CE4257

Linear Finite Element Analysis

Name: Somsak Swaddiwudhipong

Room: E1A 07-10
Tel: 6516-2173
Email: [email protected]
 Objectives
This module involves fundamentals of linear finite element
method (FEM) and its applications to solve complex physical
and engineering problems. The fundamental principles will
enable students to understand the behavior of each simple
finite element, and to be able to select the appropriate
elements for solving these problems. The fundamentals
covered will also facilitate further independent study on other
related topics in computational mechanics.

Intended Outcome:
You should be able to
 analyze linear physical problems using appropriate set of
finite elements (FE),
 study independently and engage in further study on other
linear FE procedures,
 apply the concepts to facilitate your work in Civil Engineering.
 Main Aims & Objectives
This module aims to equip civil engineering students with the ability to 
(1) model structures or physical problems, and 
(2) analyse and understand their response.

3
 Recommended Readings

1. Asghar Bhatti, M. Advance Topics in Finite Element

Analysis of Structures, Wiley, 2006, (First three chapters).
The book website at www.wiley.com/college/bhatti
contains many Mathematica and MATLAB Applications.
The text also covers nonlinear and other advanced topics
in FEM.
2. Bathe, K.J. Finite Element Procedures, Prentice Hall,
1996, (First five chapters). Two computer programs: (i)
STAP for static analysis, and (ii) Subroutine SSPACE for
eigen-solver are listed in Chapters 12 and 11,
respectively. Similar to reference #1, Bathe’s book also
covers nonlinear and other advanced topics in FEM.
3. Reddy, J.N. An Introduction to the Finite Element Method,
McGraw Hill, 2006. The text involves mainly linear topics
in FEM.
 Important Notes:

Applying FEM to solve engineering problems without knowing the behavior

of finite elements used is bad. It is, however, worse and much more
dangerous to adopt FEM to model the problems without understanding
their behavior and use the results indiscriminately. We have to
(i) know what each finite element can do, to adopt the appropriate
element type, mesh size & distribution, and
(ii) understand the behavior of the engineering problems, to be able
to model the problems correctly and get things right.

It is not the intention of this module to guide you to write a new FE program
or even subroutines for FE unless they are not available in commercial FE
programs.
 Notes on Teaching & Learning:

• Cover sufficient fundamentals  To read further, life long learning

• Facilitate learning: Difficult topics  Simple & easy to understand
Overall
• Focus and be aware of objectives
Individual
• Examples are not just for demonstration of procedure but results
discussed to cover finite element concepts for better understanding.
• Suggest a collaborative learning for deeper understanding.
• Lecture notes and references will be available on IVLE with facilities
provided for your interaction.
Date Topic

10/8 Introduction and basic requirements

17/8 1-D elements

24/8 Triangular element

31/8 Quadrilateral element

07/9 Axi-symmetric element and FE modeling

14/9 FE issues and Quiz#1 at 6 - 7 pm

 Assessment and Exercises

• Quiz #1 15%
• Quiz #2 10%
• Exercise 5+10%
• Final Exam 60% (half each portion)

Exercises are included in the notes to further enhance your understanding

if attempted. Pls spend time to find the solutions and study on the
approach used and the implication of the results obtained.
Collaborative discussion is normally effective and you are
encouraged to do so. Those who share ideas and knowledge with
others also benefit, as to explain concepts, we have to think and hence
gain deeper understanding.
Pls submit the answers a week after the topic is covered. Your solutions
will be checked and returned to you with feedback a week after that.
This also applies to the scripts of the first quiz.
1. Introduction and Basic Concepts:
1.0 Finite Element Method (FEM)
1.0.1 What Is FEM?
1.0.2 Common Terminology in FEM

1.1 A Demonstrative Simple Example

1.1.1 Direct Equilibrium Approach
1.1.2 Direct Stiffness Method
1.1.3 Energy Approach

1.2 Basic Concepts in Elasticity

1.2.1 Components of Stress and Equilibrium
1.2.2 Components of Strain and Compatibility Conditions
1.2.3 Constitutive (Stress-Strain) Relationship
1.2.4 Boundary Conditions

1.3 Applications of FEM

1.4 Concluding Remarks

1.0 Introduction:
We usually involve the following two approaches in solving physical
and engineering problems:

(1) Formulation of the physical and engineering processes; and

(2) Numerical simulation of mathematical models.

