be integrated in equation 3.4 are: p,, dF
and
/ .
4
/ s' dF JF.^1} 3
.3
«
0
-1 •1
•2
-1
Figure 8: Quadrilateral Element in a Dimensionless Coordinate System
29 *
where
u..(s,m) 13
discussed all
*
and
p..(5,m) are the fundamental ^^13
before.
In order
these functions must
coordinates ( r\ ,r\^)
is such
in Figure 8.
to perform
the integration,
transformed
to dimensionless
which are used to transform the quadri-
lateral into a square. ( n., , 12)
be
solutions
The set of dimensionless
that they vary from
Hera, the coordinate of
-1
coordinates
to
+1 as shown
the nodal points are
also interpolated such a s : X. =
N. X..
1
3
where
i = 1,2,3, r
j 1
f
3 =1,2,3,4
'
J
'
or x^ = N^ X^, . N , X^^ . N3 X j ^ - N4 X ,, X2 = N^ X,2 * "2 =^22 * ^
=^32 * ^4 ^ 42
'^•^'>
X3 = N, X,3 * N2 X23 * N3 X33 . N^ X 43 where N.1. are N^ = \
i'^ - n.^ ) (1 - n2)
N2 = ^ (1 + n^ ) (1 - 02) N3- = ^ (1 + n^ ) (1 N4 = ^ (1 - n^)(i
+02) .
n^)
and X.. is the cartesian coordinate of
the node ' j ' in the
31
'i' direction. Since terms of
the H,,
interpolation H^ coordinates,
the elements of area
dF
from
functions
are
expressed in
it is necessary to transform cartesian coordinates to the
30 n.^ ,
ri2
coordinates.
A
differential of area
dF
will
be
given by
dF =
dn•i-
1
->•
1^
3 3x.
3x.
3ni
3rii 1
3r|'1
3x.
3x-
3x.
3r|^
3n
3ri.
3x.
(3.12)
dn^ dri2
dn-
^
->
= g-li + 923 + g3k
dn^ dTi2 where 3x.
3x.
3x.
3x.
3rii
3n
3ri-i
3ri
3x.
3x.
3x.
3x.
3r|,
3r|
3n.
3n
3x.
3x.
3x-
3x,
3n-|
3ri
3n.
3n
2 ^
1 2 2
^
|G| is the Jacobian relating the elements of area in the two systems of coordinates. For the quadrilateral element. 3N
3x
1 3ri-
r-^(N.X. . ) 3 n -j 1 1T
3N
1
3n
=^11^
= r^(N.X,^) 3n-J i i 2
=
3n-,
1
=^21^ —
3N.
3N.
3x.
3^
X
12
3N^
3N
dT]^
1
=^ 31""
3N. X. 22
30^
sTT
X
41
3N X. "32
3rii
X '42
31 ^^3
3N2
9N^
3
3N2
3 ^ - 37T{^^i^i3^ = 3n7 '1 ^n"- J^ 3x.
3
9N
^22^
3N^
37i^ ^33'" T^
3N
3N
3N
3n:
3 X. 3r|2 31
21
^43
3ri2
X
41
3N. 3N. 3N ^^2 3 ^^1 -^'7 -'7 3 ^2 = 3^i^i^i2^ _ ^ ^3712 X 22^ _ + ^3n2 X _ ^ ^3n2 X"42 -M2 X x^ = a^r«n2 X **12^ ""32^
3x
3N.
3N
3N.
9n2 " 3n^^i^i3^ = 3TT^ ^13* 3n^ ^23* TT
3N
^33* ITT ^ 43
in matrix form we can write, 3N.
3N.
3N.
3N
3T?^ Sn"^ 3n"^
3n
3n
3n
3n
3x.
3N.
3N.
3N.
3N
3x
^2
3x_ 3x.^
3x_ 3x.^ ^2
3n
^2
3n.
3n
X1 1 X 21
X 22
X 23
X
X
X
31
3n
X 41
3N.
3N.
3N
3N^
X
3^ 3N.
3N.
3N.
3N^
3n"
3^
3^
3^ ;
1 2 X1 3
11
32
33
42
X 43
V
X1 2 X 1 3
X 21
X 22
'23
X 31
X 32
X 33
X
X
X ^'43
41
42
(3.13)
J
Substituting Jacobian
equation 3.13
| G | can be calculated.
^ C(s) u (s)+ I
i
m=l
{
m=l
= 2 {
JF
.m
r
into
equation 3.12,
the
Then from equation 3.4,
* p(s,m)
J Fm "-^
u.(s,m) 13
|G| dri.(m) dn..(m)} u.(m)
^
| G | dri,(m) dr|-,(m)} p. (m) ' ^ J
^ (3.14)
32 Replacing the integrals by summations again, M K ^ C(s) u (s) + Z { Z p..(s,m) |G|, (m) W, (m)} u.(m) i m=1 k=l ^^ K K 3 M K = 2 { Z m=1 k=1 where K
^ u..(s,m) |G|, (m) W, (m)} p,(m) ^^ K K 3
(3.15)
is the number of Gaussian integral points, and
W (m) is the weighting factor of m-th element, and | G L (m) is the Jacobian of m-th element. From the application of equation 3.15 to all M boundary elements, arises.
a
final system
of 3M
equations
(equation 3.7)
CHAPTER IV EXAiMPLE PROBLEMS Introduction The boundary integral equation was derived in and a numerical technique for
formulating
Chapter II
the equations in
discrete form has been described in Chapter III.
A boundary
element computer program using the Fortran language has been developed
to
solve
three
dimensional
problems where the material is and
homogeneous.
applied
to
linear, elastic,
The developed
solve the
following
continuum
a horizontal traction, embedded
foundation
subjected to
subjected
problems:
isotropic, will be
(1) an elastic
normal forces;
2) a
to a vertical traction,
and a moment on the surface;
in a semi-infinite
prescribed
continuum
computer program
solid in the form of cube subjected to semi-infinite
solid
displacements.
elastic
(3) an
continuum
The results
from
this program, are compared to those from analytical solutions or those from
the finite element method.
For each problem,
the entire surface of the continuum is disretized as surface boundary elements.
The required input data for the program
include: 1. Geometry A. Nodal numbers and B. Element numbers of surface elements 33
34 2. Material Properties A. Shear modulus and B. Poisson's ratio 3. Prescribed
boundary
values
in the form of surface
tractions or surface displacements for each element 4. Coordinates of internal points
where the
tractions
or displacements are desired. Example Problem 1; Prescribed Traction on the Top Surface of Large Elastic Soil The semi-infinite region to be solved by the discretization technique
requires
a finite boundary
at some
length
fairly far away from the region of applied tractions. classical solution, defining
the
established research, times
accurate
certain dimensions are established
finite boundary differently
for
a semi-infinite
the loaded
stresses and
for
and these lengths are to different
continuum
length as shown
results,
Using
be
problem.
In this
is dimensioned
as
in Figure 9.
5
For more
larger distances to the boundaries where
displacements are small
are necessary.
This
example deals with the prescribed tractions or moment on the boundary at the are
center of the top
surface and
the results
compared with known analytical solutions [11]. The discretizations of this example are shown in Figures
10
and 11.
elements
which
The continuum is consisting
discretized
of 57
elements
as 85
surface
at the top,
6
35
20
20
figure 9: Idealized Semi-Infinite ' Region Problem
36
20
20
iqure 10: Discretization of Example 1 Figur
37
^^'^^v^^ ^ • . : $ :
<J
^
^^i^ ^ ^
Figure 11: Discretization of Top Surface
38 elements for each side,
and 4 elements at the bottom.