The formulation normally leads to a mathematical model, mostly in

the form of differential equations governing the engineering and
physical phenomena. The solution process is, however, much more
involved than the problem formulation.
Approximate, in particular, numerical simulations are often adopted to
solve the formulated mathematical governing differential equations.
Several methods including finite difference, weighted residual,
weak formulation, finite element, etc, have been employed by
many researchers and analysts in the solution processes. Finite
element method has so far emerged as one of the most popular
approaches to study systematically the above phenomena.
1.0.1 What is Finite Element Method?

Finite Element Method (FEM) is a numerical tool used to transform

the differential equations, ordinary or partial, governing physical or
engineering problems into a set of algebraic equations that can be
efficiently solved by computers.

FEM may also be viewed as the generalization of Matrix Structural

Analysis to cover 2-D and 3-D physical and engineering problems.
Principles and concepts adopted for Matrix Structural Analysis can be
extended to FEM.

FEM is one of the most popular approaches as

(i) it is a versatile tool,
(ii) it has a high level of computerization,
(iii) it can be easily adapted to accommodate changes in design and
(iv) the availability of computers.
Similarity of Governing Equations for Physical Problems

C
r
o
s
s
-
s
e
c
t
i
o
n
a
l
a
r
e
a
d  du  du
 aA   q  0; Q  a ; A
dx  dx  dx
Physical Differential Primary/ Modulus (a) Source
problems Equation/ secondary (q)
Constitutive law variables (u/Q)
Axial d  du  Displacement/ Young’s Force per
deformation  EA   q  0 Axial stress,  modulus, E unit length
of a bar dx  dx 
du
Hooke:   E
dx
Electric d  dV  Voltage, V/ Electric Electric
current  A   q  0 Electric charge conductivity, charge
dx  dx 
dV  supply
Ohm: Q  
dx
Heat d  dT  Temperature, T/ Thermal Heat
transfer  kA q 0 heat flux conductivity, source
dx  dx 
dT k
Fourier: Q   k
dx
Other similar problems include: Flow through porous media, Ion diffusion, etc.
1.0.2 Common Terminology in FEM:

Element: 1D:

2D:

3D:

Node: Location where adjacent finite elements share the same values of
variables.
Degree of Freedom: Independent nodal variables in finite element.
1.0.2 Common Terminology in FEM (cont)
Local reference (x’,y’) axes: y’
Reference axes in each element. y’
x’ x’
Some samples are shown.

x’

Global reference (x,y) axes:

Common coordinate system y
y’ x
for the whole problem.

Cartesian (x,y) and Natural (s,t)

coordinate systems! t
y
2 1

s

3 4
x
1.0.2 Common Terminology in FEM (cont)
Boundary conditions: Restraints that have to be satisfied at boundaries.
(boundary nodes for FE)

Essential boundary conditions impose conditions on kinematic variables,

namely, displacements and/or slopes.

Natural boundary conditions specify conditions on kinetic variables, i.e.,

forces and/or moments.
Mixed boundary conditions involve both kinetic and kinematic variables.

Stiffness matrix, K, when operates on nodal displacements, provides

nodal forces as in
Fi = Ku i
Restraint sufficiency: At least, the minimum restraints to prevent rigid body
motion have to be provided, otherwise the global stiffness matrix is singular
corresponding to un-controlled displacements in non-restrained directions.
1.1 A Demonstrative Simple Example:
This simple example is to demonstrate various approaches in the
formulation of stiffness (finite element) equations. Let consider
the simple assembly of three springs. If the two ends are held in
positions, find the displacement of the central rigid block and
forces in springs or rods.
(i) Direct equilibrium approach
(ii) Direct stiffness method
(iii) Applying energy principle

Fig 1.1.1 A Simple Structure with Spring or Rod

 Spring or Rod Element Stiffness Equations
Positive sign convention is adopted for
both nodal forces, f1x and f2x and nodal
displacements u1 and u2. We impose +f1x
u1 u2 (a) and +f2x in Figs (b) and (c) respectively later.