total number of unknowns in this problem is 255 and,
The
hence,
the order of the G and H matrices is 255 X 255. Shear modulus used for this problem is 1000 psi and Poisson's ratio is 0.3. Loads are considered
to be applied at the center of the
top surface and are uniformly distributed over a square area of length 4 ft, as shown in Figure 11. Three cases of loading conditions for this problem are: 1. A uniformly distributed vertical load = 100 psf 2. A uniformly distributed horizontal load = 1OOpsf 3. A moment = 40 ft lbs applied by tractions on 4ftx4ft area The first two cases are easily incorporated in the analysis. For the third case, since the tractions are assumed constant over each element,
moment should be changed to two opposite
tractions as shown in Figure 12. Here, using three elements, a zero traction is applied
to the center element
and equal
and opposite tracions are applied to the outer elements. can
expect
that
the smaller
the distance
between
elements, the better simulation of the moment. varying stresses
are considered,
to get a better simulation, too. displacements
outer
If linearly
more elements are
needed
Normal components of
on the side surface and
bottom
We
the
surfaces are
39
'^t
fUR]]^
^2i:
^"'^Jifliiijj
imm b
mm
[ a 1^ b -mt
^ = a (a+b)
H =
Figure 12: assumed
q.
Applied Moment in Example 1
to be zero for the
boundary condition as shown
in
Figure 13. The vertical stresses beneath area due
to vertical
the corner of the
loaded
and horizontal tractions are given in
Table 1 and are plotted in Figures 14 and 15,
respectively,
where they are compared against Hell's solution [7]. Vertical stress at the Table 1.
same points
due to moment
The results are plotted
are shown
in Figure 16,
in
where the
result is compared with Giroud's solution [7]. From these results, it is shown that the boundary element solution for
a vertical traction is very good
for vertical
40 4 ' .• ,<
0
#
20
^
20' — (a) Application of Vertical Load
r—1 I
.
X
I
0
20
m
jw
Mr
m
k^
20 ' (b) Application of Horizontal Load Figure 13: Boundary Conditions and Applied Loads
41
Table 1 : Comparison of Vertical Stresses Between Analytical Solution and Boundary Element Method due to Surface Loading Vertical Stress
X3/a
Case 1 Vertical Load Hell's Solution B.E.M.
o^
(kips/ft^)
Case 2 Horizontal Load Holl's Solution B.E.M.
Case 3 Moment Giroud's B.E.M. Solution
0.2
.249
.312
.152
.140
.212
.246
0.4
.240
.261
.133
.115
.151
.189
0.5
.232
.242
.121
.105
.125
.159
0.6
.223
.226
.109
.096
.103
.132
0.8
.200
.198
.086
.077
.070
.090
1 .0
.175
.172
.067
.061
.047
.060
1 .2
.152
.149
.051
.048
.032
.041
1 .5
.121
.121
.035
.033
.023
.024
1 .6
.112
.113
.031
.031
.019
.020
1 .8
.097
.099
.024
.024
.016
.015
2.0
.084
.087
.019
.019
.012
.010
2.5
.060
.060
.011
.011
.008
.005 . ,,
—
_
42
Vertical Stress
0
0.1
0.2
o
(kips/ft ) 0.3
0.4-
0.8-
1.2-
1.6Depth (x3/a) 2.0-
2.4-
Roll's Solution B. E. iM.
2.8-
T^igure 14: Comparison of Vertical Stresses due to Vertical Traction
0
43
Vertical Stress
a
(kips/ft^
0.4 -
0.8-
1.2Depth (x3/a) 1.6-
2.0-
2.4-
2.8.-
Figure 15: Comparison of Vertical Stresse: due to Horizontal Load
44
Vertical Stress
a
(kips/ft^
0.4-
0.8-
1.2Depth (x^/a) 1 .6-
2.0-
2.^-
2.8-
Figure 16: Comparison of Vertical Stresses due to Moment
45 stresses at a distance of x-./a = 0.6 below side of the loaded area. have
about
point,
10 percent
For cases errors at
where 'a' is the
2 and 3,
the points
but as the number of elements
the results beyond
increases,
that
the error
decreases rapidly. Example Problem 2: Embedded Foundation With Prescribed Displacements on The Top Surface of Large Elastic Soil Here,
an embedded foundation such as a rigid footing on
the top surface is considered. problem are the same as
Dimensions of
in the example 1,
discretization of the top surface shown
in Figure 17.