When u2=0, f1x=ku1

f1x When u1=0, f1x=-ku2
(b) Hence, if both u1 & u2 exist:
u1 u2
f1x=ku1-ku2=k(u1- u2)

f2x
(c) Fig 1.1.2 (a) A spring or rod element;
u1 u2 (b) and (c) Forces in spring free bodies.

Fig (b): f1x  k  u1  u2   ku1  ku2  f1x   k -k  u 1 

   
Fig (c): f 2 x  k  u2  u1    ku1  ku2  f 2 x   -k k  u 2 
(1.1.1)
For rod element, k=EA/L is positive definite.
E = Young’s modulus; A = cross-sectional area; L = length of rod
1.1.1 Direct Equilibrium Approach:
1 2 3 Positive forces in each spring are imposed
and followed by forces for FBD at each joint.

FBD 3
4
F 1x
 f 1x

f1x u2 u3
FBD 1 FBD 2
F 3x
 f 3x

P  f (el1)  f (el 2)  f (el 3)

2x 2x 2x

u1 u2

4
u2 for each spring is equal.

u2 u4

Fig 1.1.3 Equilibrium at each node F  f

4x 4x FBD 4
 Total Stiffness Equations::

Equilibrium F 1x
 f P  f (el1)  f (el 2)  f (el 3)
1x 2x 2x 2x

at nodes 1 to 4 (1.1.2)
FBD 1 to FBD 4: F 3x  f 3x F 4x  f 4x

Stiffness equations
Element 1: Element 2: Element 3:

f  k 2 u 2  k 2 u3 f  k3 u 2  k3 u 4
f  k1u1  k1u2 2x 2x
1x

f   k 2 u 2  k 2 u3 f   k3 u 2  k3 u 4
f   k1u1  k1u2 3x 4x
2x
(1.1.3)

Substituting Eq (1.1.3)  F 1x
 k -k1 0 0  u1 
   1
in Eq (1.1.2) gives:
P   -k1 k1  k2  k3 -k 2 -k3  u2 
   (1.1.4)
F 3x 
0 -k 2 k2 
0  u3 
   0 - k3 0
 
k 3  u 4 
F 4x
 Boundary Conditions and Solutions:
Boundary conditions: u 1  u 3  u 4  0 (1.1.5)

Imposing Eq (1.1.5) in Eq (1.1.4) and solving yield:

u2  P /(k1  k2  k3 ) (1.1.6)

In view of Eqs (1.1.4) to Eq (1.1.6) , the reactions are:

F  -k u  -k P /(k  k  k )
1x 1 2 1 1 2 3

2u2  -k 2 P /(k1  k 2  k3 )
F 2x
 -k (1.1.7)

3u2  -k3 P /(k1  k 2  k3 )

F 3x
 -k

Are negative values of Fix reasonable?

1.1.2 Direct Stiffness Method

     f  - k 3  u 2 
 f 1x   k1 -k1  u1   f 2x  k2 - k 2  u 2   2x  k3
         
   -k3 k 3  u 4 
 f   -k1 k1  u2   f   -k 2 k 2   u3   f 
4x
3x 
2x
(1.1.8)
Expanding all three equations,
e.g., Eq (1.1.8c) gives:
Assembling Eq (1.1.8) in
0  0 0 0 0 u  view of Eq(1.1.2) provides:
    1
f  0 k 0 - k  u   F 1x   k
2x  3 3 2 -k1 0 0  u1 
        1
0   0 0 0 0  u3   P   -k1 k1  k 2  k3 -k 2 -k3  u2 
   0 -k3 0 k3  u4     
f 4x  F 3x   0 -k 2 k2 0  u3 
   0 - k3 0

k3  u4 
Assembling process = F 4x
Setting up equilibrium (1.1.4)
 1.1.3 Energy Method:

Strain energy in a spring:

1
U  k (u2 -u1) 2 (1.1.9)
u1 u2 2
Positive sign
convention of
forces and
displacements

(1.1.10)
1 2 1 2 1 2
Total strain energy: U  k (u -u )  k (u - u )  k (u -u )
2 1 2 1 2 2 3 2 2 3 4 2
Potential energy: V  -F 1xu1 - Pu2 - F 3xu3 - F 4 xu4 (1.1.11)

Total potential energy:   U V (1.1.12)

 Minimization of Total Potential Energy
1 2 1 2 1 2 (1.1.13)
  U V  k1(u2 -u1)  k2 (u3 -u2 )  k3 (u4 -u2 ) -F 1xu1 - Pu2 - F 3xu3 - F 4 xu4
2 2 2
 
 -k (u -u )  F 1x  0  k (u -u ) - k (u -u ) - k (u -u )  P  0
1 2 1 2 3 2 3 4 2
u1 1 2 1
u2
 (1.1.14) 
 k (u -u )  F 3 x  0  k (u -u )  F 4 x  0
3 4 2
u3 2 3 2 u4

Assembling Eq (1.1.14) gives:

 F 1x   k -k1 0 0  u 1 
   1
 P   -k1 k1  k2  k3 -k 2 -k3  u 2 
    (1.1.4)
 F 3x   0 -k 2 k2 
0 u 3 
   0 - k3 0

k3  u 4 
F 4x
Assembling process = Setting up equilibrium equations
= Satisfying energy principle
 1.2 Basic Concepts in Elasticity
Most engineering materials possess to certain extent the elastic
properties. When the applied forces are below the elastic limit, the
materials will resume the original configuration after the removal of
loads. Primary variables are: Forces and Displacements while
Secondary parameters are: Stress and Strain.

The governing conditions involved are:

(a) Equilibrium Equations: Forces (stress resultants) in equilibrium,
(b) Constitutive Relationship: Relating stress to strain,
(c) Compatibility Requirements and Boundary Conditions: Strain-
displacement relationship.

For a well posed boundary value problem, the number of

dependent variables (unknowns) should be equal to the number
of governing equations and the number of boundary conditions
equals the number of integration constants that is the same as
the order of differentiation in the governing differential equations.
 1.2.1 Components of Stress and Equilibrium
Equations: dx

x
 zz

Normal stress components:  zy

(x ,y ,z )

 xx ,  yy ,  zz dy  zx
dz
 xx

Shear stress components:  xy

y z

 xy   yx ,  yz   zy ,  zx   xz
yx

 yy  yz  xz

Fig 1.2.1 Positive Stress Components

Each force applied on a body produces stress, which is force per
unit area. Consider a small cube in the body, there are 3 components
of stress on each surface as partly shown in Fig 1.2.1. There are 3
sets of stress on 3 pairs of opposite surfaces constituting 9 stress
components. Applying 3 moment equations show that shear stress
tensor is symmetrical and hence,

 xy   yx  yz   zy  zx   xz (1.2.1)
 Stress and Sign Convention

The remaining 6 stress components are 3 normal stress and 3 shear

stress components.

The Timoshenko’s sign convention for stress is adopted.

Positive stress is the stress in positive coordinate direction on the
positive surface where the surface normal is in the positive
coordinate direction.
On the negative surface, positive stress is the stress in negative
coordinate direction.

The subscripts x, y, z and 1, 2, 3 are, herein, inter-changable to

facilitate the writing of tensor notation using Einstein summation
convention when necessary.
 Equilibrium Equations:

For clarity, only stress components

in x-direction are shown. bx

Body forces in x, y, z directions are

bx, by and bz respectively.

 Fx  0 : Fig 1.2.2 Free Body of Differential Cube

 xx  yx 
dx  dydz   dy  dzdx   zx dz  dxdy   bx dxdydz  0
x y z
Divide by common volume dxdydz, we get
S
i
m
i
l
a
r
l
y
,
 xx  yx  zx  Fy  0 &  Fz  0 :
   bx  0
x y z  xy  yy  zy
or in tensor notation with Einstein    by  0
summation convention, x y z
 ji  xz  yz