elements,
this example
except that the
are slightly different as
The continuum is discretized
which consist of
with 87
59 elements at the top surface,
6 elements for each side, and 4 elements for the bottom side. Two cases of boundary conditions for this problem are: Case 1. A Vertical Displacement = 1 inch Case 2. A Horizontal Displacement = 1 inch Due to the unavailability of this problem, code
a finite element solution
(SAP V) is
discretization Figure 18.
an analytical solution for
employed
of the
to
finite
compare
using the the
element model
Fortran
results. is
The
shown
in
For the finite element discretization, 219 solid
block elements which have 8 nodes per element, and 320 nodes are
used.
The total
number of
unknowns
for the
finite
46
figure
17
D i s c r e t i z a t i o n of Top Surface of Example 2
47
Figure 13
: Discretization of Finite Element Model
48 element model is 960.
As the finite element method requires
compatibility of displacements at each node between elements, it is very
difficult to
use larger elements
away from the
loaded area to reduce the number of degrees of freedom. Since this is the surface,
a displacement boundary value problem at
special boundary
prescribe these displacements
elements
must be
used
in the SAP iv code.
boundary conditions on the sides and bottom in the previous example (see Figure 19). of top surface are shown there.
to
All the
are the same as The displacements
Results obtained from these
two methods are compared at: (1) x-/a = 0.625 (x ^= 2.5) for tractions and (2) at x^/a =1.25 (x
= 5.) for displacements,
as shown in Tables 3, 4 and Figures 20, 21, 22, and 23. From the results,
it is shown that the boundary element
solution agrees remarkably well
with SAP IV solution.
the displacement at internal points due displacement, good
the boundary element
agreement with those
The displacement loaded
area
calculated
using very
to applied vertical
solution
solution.
at any point away from the
the boundary e lement
accurately
shows fairly
of the finite element
and the stress
For
because
technique can be
of the nature
of the
solution using the fundamental solution. Example Problem 3: a Solid Cube (Uniformly Loaded on one Side Only) This simple problem
when solved
by the finite
element
49
23
7W
JmT
20 (a) Vertical Displacements
h0 ^
V
23
ifmjr
mm 20
(b) Horizontal Displacements
Figure 19: Boundary Conditions and Application of Displacements
50
-
•
t —
^-'
•
r^
AJ M-l
CO (0
a
Q)
•H
m m
^
Q)
s w
a m
u w
ro
<j\
^
ro ro
CN
,_
"^
vo
in
1
C •
s w•
M (d
o
CN
00
O
•«^
r**x>
CN 1
1
O 00
CN 00
00
in
ro
1
1
CN 1
o
CD r^ m fd (d - P N •H
CO
c\
"^
• CQ
0
u
CN
• ^
•H CN Q
U
4J
l^ CN
0 ffi
CN 00
cr> •«* in
00
o
o
T—
in 00
• ^
,—
r*a\
• ^
CN CN
CN
in
KO
o
o
o
vo
vo 1
CN 1
CN 1
^
o
'^
1
o 1
CN
fo
4J 4J
•H C V^ Q)
0) s
>
(U
u
M-i rd
-
•
0 r-i
T—
a c w 0 -H
>>i^
m Q
•H
M 0 (d -P
a B 0)
0 3 U 'O *. CM
0
s• w
•
r-
VO
r-
r-
r-
t—
00
r-
00
00
00
(N
V^
^
•^
CM
^
CN
TT CN
'
vo
CN
T—
CN
f^
m
m
00
in
T—
T—
vo
in
'^
r-
00 (N
r^
CN
in
CN
CN
r—
o o
O CN
o
T—
f
OQ
a m
•H
Q iH
Q) (d
to o
fd -P U -H M 0) >
•
in
CN
in
CN
s•
CN
in
T—
vo
H •
T—
CN
CN
.H
^
1
ro
><
in CN
o
in
o
o
in
r-»
in
CN
00 1
in 1
1
CN 1
1
o
in
o
in
r^
in
in CN
CN
ro
in
00
.