 bi  0 (1.2.2)   zz  bz  0
x j x y z
1.2.2 Components of Strain and Compatibility
Conditions:
For clarity, consider a 2-D case: B’
Displacements of A to A’ are u and v
in x- and y-directions respectively.
B
u
u dx  u C’
A ' C ' AC x u
 xx    ;
AC dx x A’
A ' B ' AB v C
 yy   ; (1.2.3) A
AB y
u
A’B’ is inclined to AB by an angle
y
v Fig 1.2.3 Plane element deforms
A’C’ is inclined to AC by an angle
from square to rhombic
x
Engineering    u  v  ;
 
 
xy
shear strain:  y x  We ignore higher order terms
1  u v  in these strain components.
Physicist’s  xy     ;
2  y x 
shear strain:
 Components of Strain in 3-D
Corresponding to stress components, there are 9 strain components but
only 6 independent strain components:
u v w
 xx  ;  yy  ;
y
 zz  ;
x z (1.2.4)
1  u v  1  v w  1  w u 
 xy    ;  yz     ;  zx     ;
2  y x  2  z y  2  x z 

or in tensor notation,

1  ui u j 
 ij     (where u1 = u, u2 = v, u3 = w, x1 = x, x2 = y and x3 = z)
2  x j xi 
(1.2.5)

 xy   yx ;  yz   zy ;  zx   xz
Note that
 Compatibility of Strains
Compatibility of Strains require that: displacement components must be
single valued continuous functions (otherwise, there is a breakage or
overlapping). This is satisified if strain relations are enforced.
Differentiating Eq. (1.2.3) gives Compatibility Conditions for 2-D problems.
 2 xx   yy   xy
2 2

  ; (1.2.6)
y 2 x 2 xy
Compatibility Conditions for 3-D problems are obtained from Eq. (1.2.4).
There are 6 compatibility equations for 3-D cases.
 2 xx   yy   xy
2 2
 2 xx    yz  xz  xy 
  ; 2     ;
y 2
x 2
xy yz x  x y z 
 2 yy  2 zz   yz
2
 2 yy    yz  xz  xy 
  ; 2     ; (1.2.7)
z 2
y 2
yz xz y  x y z 

 2 zz  2 xx  2 xz  2 zz    yz  xz  xy 
  ; 2     
x 2
z 2
xz xy z  x y z 
 1.2.3 Constitutive (Stress-Strain) Relationship:
For linear isotropic elasticity,
1 1
 xx  [ xx  ( yy   zz )]  xy   xy
E 2G
1 1
 yy  [ yy  ( zz   xx )] are uncoupled  yz   yz (1.2.8)
E E
2G
1 G 1
 zz  [ zz  ( xx   yy )] 2(1   )  zx   zx
E 2G
1
 xx   1/ E  / E  / E 0 0 0   xx   xy  2 xy  G  xy
     
 yy    / E 1/ E  / E 0 0  0  yy 
 zz    / E  / E 1/ E 0 0 0   zz  (1.2.9)
   

 xy   0 0 0 1/ G 0 0   xy 
   0 0 0 0 1/ G 0   yz 
  
yz
 
 zx   0 0 0 0 0 1/ G   zx 
 Constitutive (Stress-Strain) Relationship:
E
Solving Eq (1.2.9) and noting that G  , we get:
2(1   )
2G
 xx 
1  2
  xx   yy   zz   2G xx  xy  2G xy  G xy

2G
 yy 
1  2
 
 xx   yy   zz  2G yy  yz  2G yz  G yz (1.2.10)

2G
 zz 
1  2
 
 xx   yy   zz  2G zz  zx  2G zx  G zx

2G (1.2.11)
or in tensor notation,  ij   mm  2G ij  ij
1  2
ij = Kronecker delta function = 1 for i=j, otherwise = 0

Collectively, in matrix form, σ  Dε (1.2.12)

D is the tangent stiffness

 Appropriate 2-D Models:
Depending on the structure geometries and loading conditions, it is not
always necessary to solve a problem using a full 3-D model which is resource
demanding. With negligible loss in accuracy, it is often much faster to solve
problems using a 2-D model whenever feasible. There are 3 distinct 2-D
models that can be considered, namely, (a) plane stress, (b) plane strain,
and (c) axisymmetric.