1
51
(U
en
fC
en
u1
d)
en en
tn 4J G) C 5-t (U 4J G w (U U •-( (d (T) i H
o
CU
•H en 4J •H
U C
a)
> ^ (d »+-i
( 0 —u 4J
U 0 0} >
c 0 -rH
M 0 fd -p
a fl> e 0 3 u ^ • •
o
0) i-
fr. CN <4-l
m
u -p CO
52
4J
00 CN
o
Q)
vi;
en fd
O 1
o
en o; en 4J
ress emen
t
c_->
•
CN
""•
X "" (_) •
o
o
u en •H
fd
0) j-i
-P
en
a
•H D 4J
u ^
CU ft > -^ iw
c
0
•H
o • o
en en
a
fd en
•
Q
en r e
mpar e to
1
CN
iH
r-j
1
CN
o
in fd
on o oriz
•
-P o
T—
0 3 U T!
o 00
•r-l
53
•H
^ •
T -
-'-'
s•
•
H • CQ
a
CO •H CN D
4J
(d en 4J Ti
r*
,_
T—
T—
p^
r—
^
vo vo o
r^
CT^ O
o
o
o
o
o
o
o
o
o
o
VO (Ts O
(N CN
00
^
^_
ro
1
O 1
O 1
o
1
CN
^ ^—
rsj 00
r^
«—
,—
o
o
o
1
fd -P
u 0c
B d) .H u (d
•
N •H J-I
•u H
S
•
w•
0 ffi
CUA->
cn ^1
•H
ro
d) 1—1
c c
iH
ro CN
cn fd
Q) fd
fd
T -
^
Q
(U > cn
iH
0
in
00
• ^
"^
^
T—
T—
o
o
vo
^
ro
T—
T—
T—
o
o
o
vo cr> o
o
o
o
o
o
cr>
1
[l^
1
00
1
1
1
"^
00
o ro
ro
4J
C
-p Q) o Q) BQ) •H fd
-P 3
;
U
r—
J-I T5 fd
^~'
Q) > cn
a
c
•
a cn
0 •H Q
r- -H
0 C arH
Q
fd
en rH 4J
•H
fd
CN
'^r
'^
ro
o vo ro
o
o
o
o
o
O
O
O
00
vo
in
•«^
in CN TT
vo o
t—
ro
00 00
in
ro
•^
CN
s•
CN
o ro
• PQ
o
o
• H
CT^ 00 CN
-*:J<
P4
o
in
00 in
CN
fd fd
u •u H
fd M N •H e -p J-I
4J J-I
c a (u 0 c U H
T—
T—
cn r-l
c 0
M
in
ro
iH
4J CO iw •H
0
<J^ T—
ro
o cr» ro
•
0
ffi
•
S
0) >
• ^
••sr
CN •^ in
o
o
o
o
o
o
o
o
o
in
o
o
' i *
• ^
o
o
in
in
ro
CN 1
o
o
CN
ro
• ^
vo
o
r—
a\
rH fd EH
X
^
vo 1
1
1
1
54
C
00
I o
tn
s •
•
w w
O •
s
-P c <-* "^ re?
•
•
§ b
•
Ci^ PQ .-'
CI
c^. > 0": •rH ^.^m^
\'
O •
1^
CN
X ^^ 0)
u
—'
.-d
O
C -P
c; S- - P
O •
a' G
o
• - '
>
-p fn
u
o .