Plane stress – stress in the third dimension vanishes, such as that in a thin
plate. (stress across thickness is negligible, zz  0 ).

Plane strain – strain (not stress) along the third direction is zero, such as
long structures (retaining wall, etc) or those with very thick geometries,
 zz  0 but  zz is normally non-zero.
Axisymmetric – solids of revolution subjected to axially symmetric loading.
All dependent variables including displacements, stresses, strains, forces,
bending & twisting moments, etc are independent of angular coordinate (θ).
 Problems Appropriate for 2-D Models:

Construction Load
q=10 kPa
125
q= 5 kPa

(a) plane stress

120

115

110

105

100

Reduced Level (m)

95

90

85

80

75

70

65

60

55
-55 -50 -45 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 35 40 45 50

Distance (m)

(c) axisymmetric (b) plane strain

Take advantage of axisymmetry and cyclic symmetry whenever possible.

 Axisymmetry: FE Simulation of Rigid Conical Indentation
Finer mesh at high
stress gradient region

The size of each element for

Conical
the finest mesh is 20x20 nm
Indentation

2-D vs 3-D:
Let us compare the sizes of 3-D and 2-D or axisymetric FE problems to be
solved. Let each side needs, say 50 nodes.
For a 3-D case, we have 125000 nodes with 3 DoF’s per node, i.e., 375000 DoF’s.
2-D or axisym case involves 2500 nodes with 2 DoF’s per node, i.e., 5000 DoF’s.
The latter involves a mere 1.33% of DoF’s of the former. Solving the equations of
the latter case takes significantly less than 1% of that of the former.
 Constitutive (Stress-Strain) Relationship
Eq (1.2.10):
 xx  1    0 0   xx 
0
      yy 
 yy
 1   0 0 0
 
 zz  E    1  0 0 0   zz 
     

 xy  1   1  2   0 0 0 (1  2 ) / 2 0 0   xy 
   0 0 0 0 (1  2 ) / 2 0   
  yz
   yz 
 zx   0 0 0 0 0 (1  2 ) / 2   zx 
 
(1.2.12)
σ  Dε
When out-of-plane stress vanishes, out-of-plane strain
normally survives. Solving for  zz from Eq (1.2.12) (c)
and substituting in Eq (1.2.12) (a) and (b), we get:
Plane stress   xx   1  0    xx 
  E      (1.2.13)
 
 yy  (1  2 )   1 0   yy 
  (1  ) / 2    xy 
Plane strain   xy   0 0

  xx   1   0    xx 
  E    
 
 yy  1   (1  2 )   1  0   yy 
(1.2.14)
     0 0 (1  2 ) / 2    xy 
 xy 
 General 3-D Elasticity Problems:
General 3-D elasticity problems are governed by the following 15 equations
involving 15 unknowns: 6 stress and 6 strain components and 3 displacements.
The problems can thus be solved.
u 1  u v 
 xx  yx  zx  xx  ;     ;
   bx  0 x
xy
2  y x 
x y z
(1.2.2) (1.2.4) 1  v w 
 xy  yy  zy v     ;
   by  0   ; yz
2  z y 

yy
x y z y
w 1  w u 
 xz   zz  zz  ;     
   bz  0  z 
zx
yz
z 2  x
x y z
 xx  1    0 0 0   xx 
      yy 
 yy
 1    0 0 0  
 zz  E    1  0 0 0   zz 
    

 xy  1   1  2   0 0 0 (1  2 ) / 2 0 0   xy 
   0 0 0 0 (1  2 ) / 2 0   
  yz
   yz 
 zx   0 0 0 0 0 (1  2 ) / 2   zx 
(1.2.12)
 Plane Problems in Elasticity
Plane problems in elasticity are governed by the following 8 equations
involving 8 unknowns: 3 stress and 3 strain components and 2 displacement
components. The problems can thus be solved.
 xx  yx u 1  u v 
  bx  0  xx  ;  xy     ;
x 2  y x 
x y
(1.2.2)
 xy  yy v (1.2.3)
  by  0  yy  ;
y
x y
Plane stress   xx   1  0    xx 
  E    
 
 yy  (1  2 )   1 0   yy  (1.2.13)
  (1  ) / 2    xy 
or plane strain   xy   0 0

  xx   1   0    xx 
  E    
 
 yy  1   (1  2 )   1  0 (1.2.14)
  yy 
     0 0 (1  2 ) / 2    xy 
 xy 
 1.2.4 Boundary Conditions:
Either force or displacement boundary conditions have to be specified on a
boundary for any well-posed problems. Only one (either force or displacement)
is permitted to be imposed at a location. Let’s consider force boundary
conditions.