O
•H
0) cn fd
-P
I—I c C (Ti Q)
Q
o
C F p ^J
«
CN
CD ' J
P -P t t c ,-^ o
C -P -H r j fd c •• fN CN 'J
I
o 00
1 o
r o
<^
CO
I
O
o
o
c CJ
^ -^
e
-p
OJ u
U-J
fd
OJ
,— O
I>
55
cn r -p ^ C 0-'
CD c; S 0 O y O -H fd ^r-{ C
an:
U) •H 0 C -P
--J (1) fd r CN C T? •p-i 0)
J-) en rn P -P fd
a' c n
>
~ 0 U-i Oi
0
en c:
4J
^^ C T3
0 c B en u
•H p fd Qi
e
a) u -p fd c H M a
0 ^ u fd •* r^ CN fl' P
(D c
Q
a
^
en
•H C
56 method yields exact solution.
But when the boundary element
method is employed to solve it, the results are exact,
even
though
this
the errors
problem,
in the
the boundary
solution.
results are
small.
element method will
And to demonstrate this,
For
not yield exact
this particular problem
is selected. A cube of dimensions of 2 inch on each side Figure
24,
is loaded
in a state
of uniaxial
applying 100 psi of normal traction on the top
as shown in tension by surface. The
cube is divided into 6, 12, or 2 4 surface elements, as shown in Figure 24.
If 6 elements are
used in a
problem,
there
will be a total number of 18 unknowns and the order of the G and H matrices will be 18x18. configuration is used,
When the 12 and
24
element
the order of the matrix will be 36 x
36 and 72 x 72 matrices, respectively. The results Table 4-
for this problem when v = 0.3
are given in
The internal stresses are calculated
at 2 points
namely, (.4, .4, .4) and (.6, .6, .6). Comparing reactions
the
results
with
the analytical solution,
at the bottom are found to be within 2 percentage
of the analytical
solution.
Maximum axial
which occur at the top surface, for 6 element
give
displacements,
17.6 percentage error
and 7.4 percentage error for 24 elements.
twelve element model gives
the most accurate
results.
A
57 Table 4: Results For Example 3 Subjected To Axial Tension Units : in, lb or psi N = Reaction at bottom ele. ( % error) Displacement at top ele. ( % error)
6
N = 12
N = 24
EXACT
100. ( 0.0)
102. ( 2.0)
100. ( 0.0)
100. ( 0.0)
.00122 (17.6)
.00130 (14.2)
.00137 ( 7.4)
.00148 ( 0.0)
94.60 ( 5.4)
96.47 ( 3.6)
94.29 ( 5.7)
100. ( 0.0)
94.55 ( 5.5)
96.18 ( 3.8)
94.29 ( 5.7)
100. ( 0.0)
Int. point 1 ( % error) Int. point 2 ( % error)
internal point-1:(.4, .4, .4) internal point-2:(.6, .6, .6)
53
(a)
1 1 1 1
1
1 1
1
— •^^ -
1
i
1
1
1
1 1
•
1
.....
(b) 6 elements
(c ) 12 e l e m e n ts
)
( d ) 2 4 e 1 e m e .T •:
Figure 24: Surface Elements Arrangements for Example 3
CHAPTER V SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS Summary In this research, of boundary program
element
for
technique
solving
three-dimensional
basic principles of the development are
for stresses
isotropic,
reviewed.
and displacements
homogeneous,
due to Kelvin is
employed
displacements and tractions are quadrilateral element. tested
for three
in a
linear,
elastic, and solid continuum is developed. solution
A computer
and
The fundamental
in this program.
The
assumed constant within the
A constant quadrilateral element is
dimensional
elastic problems and results
are presented. A 6-point Gaussian quadrature integration formula is used for the formulation of the equations. Three example problems are solved and results are compared with known solutions. Conclusions Following conclusions are made in this research. 1 . The boundary
element method
can be
applied to three
dimensional elastostatic problems accurately. 2. For the problems
dealing
loaded
in a relatively small
method
offers
analysis.
a very
with the area,
convenient
large domains and
the boundary
and accurate
element
method for
The convenience is in the input data preparation. 59
60 The method is engineering
applicable to many problems
where
large
mass
of
soil
in foundation
medium
is
to be
considered. 3. Use
of 4-point
Gaussian
integration
inaccuracy in the equations and hence
leads to
some
6-point integraton is
found to be more accurate. 4. The method
yield poor results
for problems where the
surface volume ratio is large. Recommendations Following research are made for future research area. 1 . The method can be used
very
efficiently
many soil-structure interaction problems. that the finite
element method be
for solving
It is recommended
used for structures
and
where the surface volume ratio is large. 2. For numerical Gaussian
integration
computations.
integration, can be
both 4-point and 6-point
used to reduce
the number of
The 6-point formula can be used for integra-
tion of a equation near the area of stress concentration and 4-point integration stress concentration.
formula
for all
other area
away from
REFERENCES 1.
C. A. Brebbia. The Boundary Element Method For Engineers. London: Pentech Press, 1978.
2.
J. c. F. Telles. The Boundary Element Method Applied to Inelastic Problems. New York. Springer-Verlag, 1983.
3.
C. V. G. Vallabhan, J. Sivakumar, and N. R. Radhakrishnan, "Application of Boundary Elements for Soil-Structure Interaction Problems," 5th International Conference on Boundary Elements, University of Southampton, Southampton, UK. July, 1984
4.
C. A. Brebbia, J. C. F. Telles, and L. C. Wrobel. Boundary Element Techniques. New York,. SpringerVerlag, 1984.
5.
R. K. Nakaguma. "Three Dimensional Elastostatics Using The Boundary Element Method." Doctoral Dissertation. Dept. Of Civil Engineering. The University Of Southhampton. England. 1979.
6.
p. K. Banerjee and R. Butterfield. Boundary Element Methods in Engineering Science. New York, McGraw-Hill Co., 1980.
7.
H. G. Poulous and E. H. Davis. Elastic Solutions for Soil and Rock Mechnics. New York. John Wiley And Sons. 1974.
8.
A. I. Lur'e, D. B. McVean, and J. R. M. Radok. Three Dimensional Problems of The Theory of Elasticity. New York. Interscience Publishers. 1964.
9.
C. A. Brebbia and S. Walker. Boundary Element Techniques in Engineering. England. Pentech Press. 1978.
10.
Arthur P. Boresi. Advanced Mechanics of Materials. 3rd ed. New York. John Wiley and Sons. 1978
11.
D. J. Dansen. "Elasticity Problems With Body Forces." Boundary Element Techniques in Computer Aided Engineering. 5th International Conference on Boundary Elements, University of Southampton, Southampton, London Sep. 1983. 61
62 12.
C. A. Brebbia and J. Doningnez. "Boundary Element Methods Versus Finite Elements." Boundary Element Techniques in Computer Aided Engineering. 5th International Conference on Boundary Elements, University of Southampton, Southampton, London Sep. 1983.
13.
J. C. F. Telles and C. A. Brebbia. "Boundary Element Solutions For Half Plane Problems." Boundary Element Techniques in Computer Aided Engineering. 5th International Conference on Boundary Elements, University of Southampton, Southampton, London Sep. 1983.
14.
C. A. Brebbia. "Simulation of Engineering Problems Using Boundary Elements." Boundary Element Techniques in Computer Aided Engineering. 5th International Conference on Boundary Elements, University of Southampton, Southampton, London Sep. 1983.
15.
-M. Van Laethern and E. Back. "The Use of Boundary Elements to Represent The Far Field in Soil-Structure Interaction." Boundary Element Techniques in Computer Aided Engineering. 5th International Conference on Boundary Elements, University of Southampton, Southampton, London Sep. 1983.
16.
S. P. Timoshenko and J. N. Goodier, Theory of Elasticity. 3rd ed.New York, McGraw-Hill Co., 1982
17.
A. E. H. Love, A Treatise on the mathematical Theory of Elasticity. Dover Publication, N.Y. 1944