Fig 1.2.4 Free body of tetrahedron

n= unit normal vector to surface dΓ
d 1 , d  2 , d 3 are areas of the three adjacent surfaces.

d 1  nx d  d  2  n y d  d  3  nz d 
1.2.4 Boundary Conditions
Force boundary conditions:

F x  0: df1 -  11 d 1 -  21 d  2 -  31 d 3  0
t1 d    11 nx d    21 n y d    31 nz d 
I
n
c
l
u
d
i
n
g

,  Fy  0 &  Fz  0 :

or in tensor notation ti d    ji n j d  (1.2.15)

or in matrix form σ Tnd   td   df

n= unit normal vector to surface d
Imposing displacement boundary conditions is normally straight forward.
 Example 1: Introduction, Rod Elasticity:
Example Ex1: A rod of length L=100 mm is fixed at the left end.
The cross sectional area A* ( x)  10(1  0.009 x) mm2 and the Young’s
modulus, E=200 kN/mm2, where x is in mm. The rod is subject to
uniformly distributed force of 0.1 kN/mm and a load P of 20 kN at
the free end, both in x-direction as shown in Fig. Ex1. Establish
the expressions for displacement, u, and strain,  , as functions of
x along the rod.
x

P
E, A*(x)
L

Fig. Ex1 Non-uniform rod under uniformly

distributed force and load P in x-direction
 x
Example 1: Rod Elasticity F
F 20  0.1100  x  30  0.1x
   L
A 10 1  0.009 x  10 1  0.009 x 
 30 0.1x
   x’
E 2000 1  0.009 x  2000 1  0.009 x 


x
30 0.1x 
u     dx P
0 2000 1  0.009 x  2000 1  0.009 x   Free body diagram
 30  1  0.1  x 1 
   ln 1  0.009 x      ln 1  0.009 x    C
 2000  0.009  0.009 
2
 2000  0.009 
Apply boundary condition to find C: u  x  0   0; C  0

 30  1  0.1  x 1 
u    ln 1  0.009 x      ln 1  0.009 x  
 0.009 
2
 2000  0.009  2000  0.009 
u  x  L   2.97 mm dx 1
 ax  b  a ln  ax  b 
xdx x b
 ax  b a  a 2 ln  ax  b 

 1.3 Applications of FEM
The following few figures demonstrate the applications of FEM in
engineering problems:

• Analysis and design of tall buildings including State Tower

(Silom Precious) project in Bangkok, Thailand.
• Analysis and design of the cable-stayed roof of Yishun
stadium, Singapore. Basic concept of modeling a truss structure
with hundreds of members into a space frame comprising tens of
thousands of members simplifies tremendously the finite
element simulation of the large-DoF structure without affecting
the accuracy of the results for design purposes.
• Finite element was used in the simulation of a large movement
of excavation support systems including the movement of
soils in the application of spud-can piles for offshore structures.
FE is also commonly used in the analysis and design of various
offshore structures under severe environments.
State Tower, Bangkok, Thailand:
FE Application in Basement Excavation
 Yishun Stadium Roof
under Construction

FE Model
“Water Cube” During Construction
Structural Model
Structural Model
 1.3 Applications of FEM (cont)

• Finite element was used in the simulation of a large

movement of excavation support systems including the
movement of soils in the application of spud-can piles for
offshore structures. FE is also commonly used in the
analysis and design of various offshore structures under
severe environments.
• Finite element method is essential for the analysis and
design of structures with complex geometry, such as the
design of the roof of the Sydney Opera House in Australia.
• Finite element is used to simulate the drop test and car
crash experiment including severe impact problems.
• Penetration, indentation, fracture and the development of
natural forms and patterns such as those of leaves and
fruits may be simulated by the finite element method.
Large Movement of CBP Wall during Excavation:
 FEM Mesh and Material properties
Construction Load
q=10 kPa
Contiguous bored Pile Wall (New Piles)
125 (900 mm dia. to depth of 10 m,
q= 5 kPa
600 mm dia. to depth of 28 m)
120
N=5, E= 6 MN/m2,
115 c= 5 kPa,  = 24o Old contiguous bored pile

110 1st Floor Slab

N=8, E= 9.6 MN/m2,
105 c= 5 kPa,  = 25o
100
N=15-25, E=18-30 MN/m2,
Reduced Level (m)

95 c= 7-10 kPa,  = 26-28o

90
Piles
N=40, E= 48 MN/m2,
85 c= 10 kPa,  =30o
80

75
N=60-100,
E= 72 – 120 MN/m2,
70 c= 10 kPa,  = 31-32o
65

60

55
-55 -50 -45 -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30 35 40 45 50

Distance (m)
Spudcan Supporting Mobile Jackup Unit and
Spudcan-Pile Interaction

Spudcan

Image Credit: Keppel Offshore

and Marine Ltd
 Convergence Studies and Effect
of Spudcan Penetration Rates
ABAQUS Number of Number of
0 2 4 6
Penetration Resistance, MN
8 10 12 14 16 18
EC3D8R Elements DoF’s in Model
0

2
Mesh 1 55610 179430
4
Mesh 2 133239 420018
6
Penetration Depth, m

8 Mesh 3 355298 1101534

10
Mesh 4 456884 1411983
12

14 Mesh 1 Penetration Resistance, MN

Mesh 2
0 2 4 6 8 10 12 14 16 18
16
Mesh 3 0
Mesh 4
18
2

Penetration Depth, m
6

10
Typical spudcan penetration
12
rate is in the order of 1m/hr.
14
Penetration Rates - 0.1672 m/s
16 Penetration Rates - 0.3344 m/s
Penetration Rates - 1.672 m/s
18
Max Deviation
of < 2%
 Verification of Results
Penetration Resistance, kN
0 5000 10000 15000 20000
0

4
Penetration Depth, m

10

12
Eulerian FE Analysis

14
Hossain and Randolph (2009) -
Shallow Penetration
16
Hossain and Randolph (2009) -
Deep Penetration
18

Tho, et al, Int. J. Geomechanics (2012) 64.

 Spudcan Penetration on
Sand over Soft Clay Soil Profile
Initially Void
3m
Sand
Rigid Spudcan Spudcan Diameter = 12 m
’ (Sand) = 30
E’ (Sand) = 20000 kN/m2
Soft Clay
34m Poisson’s ratio (Sand) = 0.3
cu (Clay) = 10.7 kPa
Eu (Clay)= 300 x cu
Poisson’s ratio (Clay) = 0.495

Tho, et al, Can. Geotech, (2013) 595.

Experimental Observation and Numerical Simulation

(Image Credit Dr Tho K K, NUS)

(Image Credit Dr Teh Kar Lu, NUS)

Test Set-up for In-plane Bending of Brace:

Grouting

brace

Saddle
chord support

1000-ton test rig Roller

Chord: 508x15.1 (d0 x t0)
Brace: 407x22.9 (d1 x t1)
Test vs. FE Results
FE Model for Drop Test
Car Crash Experiment:
Car Crash Experiment
1.4 Concluding Remarks
1. An analyst should (i) know what each finite element can do,
adopt the appropriate element type, mesh size & distribution for
the problem to be solved, and (ii) understand the behavior of
the engineering problems;
2. FE equations may be formulated either via equilibrium, direct
stiffness method, or energy principle.
3. The governing conditions in elasticity involves: Equilibrium
equations, constitutive relationship, strain-displacement
relationship and compatibility requirements.
4. Whenever appropriate, 3-D problems should be modeled and
solved via 2-D or 1-D analyses for efficiency and ease of
interpretation of results.
5. Examples of FE models for various engineering applications are
demonstrated.